Pauli Valkjärvi Integrated Data Acquisition for State-of-the-Art Large-Bore Engine Test Cell Vaasa 2022 School of Technology and Innovations Master’s thesis Smart Energy 2 VAASAN YLIOPISTO Tekniikan ja innovaatiojohtamisen yksikkö Tekijä: Pauli Valkjärvi Tutkielman nimi: Integroitu tiedonkeruujärjestelmä suurisylinterisen moottorin tutki- mukseen Tutkinto: Diplomi-insinöörin tutkinto Koulutusohjelma: Smart Energy Työn valvoja: Apulaisprofessori Maciej Mikulski Ohjaaja: Tohtoriopiskelija Michaela Hissa Vuosi: 2022 Sivuja: 118 TIIVISTELMÄ: Polttomoottoreilla tulee olemaan tärkeä rooli hiilidioksidipäästöjen vähentämisessä ja kestävän voimansiirtojärjestelmän toteuttamisessa merenkulkualalla. Merenkulkualan sähköistäminen on nykyisellään hankalaa valtavan energiantarpeen vuoksi. Sen vuoksi polttomoottorit tulevat pysymään lähitulevaisuudessakin laivojen tärkeimpänä voimanlähteenä. Uutta palamismenetel- mää, reaktiivisuudella hallittua puristussytytystä (RCCI), voidaan pitää yhtenä lupaavista poltto- moottoriteknologioista, jonka avulla voidaan samanaikaisesti saavuttaa erittäin alhaiset typen oksidi- ja hiukkaspäästöt, sekä korkea hyötysuhde. Vaikka konseptia on kehitetty pitkään, sovel- tuvuutta isosylinterisissä moottoreissa ei ole osoitettu julkisesti. Tämän opinnäytetyön tavoitteena oli suunnitella ja toteuttaa uusi tiedonkeruujärjestelmä isosylinteriseen RCCI -testipenkkiin Vaasan yliopiston VEBIC moottorilaboratoriossa osana Clean Propulsion Technologies (CPT) -projektin työpakettia 3. Testipenkki instrumentoitiin uusilla an- tureilla, analysaattoreilla ja tiedonkeruulaitteilla. Järjestelmän rakentamiseen tarvittavat lait- teet hankittiin ja laiteasennukset sekä sähköliitännät toteutettiin. Lisäksi mahdollistettiin uuteen järjestelmään soveltuva tiedon tallennusprosessi. Järjestelmän suorituskyvyn arvioimiseksi suo- ritettiin osittainen järjestelmätesti, koska moottoria ei ollut mahdollista käynnistää vielä opin- näytetyön aikana. Osittaisen järjestelmätestin tulokset osoittivat, että uusi tiedonkeruujärjestelmä kykenee mit- taamaan korkealla näytteenottotaajuudella ja tallentamaan mittaukset kampiakselin asennon suhteen. Opinnäytetyössä suunniteltu ja toteutettu järjestelmä tarjosi useita parannuksia edel- liseen järjestelmään verrattuna. Käytettävissä olevien korkean näytteenottotaajuuden kanavien lukumäärä kasvoi 8:sta 16:een ja järjestelmä tarjoaa joustavamman reaaliaikaisen tiedon jälki- käsittelyn. Päivitetty järjestelmä tarjoaa myös merkittävän parannuksen datan integroimiseen, koska nopeat ja hitaat mittaukset voidaan tallentaa samaan tiedostoon. Välittömien järjestel- män parannusten lisäksi uusi järjestelmä kykenee mukautumaan tulevaisuuden tarpeiden mu- kaan. AVAINSANAT: Tiedonkeruu, mittaaminen, anturit, RCCI, polttoanalyysi, moottoritestisolu, in- tegrointi. 3 UNIVERSITY OF VAASA School of Technology and Innovations Author: Pauli Valkjärvi Title of the thesis: Integrated Data Acquisition for State-of-the-Art Large-Bore Engine Test Cell Degree: Master of Science in Technology Degree Programme: Smart Energy Supervisor: Associate Professor Maciej Mikulski Instructor: Doctoral student Michaela Hissa Year: 2022 Pages: 118 ABSTRACT: Internal combustion engines will have an important role on a road to decarbonization and a sus- tainable powertrain system in the maritime sector. Electrification of the maritime sector is cur- rently difficult due to its excessive energy density demand. Therefore, internal combustion en- gines will remain a primary power source for ships in the near future. A novel combustion con- cept, reactivity-controlled compression ignition (RCCI), can be seen as one of the promising com- bustion technologies that enables simultaneous ultra-low NOx and soot emissions, as well as high thermal efficiency. Although the concept has been developed for a long time, its feasibility for large-bore engine applications has not been publicly demonstrated. The goal of this thesis was to design and implement a new data acquisition system for the large- bore RCCI test bench in University of Vaasa’s VEBIC engine laboratory, as part of the Clean Pro- pulsion Technologies (CPT) project’s work package 3, novel combustion and advanced aftertreat- ment. The test bench was instrumented with new sensors, analyzers and data acquisition hard- ware. Devices required to build the system were acquired and device installations, as well as electrical connections were established and supervised. Additionally, data storing workflow, suit- able for the new system, was developed. In order to validate the system performance, a partial system test was carried out due to the inability to start up the engine during the thesis. The results from the partial system test proved that the new data acquisition system is able to measure high sampling frequency signals and record them in reference to crank angle. The sys- tem that was designed and implemented in the thesis provided several improvements when compared to the previous system. The number of available high sample frequency channels in- creased from 8 to 16 and the system provides more flexible real-time post-processing capabilities. The upgraded system also provides a significant improvement in integration, as the high-speed and low-speed measurements can be recorded into a single file. In addition to immediate system improvements, the new system is able to expand according to future requirements of the test bench. KEYWORDS: Data acquisition, measurement, sensors, RCCI, combustion analysis, engine test cell, integration. 4 Foreword This Master’s thesis was done for the University of Vaasa as a part of the CPT project’s work package 3. The work done in the thesis was meant as an enabler for the demon- stration of a medium-speed engine utilizing low-temperature RCCI. Firstly, I want to thank my instructor and team leader, Doctoral student Michaela Hissa for her valuable guidance throughout the thesis. I want to also thank my supervisor, As- sociate Professor Maciej Mikulski for supervising the work and providing useful com- ments and suggestions. I am also grateful for the help and support I received from the VEBIC engine laboratory staff during the implementation phase of the work. Lastly, I am forever grateful for the support I received from my family and friends throughout my studies. 5 Contents 1 Introduction 14 1.1 Background 14 1.2 Problem formulation 16 1.3 Thesis structure 17 2 Literature review 19 2.1 State-of-the-art in engine testing 19 2.1.1 Development trends of internal combustion engines 19 2.1.2 Engine testing in general 20 2.1.3 Modern test bench and control development methodologies 22 2.2 Test bench transducer and analyzer technologies 25 2.2.1 Angular position and speed measurement 25 2.2.2 Linear displacement measurement 26 2.2.3 Pressure measurement 27 2.2.4 Temperature measurement 29 2.2.5 Smart sensors 31 2.2.6 Fuel consumption and flow measurements 32 2.2.7 Exhaust gas emission measurement and analysis 35 2.3 Signal conditioning and acquisition 39 2.3.1 Signal amplification 40 2.3.2 Measurement error reduction 42 2.3.3 Analog-to-digital conversion 43 2.3.4 Test cell data communication technologies 45 2.3.5 Sampling 47 2.4 Combustion analysis and post-processing 48 2.4.1 Combustion measurements 49 2.4.2 Top dead center detection 50 2.4.3 Zero-level correction 51 2.4.4 Mean effective pressure 53 6 2.4.5 Heat release rate 55 2.4.6 Gas exchange analysis 57 3 Objects and methods 59 3.1 VEBIC W4L20 engine test cell 59 3.1.1 Engine overview 60 3.1.2 Fuel system 60 3.1.3 Intake and exhaust system 61 3.1.4 Cooling and lubrication system 61 3.1.5 Control system 62 3.2 RCCI engine test bench instrumentation 63 3.2.1 Electro-hydraulic valve actuation (EHVA) 65 3.2.2 Crank angle encoder 65 3.2.3 High sampling frequency measurement instruments 66 3.2.4 Low sampling frequency measurement instruments 67 3.2.5 Fuel consumption measurement system 67 3.2.6 Emission measurement system 68 3.3 Data acquisition 70 3.3.1 High-frequency data acquisition system 70 3.3.2 Low-frequency data acquisition system 71 3.4 System design and validation methodology 72 3.4.1 System design 73 3.4.2 Performance validation 73 3.4.3 System prechecking 74 3.4.4 Reference measurements 75 4 System design and implementation 77 4.1 System requirements 77 4.2 Device installations and connections 78 4.2.1 Dewesoft data acquisition hardware 79 4.2.2 Installing the piezoresistive amplifiers 82 4.2.3 Electrical signal connections to data acquisition hardware 83 7 4.3 Data acquisition system implementation 85 4.3.1 Analog input channel configuration 85 4.3.2 Counter channel configuration for encoder 87 4.3.3 CEA module configuration and parametrization 88 4.3.4 Connecting the Modbus TCP/IP interface 92 4.3.5 Monitoring interface 92 4.3.6 Data storing 93 5 System consolidation and validation 95 5.1 System consolidation 95 5.2 System performance validation 96 5.2.1 High sampling frequency measurement system evaluation 96 5.2.2 The EHVA unit test conditions 96 5.2.3 Results of the EHVA unit test 97 5.3 Prospects for system expansion 99 6 Conclusions 102 7 Summary and outlook 105 References 106 Appendices 113 Appendix 1. Specification of Dewesoft SIRIUS STG module. 113 Appendix 2. Part of the Modbus channel list displayed in DewesoftX. 115 Appendix 3. Example text-file export from DewesoftX. 116 Appendix 4. Analog signal error calculations in the high sampling frequency measurement chain. 117 Appendix 5. Measured analog signal curves during the EHVA unit test representing the valve position and control signals in time domain. 118 8 Figures Figure 1. Overview of the engine research and development process. 21 Figure 2. Principles of MIL (a) and RCP (b) methods in engine control development (Isermann, 2014). 23 Figure 3. Principle of HIL simulation for engine control development (Isermann, 2014). 24 Figure 4. Principle of EIL simulation setup. 24 Figure 5. Measurement principles of gear tooth sensor and optical encoder (Merker et al., 2012). 26 Figure 6. Piezoelectric principle (Kistler, 2020). 29 Figure 7. Structure of direct weighing gravimetric fuel gauge (Martyr & Plint, 2012). 33 Figure 8. Coriolis mass flow meter (Martyr & Plint, 2012). 34 Figure 9. Structure of chemiluminescence detector measuring NO concentration (Nakamura & Adachi, 2013). 37 Figure 10. Structure of flame ionization detector measuring THC concentration (Nakamura & Adachi, 2013). 37 Figure 11. Principle of data acquisition process (adapted from Martyr & Plint, 2012). 40 Figure 12. Basic charge amplifier circuit for a piezoelectric pressure transducer (Rogers, 2010). 41 Figure 13. Most common types of errors resulting from ADCs (Measurement Computing, 2012). 45 Figure 14. VEBIC W4L20 engine test cell. 59 Figure 15. Instrumentation of the RCCI cylinder. 63 Figure 16. Structure of the intended exhaust gas measurement system of the RCCI cylinder. 70 Figure 17. The layout of the W4L20 engine test cell data acquisition system prior to modifications. 72 Figure 18. Topology of Dewesoft SIRIUS AI channel (Dewesoft, 2022a). 80 Figure 19. Dewesoft Super Counter architecture (Smith, 2021). 81 9 Figure 20. Dewesoft SIRIUS data acquisition hardware, installed in the VEBIC laboratory within the scope of the present work. 82 Figure 21. Channel setup window from DewesoftX. 86 Figure 22. Counter sensor editor window in DewesoftX. 87 Figure 23. Selected parameters and additional channels assigned for the W4L20 RCCI test bench. 89 Figure 24. Result definition settings from DewesoftX showing the available and enabled basic calculations. 91 Figure 25. The designed folder structure for storing measurement data. 94 Figure 26. The layout of the W4L20 engine test cell data acquisition system designed in the thesis. 95 Figure 27. Position and control signal curves of intake (blue) and exhaust (red) valves with respect to crank angle. 98 Tables Table 1. Wärtsilä 4L20 engine specifications. 60 Table 2. Specification of measurement instruments used in the RCCI cylinder. 64 Table 3. Comparison of Kistler KiBox and Dewesoft SIRIUS data acquisition systems. 71 Table 4. Electrical signal connections to Dewesoft AI channels. 84 10 Abbreviations ACT Advanced combustion technology ADC Analog-to-digital converter AI Analog input AO Analog output BDC Bottom dead center BMEP Break mean effective pressure CAD Crank angle degree CAN Controller area network CDC Conventional diesel combustion CEA Combustion engine analysis CLD Chemiluminescence detector CO Carbon monoxide CO2 Carbon dioxide COVIMEP Covariance of IMEP CPT Clean Propulsion Technologies CSV Comma-separated values DI Digital input ECU Engine control unit EHVA Electro-hydraulic valve actuation EIL Engine-in-the-loop EOI End of injection FID Flame ionization detector FMEP Friction mean effective pressure FSO Full-scale output FTIR Fourier transform infrared H2O Water vapor HIL Hardware-in-the-loop HT High temperature 11 IMEPg Gross indicated mean effective pressure IMEPn Net indicated mean effective pressure I/O Input/output LT Low temperature LTC Low-temperature combustion MIL Model-in-the-loop NDIR Nondispersive infrared N2 Nitrogen NOx Nitrogen oxides O2 Oxygen PLC Programmable logic controller PM Particulate matter PMD Paramagnetic detection PMEP Pumping mean effective pressure PRT Platinum resistance thermometer RCCI Reactivity-controlled compression ignition RCP Rapid control prototyping rpm Revolutions per minute RTD Resistance temperature detector RTU Remote terminal unit SiO2 Silicon dioxide SOI Start of injection SOx Sulfur oxides TCP/IP Transmission control protocol/internet protocol TDC Top dead center TEDS Transducer electronic data sheet THC Total hydrocarbons TTL Transistor-transistor logic USB Universal serial bus VDC Voltage direct current 12 VEBIC Vaasa Energy Business Innovation Center ZrO2 Zirconium dioxide Greek Letters γ Polytropic coefficient ε Emissivity θ Crank angle ρv Fluid density σ Stefan-Boltzmann constant σs Standard deviation Other Symbols Acyl Area of the cylinder wall Aeff Effective valve area Cc Cable capacitance Cr Amplifier feedback capacitor capacitance Ct Sensor capacitance E Measurement error f Frequency hc Heat transfer coefficient m Mass n Crank angle interval resolution N Engine rotational speed p Pressure Δp Zero-level correction offset value pn Cylinder pressure at reference point Pb Brake power Pr Pressure ratio 13 Q Piezoelectric charge Qgross Gross heat release rate Qht Heat loss Qnet Net heat release rate R Gas constant Ri Cable resistance Rt Amplifier resistance t Time Tg Temperature of in-cylinder gas Tin Temperature of intake air Tw Temperature of cylinder wall Uis Reference isentropic fluid velocity Uo Amplifier output voltage V Volume Vn Cylinder volume at reference point Wb Brake work Wi Indicated work xi Measured variable Δxi Uncertainty of measured variable xj Systematic error component Y Calculated variable ΔY Uncertainty of calculated variable 14 1 Introduction 1.1 Background The special report on global warming by the Intergovernmental Panel on Climate Change states that if global warming were to exceed 1.5 °C compared to the pre-industrial era, there would be significant harm to the ecosystem and societies (Masson-Delmotte et al., 2018). In 2019, the shipping industry’s global carbon dioxide (CO2) emissions accounted for roughly 2.5% of overall global CO2 emissions (Teter et al., 2020). In addition to this, the shipping industry has a significant impact on human health and air quality near ports due to the high rate of air pollutants like nitrogen oxides (NOx), particulate matter (PM) and sulfur oxides (SOx) (Chen et al., 2021). Without major development steps toward more sustainable solutions, the ratio of emissions resulting from shipping compared to other transportation subsectors will increase in the future (Mestemaker et al., 2020). Paris agreement (United Nations, 2015) together with tightening emission legislation drives engine and powertrain development towards higher efficiency and lower emis- sions. In the heavy-duty road transport sector, the introduction of Euro VI emission standard (European Union, 2009) presented a significant NOx emission reduction for die- sel engines in order to improve the quality of air. The reduction of NOx emissions needs to be carried out without sacrificing the inherent advantages of diesel engines or adding to other emissions. The limit of heavy-duty vehicle NOx emissions in Euro VI compared to Euro V was reduced by 80% in steady-state testing and 77% in transient testing (Wil- liams & Minjares, 2016). Emission legislation is one of the key drivers in the engine de- velopment process, and progress in the automotive segment inspires emission mitiga- tion in other transport segments. Diesel engines, commonly used in ships and heavy-duty vehicles, have a high thermal efficiency because of the high compression ratio and low heat rejection. However, they 15 suffer from large NOx and PM emissions due to high in-cylinder temperatures and diffu- sion combustion inherent to the diesel combustion process (Wei & Geng, 2015). Ad- vanced combustion technology (ACT) is a concept which aims to mitigate emissions while improving engine efficiency. According to Shim et al. (2020), novel combustion technologies allow for reducing NOx and PM emissions by controlling the maximum in- cylinder temperature, as well as the local equivalence ratio. Shim et al. (2020) stated, that single-fueled ACTs offer lower emissions. However, in addition to having a narrow operating range, there are difficulties in the control of the combustion phase and main- taining combustion stability. To overcome issues of single-fueled ACTs, researchers have investigated the possibilities of dual-fueled ACTs. According to Baškovič et al. (2022), RCCI has the highest potential among several low-temperature combustion (LTC) ACT concepts when it comes to fuel- flexibility and wide operating range. RCCI uses in-cylinder blending of two fuels with dif- ferent reactivities to control combustion, which is driven by chemical kinetics (Paykani et al., 2016). In dual-fuel RCCI, low reactivity fuel is supplied into the cylinder during the intake process through a port fuel injector to create a uniform mixture of air and fuel. High reactivity fuel is injected directly into the cylinder using a single or multiple injection strategies early enough to form a premixed charge inside the cylinder prior to auto-igni- tion (Li et al., 2017). Changing the ratio of these two fuels, as well as the injection timing enable efficient control over the ignition timing and combustion rate (Panda & Ramesh, 2022). With RCCI, it is possible to simultaneously achieve high thermal efficiency and low NOx and PM emissions (Krishnamoorthi et al., 2019). Kokjohn et al. (2011) demonstrated that when compared to a conventional diesel engine, RCCI provides improved control of the combustion process, higher efficiency and signif- icantly reduced NOx and PM emissions in the span of a wide engine operating range. In contrast to the benefits of RCCI, there are challenges to achieve high and low engine loads. During low load operation, RCCI engines tend to have an increased rate of total 16 hydrocarbons (THC) and carbon monoxide (CO) due to low temperatures inside the cyl- inder during combustion. Additionally, operating an RCCI engine at high loads can result in excessive pressure rise rates that could damage the engine (Li et al., 2017). Novel combustion technologies give new requirements to data acquisition system of an engine test facility. In contemporary engine research, test cells are complicated systems filled with specific machinery and instrumentation. Accurate and precise measurements need to be made in order to accurately control and capture the performance character- istics of the engine and related subsystems. A common data acquisition system com- prises of an instrument that produces a signal representing the physical phenomena, signal conditioning to protect the signal from corruption and ensure signal compatibility with the rest of the system, acquisition hardware for converting analog signals to digital format, and software that is used to record, analyze and store the digital values repre- senting the real phenomenon (Osman & Massoud, 2013). In order not to compromise the data produced in the test cell, all the devices connected to the measurement chain are desired to be integrated (Martyr & Plint, 2012). Instrumentation of modern engine test cell measurement chain includes a large variety of sensors, actuators and analyzers, each with their preparatory sensor output types and communication protocols (Asad et al., 2011). 1.2 Problem formulation In the CPT consortium (Clean Propulsion Technologies, 2021; Mikulski, 2021), the goal is to create a shared vision and sustainable business solutions in the marine and off-road segments in order to secure the global position of the Finnish powertrain industry. The goal will be achieved by developing innovative powertrain technologies that comply with emission regulations until 2035. Technological roadmap providing a consolidated plan in securing emission compliance for both segments until 2050 will achieve the goal in the long term. The CPT project consists of six work packages. In work package 3, novel com- bustion and advanced aftertreatment, the goal is to provide significant progress in the 17 technological roadmap for novel combustion technologies. This will be made possible by demonstrating a medium-speed engine that uses low-temperature RCCI. In order to build a large-bore RCCI marine engine platform as a part of the CPT project, many modifications are needed in the University of Vaasa’s VEBIC (Vaasa Energy Busi- ness Innovation Center) engine laboratory. In this regard, the large-bore engine test cell requires new data acquisition system and measurement workflow. The goal of this thesis is to design and implement a data acquisition system for large-bore state-of-the-art en- gine test bench. New sensors, analyzers and data acquisition hardware were studied and implemented to the current engine test cell. The objective of this research is to discover the requirements for the new data acquisition system in order to provide more fluent low-temperature combustion research for the RCCI engine. Based on the requirements, a new measurement system was designed and implemented into University of Vaasa’s large-bore engine test cell located in the VEBIC laboratory building. While realizing this objective, this work aims to answer the four following research questions: 1. What are the measurement system requirements to enable LTC research on a multi-cylinder engine? 2. How to integrate different measuring instruments into a common data acquisi- tion platform, bearing in mind their individual sampling rates and communication protocols? 3. How to carry out efficient post-processing of the measured quantities, that ena- bles insight to the in-cylinder phenomena? 4. How to effectively store large amount of measurement and analysis data coming from an engine test bench? 1.3 Thesis structure 18 This thesis consists of five main chapters which are literature review, object and methods, system design and implementation, system consolidation and validation, and conclu- sions. The literature review focuses on state-of-the-art engine testing and the main con- cepts around the engine test cell data acquisition systems. In addition, central post-pro- cessing methods and combustion analysis are discussed. The chapter provides insight regarding modern measurement technologies and analysis methodologies for conduct- ing modern combustion engine research. Ultimately, this gives a solid background for understanding the objects and methodologies used in this work. Objects used in this research, as well as system design and validation methods are introduced and discussed in chapter 3. That chapter provides detailed information regarding the status of the test cell prior to modifications and methods to design and build a robust test cell data acqui- sition system. chapter 4 focuses on describing the system design and implementation phase. There, configuring and adding new measurement devices to a common system, as well as the relevant functions of the system are explained in detail. Results regarding system consolidation and validation are discussed in chapter 5. The focus is on the over- all performance of the upgraded system and its adaptability for the future. The conclu- sions of this thesis are presented in chapter 6, which aims to give concrete answers to research questions. The final chapter of the thesis provides a summary. 19 2 Literature review 2.1 State-of-the-art in engine testing Modern engine testing involves the combination of different research and development methodologies for different purposes. Simulation models coupled together with the en- gine test bench offer a flexible and efficient platform for engine development. This type of setup requires accurate measurement of many different parameters that are used in the calibration and optimization of the engine control system. 2.1.1 Development trends of internal combustion engines The objective of engine development today is to produce engines with high efficiency and performance while having low emissions. Characteristics of modern diesel engines include a combustion chamber utilizing four valves, multiple common-rail injectors with high pressure of up to 2000 bar, advanced supercharging, controlled exhaust gas recir- culation, oxidation catalyst, and a high level of electronic engine management (Heywood, 2018). Development trends of diesel engines aim towards improvements in common- rail direct injection, turbocharging, exhaust aftertreatment systems, and modification of the combustion process (Isermann, 2014). Improving the combustion process and lowering emissions and noise with common-rail direct injection is possible with piezoelectric injectors that allow efficient use of different combinations of injection pulses (Wang et al., 2020). More efficient turbocharging is at- tained with regulated two-stage turbocharging or variable geometry turbocharging. These technologies allow high-pressure boosting, thus improving the power density and low-speed torque characteristics of the engine (Wang et al., 2020). According to Iser- 20 mann (2014), exhaust aftertreatment development is focusing on oxidation catalyst con- verters and particulate filtering in order to mitigate CO, THC, NOx and PM tailpipe emis- sions. For heavy-duty engines, selective catalytic reduction appears as an alternative af- tertreatment technology. With advanced combustion technologies, it is possible to sig- nificantly reduce emissions, while simultaneously having high thermal efficiency. ACTs related to diesel engine development listed by Wang et al. (2020) include homogenous charge compression ignition, dual fuel RCCI, and dual fuel highly premixed cool combus- tion. These combustion concepts require more sophisticated control compared to con- ventional diesel combustion (CDC). A good overview of these enabling control technol- ogies can be found in the work by Duan et al. (2021), including multi-pulse injection, fast thermal management, variable valve actuation, variable compression ratio, on-board fuel reforming, and many other control functions. 2.1.2 Engine testing in general The purpose of engine testing and research is to compare the performance of engines in different conditions and states. As outlined in the previous subsection, internal combus- tion engines are complex machines that require many auxiliary systems and support ser- vices in order to perform as desired. A facility designated for engine testing requires ad- vanced control and a data acquisition system to carry out the required test procedures (Atkins, 2009). Tightening regulations, especially emission legislation has considerably shaped the scope and methods of engine testing in the individual development stages. The first stage in engine development process is to gain an understanding of the engine performance at a fundamental level. At this level, performance is mainly influenced by the combustion event occurring inside the engine cylinder. Fundamental research is of- ten first conducted on a single-cylinder research engine test bench. The single-cylinder engine gives a possibility of finding out performance characteristics and refining certain designs before moving to multi-cylinder configurations (Martyr & Plint, 2012). The main 21 advantages of a single-cylinder setup are relatively low cost and effective implementa- tion. Optical versions of single-cylinder configurations usually have cylinder walls and piston crown made out of transparent material. This allows visual access into the com- bustion chamber by using a high-speed camera and enables visual observation of several in-cylinder phenomena, like combustion, flame growth and propagation, and injection- related events (Martyr & Plint, 2012). An example of an optical setup done on a medium- speed engine can be found in a publication from Merts et al. (2021). The authors con- ducted an optical study that enabled visualization of ignition and combustion events of conventional dual fuel and RCCI operation. After fundamental experimentation is done with a single-cylinder setup, more applied level research is required. As stated by Asad et al. (2011), single-cylinder engine test benches often require the sub-systems to be externally established and controlled. It was also noted that intake and exhaust flows can be highly pulsating in single-cylinder configurations, which may lead to uncertainty during the measurement campaign. To properly develop and calibrate the engine control system with accurate and reliable measurements from the engine on an applied level, multi-cylinder testing is necessary. The final phase in the engine development process is to optimize the operating param- eters and control functions, as well as get the certifications in a real-world test (Isermann, 2014). Figure 1 elaborates the engine research and development process discussed above and highlights the requirements for experimental setup in a specific development phase. Figure 1. Overview of the engine research and development process. 22 2.1.3 Modern test bench and control development methodologies According to Isermann (2014), a typical test bench consists of an internal combustion engine under investigation and a dynamometer used to convert the mechanical power generated by the engine into electrical energy. This setup is then equipped with specific instruments for control activities and required measurements. Depending on the appli- cation and desired operation profiles, different types of dynamometers are used. For example, eddy-current dynamometers are common in steady-state tests, while electric motors with grid coupling can be further applied in transient campaigns (Atkins, 2009). The main controllable parameters in the engine test bench are engine load and speed (Martyr & Plint, 2012). Aside from this, the test bench automation system regulates cool- ant, lubrication oil, fuels, and intake air. In addition, information is acquired from the measurement devices. Acquired information is then used in the implementation of safety and emergency procedures. Modern engine control parameters need to be cali- brated to be suitable in certain operating conditions while maintaining the desired per- formance and emission characteristics. This procedure known as engine mapping is usu- ally done in test benches capable of simulating real-world conditions, which helps to determine engine boundaries and optimal operating parameters in different circum- stances (Martyr & Plint, 2012). However, developments in computation capacity has en- abled efficient real-time calculation of combustion parameters. This in turn has led to development of engine control concepts such as closed-loop combustion control, which is crucial in RCCI to achieve stable combustion (Indrajuana et al., 2016). Modern engine development relies on the integration of simulations and modeling to- gether with physical testing. The early phase of engine control function and software development can be done with model-in-the-loop (MIL) simulations, where both engine and engine control unit (ECU) are simulated. This enables the design of new control func- tions before implementing them to an engine on a real test bench (Isermann, 2014). According to Isermann (2014), considerable time savings can be achieved with rapid con- trol prototyping (RCP) by testing new control functions with the help of a special real- 23 time computer in parallel with the ECU. These new control functions operate through ECU in a bypass mode utilizing its interfaces. RCP is possible if some control functions from a previously developed ECU are usable on an engine test bench to enable testing of new functions together with the real engine and ECU. Figure 2 visualizes the principle of MIL and RCP development methods. Figure 2. Principles of MIL (a) and RCP (b) methods in engine control development (Isermann, 2014). Jiang et al. (2009) say that hardware-in-the-loop (HIL) testing is commonly used meth- odology for control system prototyping, calibration, and validation in modern engine testing and development. HIL enables the calibration of ECU in parallel with the mechan- ical development of the engine. In the HIL methodology for engine control development, real ECU is used to operate a real-time engine model. The main advantage achieved in HIL testing is that the full operating range of ECU can be tested under real-time con- straints without the high cost associated with real engine tests, or possible risk of dam- age involved in the process of exploring the boundary conditions and abnormal opera- tion (Martyr & Plint, 2012). According to Isermann (2014), special electronic modules are used to simulate the necessary sensor signals and outputs from the ECU can be used to operate real actuators. Figure 3 demonstrates the principle idea behind HIL testing, where a real ECU is used to operate real components together with a simulated real- time engine model. 24 Figure 3. Principle of HIL simulation for engine control development (Isermann, 2014). Engine-in-the-loop (EIL) simulation is a specific form of HIL simulation, as stated by Jiang et al. (2009). The real component in the EIL simulation is the engine control system to- gether with the engine, while the simulated component is the application that the en- gine is coupled to. Figure 4 shows the principle of an EIL setup, where vehicle simulation is coupled to an engine test bench. Measured values from the test bench are inputs to the simulation. ECU controls the engine according to the simulation. Figure 4. Principle of EIL simulation setup. There are several advantages associated with EIL testing according to Jiang et al. (2009). It enables fast and efficient evaluation, verification, and debugging of engine control sys- tem at an early stage of development. Development of transient engine control and ini- tial calibration focusing on the vehicle system becomes possible prior to physical vehicle integration. Simulating the vehicle allows for a fast modification of its parameters in or- der to learn about their impact on engine performance and behavior. Generally, the EIL test bench provides consistent and repeatable test runs (Jiang et al., 2009). Klein et al. (2017) discuss the quality of EIL testing and say that it is dependent on the simulation models, parameter accuracy, and implementation with the test bench. They also point 25 out that despite the benefits of EIL, model integration with the test bench may require too many resources to justify the effort. 2.2 Test bench transducer and analyzer technologies Accuracy of the measurements in modern engine research and testing is important. Even though the measurement chain depends on the experimental setup, a large number of different parameters exhibiting dependency on each other need to be measured from internal combustion engines (Asad et al., 2011). There is a substantial number of differ- ent transducers and analyzers needed in order to enable required measurements during modern engine testing. 2.2.1 Angular position and speed measurement According to Rogers (2010), basic engine speed and phase measurements can be done using the inductive measurement principle where an inductive position sensor consist- ing of a permanent magnet and coil is mounted near the engine flywheel in a way it detects the change in magnetic flux due to passing gear teeth. Flux density is at its peak when the air gap between the tooth and the sensor is at its smallest, indicating that a tooth is positioned perpendicular to the sensor. This changing magnetic flux induces an alternating sine current in the coil that represent the engine speed and phase (Rogers, 2010). A digital Hall effect sensor can also be used similarly to generate a signal for the indication system (Merker et al., 2012). Gear tooth-based sensors are often used only in monitoring and speed measurement applications due to relatively low angle resolution for combustion measurements (Rogers, 2010). The use of a crank angle encoder is crucial for combustion measurement as it is used to provide a reference crank angle for the measurement system (Rogers, 2010). An angular 26 encoder is mounted to the end of the crankshaft and consists of a disk that spins to- gether with the crankshaft. Because the position of the encoder is used to indicate the position of the crankshaft, it is crucial to properly install the encoder (Martyr and Plint, 2012). The disk has light gates engraved in it and the encoder produces a pulse each time that light passes through a gate. Spinning movement of the disk results in a square wave signal. When the exact number of pulses that the encoder produces during one revolu- tion is known, pulse count and the frequency of the wave signal will determine the an- gular position and the speed of the crankshaft (Rogers, 2010). There is usually one addi- tional output signal that is used as a trigger signal for determining when the crankshaft has rotated a full revolution. This trigger pulse is generated with an additional light gate in the disk. Figure 5 below illustrates the working principle of the gear tooth sensor and optical encoder. Figure 5. Measurement principles of gear tooth sensor and optical encoder (Merker et al., 2012). 2.2.2 Linear displacement measurement Movement in one direction is called linear displacement. Displacement and position are measurements often needed in engine testing for determining the status and movement of actuators and valves (Martyr & Plint, 2012). The sensor that measures linear position or displacement has an output signal that represents the movement of the measured object related to the reference point. In engine research and development, there is often 27 a need for dynamic displacement measurements, such as injector needle lift and valve movement (Rogers, 2010). According to Martyr and Plint (2012), inductive transducers can be used in this type of measurement. The working principle of an inductive displace- ment transducer is based on changes in impedance between the coils of the sensor and a conductive object whose movement is being measured (National Instruments, 2021). Transducers that are based on the Hall effect are also used (Martyr and Plint, 2012). In a Hall effect transducer, a movement of a magnetic object causes a change in magnetic flux sensed by a Hall effect sensing element, which outputs a voltage signal proportional to the movement of the object (Rogers, 2010). 2.2.3 Pressure measurement As stated by Smith (2009), pressure can be determined as a force resulting from mole- cules of a fluid or gas hitting on a surface of a container. Pressure can be measured in several ways. Absolute pressure measurement means that the pressure is measured rel- ative to zero. Gauge pressure means that ambient air pressure is used as a zero reference. Differential pressure measurement is used as a difference between two points. There are typically two elements present in an electronic pressure transducer. One is the ele- ment that is used to sense the pressure, such as a diaphragm that converts the pressure into displacement. Another element is for converting the displacement into measurable electrical property. Measuring the pressure of different mediums gives insight into the operation of the en- gine. According to Martyr and Plint (2012), typical mediums in an engine test cell from which absolute pressure is measured include fuel, lubrication oil, coolant, intake air, and exhaust gas. Common transducer types used for absolute pressure measurements from the engine are strain gauge and a capacitive transducer. A strain gauge transducer con- sists of a diaphragm that deforms according to the applied pressure. Another component of a strain gauge transducer is a strain gauge made from conductive material whose elec- trical resistance varies according to the change in diaphragm geometry. Transducers that 28 have their strain gauge made out of semiconductor material are called piezoresistive transducers (Morris & Langari, 2012). Pressure value can be derived directly from the resistance of the strain gauge, typically by using a Wheatstone bridge circuit. The sensi- tivity of a strain gauge transducer depends on the material of the strain gauge and is quantified by the gauge factor that describes the change in resistance relative to applied strain (Rogers, 2010). Piezoresistive transducers have significantly higher gauge factors compared to conventional metal strain gauge transducers (Morris & Langari, 2012). A capacitive pressure transducer includes a capacitor as its diaphragm and is designed in a way that allows for the capacitor’s capacitance to vary according to the applied pres- sure (Smith, 2009). In modern combustion engine research and development, pressure transducers that are based on the piezoelectric effect are a choice of instrumentation for in-cylinder pressure measurements because they can produce in-cylinder pressure curves with great accu- racy and repeatability (Rogers, 2010). The piezoelectric effect refers to an electric charge that accumulates in certain materials, such as quartz (SiO2), when it is placed under stress. Piezoelectric pressure transducers have a piezoelectric crystal as an element that converts the displacement into an electrical signal. The magnitude of the produced charge depends on the change in pressure and requires an amplifier circuitry to produce a strong enough signal for the acquisition system (Kistler, 2020). Figure 6 visualizes the principle of piezoelectricity, where the load applied to a quartz crystal produces a directly proportional electrical charge into the measurement circuit. 29 Figure 6. Piezoelectric principle (Kistler, 2020). Piezoelectric crystal inside the pressure sensing element responds only to dynamic changes in pressure and are not suitable for absolute pressure measurements. Piezoe- lectric pressure transducers are relatively small in size, have a high natural frequency and large measuring range (Kistler, 2020). 2.2.4 Temperature measurement Temperature is a parameter that provides essential information about the condition of the engine. According to Smith (2009), temperature means the ability of an object to transfer heat. The temperature sensing element needs to be in thermal equilibrium with the object or substance and therefore, the fastest response is gained when the meas- urement devices are in direct contact with the object of measurement. Sometimes the temperature sensing element is installed inside a thermowell. Thermowell is used to protect the sensing element from forces and chemical effects induced by the medium. A disadvantage of using a thermowell is slower response time and reduced accuracy of temperature measurement (Smith, 2009). Almost every temperature that needs to be measured during engine testing typically remains in a steady state or do not change very 30 much during short periods of time (Martyr & Plint, 2012). However, temperatures, espe- cially exhaust gas temperatures, are almost never exactly stable and may vary greatly between different operating points. Averaging multiple measurements can reduce the problem of unstable gas flow resulting from pulsations coming from different cylinders. When moving from one operating point to another, it is needed to wait for the measured temperatures to stabilize in order to get reliable measurements when changing from one operating point to another. This reduces the error resulting from heat absorbed into en- gine surfaces and possible sensor shielding (Martyr & Plint, 2012). Thermocouples are well suited for most temperature measurements inside an engine test cell. Thermocouples of every type consist of two wires made out of different metals that form a closed circuit consisting of two junctions, one is for reference and the other in thermal equilibrium with the object. When a temperature difference occurs between the two junctions, a current begins to flow in the closed circuit according to the Seebeck effect (Smith, 2009). The temperature value is obtained by introducing a resistor with a large resistance into the circuit and reading the voltage that is produced. This voltage across the resistor is proportional to the temperature difference between the two junc- tions. In theory, the Seebeck effect occurs between any two dissimilar types of metal, but due to practical reasons, current flow in relation to the temperature difference must be repeatable, high enough and as linear as possible (Smith, 2009). This means that only few different metal pairs can be used in practice. The most common thermocouple types used in engine testing facilities are types T (copper-constantan), J (iron-constantan) and K (chromel-alumel) (Martyr & Plint, 2012). As stated by Smith (2009), main advantages of thermocouples are that they can be used at high temperatures up to 1700 °C, they can withstand vibration and shock, and they are cost-effective compared to other avail- able temperature measurement devices. Disadvantages compared to resistance temper- ature detectors (RTDs) are low signal level, required compensation of a reference junc- tion, lower accuracy, and deterioration over time resulting to drift. 31 When temperature measurement requires accuracy and reliability beyond the capabili- ties of thermocouples, platinum resistance thermometers (PRTs) can be used (Martyr & Plint, 2012). PRTs are resistance temperature detectors that use platinum as a resistor element. It is known that the resistance of metals increases with temperature. The prin- ciple of PRTs is the relationship between the temperature and the resistance of the re- sistor. This relationship is not exactly linear and usually the nonlinearity is greater at higher temperatures (Smith, 2009). Resistance is measured by running a small current through the resistor and measuring the voltage drop. Temperature can be derived from the relationship between the resistance and the temperature. The most used type of PRT is called 100-Ω platinum RTD, which has a resistance of exactly 100 Ω at 0 °C. Because the wiring between the resistor element and the resistance reading device can introduce major error to the measurement, often 3- or 4-wire systems are used to compensate resistance of the wires. According to Smith (2009), the use of platinum as the resistor element is justified by several arguments. PRTs have a high sensitivity, meaning that the resistance of platinum changes according to temperature relatively large amount com- pared to other metals. Platinum also has a good resistance against corrosion. PRTs have a relatively large temperature range even up to 850 °C and linearity between tempera- ture and resistance is decent across the whole usable temperature range (Martyr & Plint, 2012). PRTs have several advantages compared to thermocouples, such as better accu- racy, stability over time and the ability to handle possible electrical interference within a test facility. Engine temperature measurements are usually required to be measured from fluid and gas flow systems and from different mechanical components (Martyr & Plint, 2012). These include intake, exhaust, coolant, oil and fuel systems, as well as re- lated mechanical parts. In reality, the locations of the measurements may vary depend- ing on the engine application and the purpose of testing. 2.2.5 Smart sensors Sensors that have smart features embedded in them are called smart sensors. The Insti- tute of Electrical and Electronics Engineers proposed an international standard called 32 Transducer Electronic Data Sheet (TEDS) to standardize communication between any TEDS sensor and smart signal conditioning device regardless of the manufacturers. Ac- cording to Rogers (2010), the usage of TEDS depends heavily on the application area. Data contained in TEDS typically includes basic information, such as data about the sen- sor itself and information regarding calibration. TEDS is stored on a microchip that is usually embedded in the sensor itself or the connector and accessing the information requires the use of a specific connection. In addition to a typical connection carrying an analog signal from the transducer, a parallel connection that provides the signal condi- tioning system with a connection interface to the microchip is required (Rogers, 2010). A useful feature that usually comes with smart sensors and conditioning systems is the ability to automatically record the operating data of a sensor. This means that the oper- ating hours and cycles can be recorded in order to keep track of the planned regular checking and recalibration of the instruments (Rogers, 2010). An additional feature that has significant value in engine testing is the ability of amplifier circuitry to recognize the sensor. When it is possible to assign a sensor to a specific application, risks involving incorrect parametrization and connection of a wrong sensor are eliminated (Rogers, 2010). 2.2.6 Fuel consumption and flow measurements Fuel consumption is an important parameter to be measured during engine testing. Ac- curate measurement of fuel consumption enables research and development of efficient engines with reduced emissions. According to Martyr and Plint (2012), modern engine testing requires the actual and transient measurement of fuel consumption during dif- ferent test sequences. Modern engines usually incorporate fuel return strategies or fuel spillback to provide more accuracy, but in turn, increases the complexity of the engine test cell fuel consumption measurement system. In modern fuel measurement systems, it is not accurate to measure only the fuel that is being supplied to the engine. The sys- 33 tem needs to be able to measure the amount of fuel that enters the metering and con- ditioning system. This means that in order to mitigate possible sources of measurement error, the system must be able to control the pressure in the return fuel lines and remove possible vapor bubbles and heat energy that have formed while passing through the en- gine fuel system (Martyr & Plint, 2012). According to Martyr and Plint (2012), the two most usual types of fuel gauges used for liquid fuel measurement are volumetric and gravimetric gauges. A volumetric gauge measures the volume of the fuel that is consumed in a specific period. Volume can be measured from a container that has a known volume or from the flow through a flow- meter. Gravimetric gauges measure the mass of consumed fuel during a specific period. As illustrated in Figure 7 below, gravimetric fuel gauges consist of a container that is mounted on a load cell that measures the weight of the container. Fuel is supplied to the engine through this container and the mass of the supplied fuel can be measured against time. Return fuel from the engine injection system is also returned to this container through spillback. Figure 7. Structure of direct weighing gravimetric fuel gauge (Martyr & Plint, 2012). 34 The mass flow rate of the supplied fuel to the engine can be measured with mass flow meters. According to Martyr and Plint (2012), mass flow meters in the market are typi- cally Coriolis flow meters that utilize the Coriolis effect. Figure 8 shows the structure of the Coriolis effect mass flow meter. Figure 8. Coriolis mass flow meter (Martyr & Plint, 2012). From Figure 8 can be seen that a Coriolis meter consists of two parallel U-tubes that the fuel passes through. Electromagnets are used to make the tubes vibrate at a specific frequency. When there is mass flow present inside the tubes, Coriolis force will produce an additional vibration to the tubes that results in slight differences in the relative vibra- tion of the tubes. These differences are detected with sensors that measure the position, velocity, or acceleration of the U-tubes and produce a sinusoidal voltage signal repre- senting the mass flow of the fuel (Martyr & Plint, 2012). The advantage of the Coriolis effect mass flow meter is its ability to continuously and accurately measure many differ- ent kinds of liquid and gaseous substances. Air flow in the test cell can be measured with several kinds of methods such as venturi gas meter. The operating principle of a venturi meter is based on the pressure drop across a venturi inside the pipe that the air is flowing through (Martyr & Plint, 2012). According to Atkins (2009), a preferred method to measure airflow inside the engine test cell is to use square-edged orifice plates. The working principle of this method is based on the pressure drop resulting from an orifice inside the pipe. As the air flows thorough 35 an orifice, it causes a slight pressure difference between the different sides of the orifice. Actual air flow can be derived from the measured pressure drop across the orifice. 2.2.7 Exhaust gas emission measurement and analysis According to Martyr & Plint (2012), there exists emission legislation for every type of machinery that uses internal combustion engines. This means that every engine needs to go through a defined test sequences in order to get the analysis of the engine exhaust composition and emissions according to procedures. The emission composition depends heavily on engine design, operating conditions and the properties and composition of the fuel used (Martyr & Plint, 2012). Gaseous components of complete combustion are nitrogen (N2), water vapor (H2O), oxygen (O2), and CO2. During incomplete combustion many other components are formed inside the cylinder and are regarded as emissions. Typical emissions are particulate matter (PM), carbon monoxide (CO), total hydrocar- bons (THC), and nitrogen oxides (NOx). Additionally, the use of diesel fuel results in SOx emissions (Martyr & Plint, 2012). In order to measure every different component from the exhaust stream, the engine test cell needs to be equipped with measurement and analysis equipment for both, gaseous compounds and particle emissions. Instruments best suited for specific test procedures depend on individual requirements. As stated by Martyr and Plint (2012), analyzers used to measure steady-state operation are required to exhibit accuracy, sensitivity, and stability which often leads to slow response times. When emissions are measured during transient conditions, analyzers with much faster response times of a few milliseconds are required. The main factors affecting the re- sponse time are the distance between the sampling point and the analyzer, as well as the time taken to analyze the sample. Atkins (2009) lists many elements that need to be considered in the exhaust emission sampling system regardless of the technologies used. Heated lines are required to keep the temperature of the sample gas high enough to prevent water vapor and hydrocar- 36 bons from condensing to the walls of the sample line. Because many analyzers are sen- sitive to variations in pressure and flow rate, they need to be kept constant with regula- tors and sample pump. Water vapor and solid particles in the sample may cause errors in some analyzers, usually those based on the absorption of infrared radiation. In the case of possible errors, they need to be filtered from the sample. The sampling system needs to be sealed properly and flushed regularly in order to prevent errors caused by leakages of air into the sampling system and contamination from previous samples. Fourier Transform Infrared (FTIR) analyzer can be used to measure multiple gas com- pounds simultaneously from the exhaust stream of an engine (Heywood, 2018). The working principle of the FTIR analyzer is based on the unique spectrum produced by different atoms and molecules when they are exposed to infrared radiation. FTIR gas analyzers expose the sample gas to infrared radiation. Based on Fourier analysis of the gathered spectrum, different compounds in the gas can be identified and their amounts determined (Martyr & Plint, 2012). Almost every gaseous compound from the sample gas can be identified using an FTIR analyzer, except diatomic elements and noble gases. They are transparent in the spectrum due to them not absorbing any infrared radiation. These gases are in return used to record the background spectrum in order to eliminate it from the real test results (Gasmet, 2020). Nondispersive Infrared (NDIR) analyzer is also based on the unique infrared absorbance band of different compounds (Heywood, 2018). However, the NDIR analyzer cannot measure multiple compounds as it does not perform Fourier analysis on the complete absorption spectrum of the sample gas. Spe- cific filters selected according to the absorbance band of the measured compounds are required. Typically, CO2 and CO are measured using an NDIR analyzer (Martyr & Plint, 2012). A chemiluminescence detector (CLD) is based on a phenomenon in which light is pro- duced by certain chemical reactions. With this method, the amount of NOx can be meas- ured from the exhaust gas by using ozone to react with NO in the sample gas (Heywood, 2018). Following reactions produce light which is detected by a photomultiplier and the 37 resulting intensity is proportional to the concentration of NOx in the sample gas (Naka- mura & Adachi, 2013). The structure and operating principle of the CLD analyzer are il- lustrated in Figure 9. Figure 9. Structure of chemiluminescence detector measuring NO concentration (Nakamura & Adachi, 2013). A flame ionization detector (FID) can be used to measure carbon-containing compounds from the sample gas (Heywood, 2018). It is used to measure THC concentration in ex- haust gas and has a wide dynamic range and sensitivity (Martyr and Plint, 2012). The basic structure and operation principle of the FID instrument is shown in Figure 10. Figure 10. Structure of flame ionization detector measuring THC concentration (Nakamura & Adachi, 2013). Inside FID, the sample gas is burned in a hydrogen flame. Burning hydrocarbons produce free electrons and positive ions that are collected with a collector electrode. When this 38 collector containing ions is introduced to an electric field, current will start to flow through the collector. The magnitude of the current is proportional to the ionization of the hydrocarbons and can be used to determine the concentration of hydrocarbons in the sample gas (Nakamura & Adachi, 2013). An important part of making successful emission measurements is selecting the proper measuring range with respect to ex- pected concentrations and calibrating the analyzers according to this range (Nakamura & Adachi, 2013). Oxygen concentration can be measured by using paramagnetic detection (PMD) analyzer (Heywood, 2018). PMD analyzer relies on strong paramagnetic susceptibility of oxygen. PMD analyzer contains a measuring cell that has a strong magnetic field inside to which the oxygen molecules will be drawn. This movement of oxygen molecules will displace a balanced detector instrument inside the measuring cell and this displacement can be measured. Displacement of the detector is proportional to the oxygen concentration in the sample gas (Martyr and Plint, 2012). Zirconium dioxide (ZrO2) oxygen sensors are also used to compute oxygen concentration from the sample gas. It is based on the prop- erty of zirconium dioxide, where voltage is generated across the ZrO2 element when dif- ferent oxygen concentrations occur on both sides of the element (SST Sensing Ltd, 2017). The ZrO2-based lambda sensors, such as the wide band universal exhaust gas oxygen sensor, are used in determining the air-fuel ratio to be used in engine performance and emission control. Modern and robust lambda sensors enable lambda control, which is important for combustion control in LTC concepts, such as RCCI (Kasprzyk et al., 2020). The advantage of deriving air-fuel ratio based on the output from ZrO2 sensor when com- pared to multicomponent gas analyzer is that the calculation is based on a single output. This eliminates the need to consider the response times of different gas components. Additionally, on-board lambda sensors enable easier installment and good response time (Nakamura & Adachi, 2013). Particles present in the exhaust gas are indicated by the color of the exhaust gas. As stated by Martyr and Plint (2012), opacity meters are widely used to measure the opacity 39 of exhaust gas. The opacity meter operates by forming a light beam across undiluted exhaust gas to the detector. The amount of light reaching the detector depends on the number and size of the smoke particles, as well as the distance between the light source and the detector. Light absorption characteristics of the smoke particles also contribute to the results. The output of an opacity meter usually represents the percentage of the light being absorbed by the exhaust gas (Martyr & Plint, 2012). The PM measuring prin- ciple can also be based on the amount of light absorbed by soot particles caught in the filter paper. Smoke meters consist of a light source, filter paper, and a light detector. The amount of light that reaches the detector through the filter paper is dependent on the amount of smoke that is filtered from the sample gas (Atkins, 2009). There are also par- ticulate samplers that can measure the mass of the particles caught in the filter paper and optical particulate counters (Martyr & Plint, 2012). 2.3 Signal conditioning and acquisition Signal conditioning is an essential part of a measurement chain and it is located between the sensor sending an electrical signal and a device that samples and records the signals. Signal conditioning has a huge part in overall system accuracy and different sensors each having their specific output may have different requirements for conditioning. In addi- tion to analog signal transmission, some instruments require a digital interface in order to send the data forward. In an engine test cell, different digital transmission technolo- gies specific to the instruments usually have to be incorporated into a data acquisition system. Figure 11 below shows the typical process of data acquisition. 40 Figure 11. Principle of data acquisition process (adapted from Martyr & Plint, 2012). Signals produced by a sensor or a transducer often require proper amplification (Morris & Langari, 2012). After signal processing, signals are sampled and digitized. After con- verting the analog signal to digital samples, data is transmitted for monitoring, pro- cessing, and storing. 2.3.1 Signal amplification An electronic amplifier is a commonly used instrument in the measurement chain be- cause a signal coming straight from the sensor often requires some form of amplification before the acquisition. This is due to the amplitude of a sensor output being too low for being measured accurately and low-level signals are more prone to electrical interfer- ence coming from outside the measurement chain. Amplifiers, therefore, improve the sensitivity and resolution of the signal (Morris & Langari, 2012). It is important to avoid any noise affecting the signal before amplification to prevent the noise from being am- plified in the process. Amplifiers are best to be located as near to the sensors as possible 41 in order to reduce the length that the low-level signal has to travel and possibly get cor- rupted by electromagnetic interference (Martyr & Plint, 2012). Specific requirements and characteristics of an amplifier depend on the type of signal they are used to amplify. According to Rogers (2010), the combustion measurement chain requires charge ampli- fiers that convert the charge coming from the piezoelectric pressure transducer into volt- age that can be read by the measurement system. Due to the importance of in-cylinder pressure measurements in engine research, charge amplifiers are almost always present in the test cell measurement system. Figure 12 shows the main components present in the charge amplifier circuit designed for piezoelectric pressure transducers. Figure 12. Basic charge amplifier circuit for a piezoelectric pressure transducer (Rogers, 2010). As illustrated in Figure 12, a piezoelectric crystal produces an electric charge, Q, that results in a slight increase in voltage at the amplifier input which is then amplified signif- icantly. Because the gain of the amplifier is large, capacitances resulting from the sensor (Ct) and cable (Cc) do not have a significant effect on the amplified output voltage (Uo). This leaves the output voltage to be dependent only on the magnitude of the input charge and the capacitance of the feedback capacitor (Cr) required to restrict the voltage rise at the amplifier input and reduce unwanted noise (Rogers, 2010). 42 2.3.2 Measurement error reduction Measurement errors always exist to some extent in the measurement system. With a good design of the system together with proper data post-possessing and analysis meth- ods, the impact of these errors can be mitigated. Morris and Langari (2012) divide meas- urement errors into two categories, errors resulting from the measurement process and errors occurring during signal transmission and conditioning. Errors that occur during the measurement process are either systematic or random. Sys- tematic errors consistently fall on either side of the correct reading. One source of sys- tematic error is the changes in the physical properties of the measured process that are a result from the measurement situation. The significance of the error depends on the instruments and the measured process. In order to mitigate the error, the type of instru- ments must be chosen for the specific measurements. The environment will also have an effect on the measurement system by introducing an environmental input to the measurement system. These inputs are changes in the environmental conditions that have an effect on the process being measured. The effect of environmental inputs can be considered by correcting them in the output reading of the instrument or increasing the environmental resistance of the instruments (Morris & Langari, 2012). Errors induced by the environment are especially relevant in engine test conditions. Drastic tempera- ture changes expose the transducers to the effect of drift. Also, improper mounting of the transducers together with the engine vibrating can lead to additional noise in the transducer output (Rogers, 2010). Systematic errors also arise as a result of instruments wearing down and additional resistance of the wiring used to connect the sensors to the acquisition system. The wear of the measurement instruments can often be solved by calibrating the instruments regularly. Errors due to wire resistance are mitigated by using cables with properly large cross section and planning the cable routes to be as short as possible (Morris & Langari, 2012). 43 Random errors are experienced as a disturbance in measured value and are a result of unpredictable variations in the measurement system. These variations can be a result of false human observations, noise caused by electrical disturbance or sudden changes in the measurement environment. According to Morris and Langari (2012), random errors in static measurements can be reduced by calculating the average result of multiple measurement points. Errors that occur during signal transmission and conditioning mainly arise because of electrical interference resulting from nearby power cables, ground loops, poor wiring practices, analogue-to-digital conversion or other electronic devices. This interference is seen as noise in the signal and the resulting error is more significant in situations where the range of the measured signal is relatively low. This will result in a poor signal-to-noise ratio (Rogers, 2010). Current signals are inherently more resistant to noise compared to voltage signals, therefore they are preferred for long distances in noisy environments (Osman & Massoud, 2013). To prevent noise from entering the measurement system, properly shielded cables should be used for transmitting the signals and these cables should be placed as far as possible from any source of possible interference (Measure- ment Computing, 2012). The appearance of high-frequency noise in the signal is a com- mon problem in most data acquisition systems and analog filtering is an effective method of getting rid of unwanted frequencies. Basic types of filters are the low-pass filter that blocks high-frequency components, a high-pass filter that blocks low-frequency compo- nents, a band-pass filter that passes only a certain frequency band, and a band-reject filter that blocks a certain frequency band. Filters can be passive or active, the difference being that active filters consist of active components, such as transistors and operational amplifiers, in addition to passive components, like resistors and capacitors. The opera- tion of active filters is superior to passive filters, because they do not introduce as much resistance to the signal conditioning circuit as passive filters (Morris & Langari, 2012). 2.3.3 Analog-to-digital conversion 44 An important part of any data acquisition system is an analog-to-digital converter (ADC). The purpose of an analog-to-digital conversion is to turn analog sensor output into a binary number representing the analog measurement value. This binary number is then transformed into a base 10 digital number in the computer and displayed on a monitor (Measurement Computing, 2012). Resolution and accuracy are two main characteristics involving ADCs. The resolution of ADC is given as a number of bits that corresponds to the number of discrete steps that the digital value is able to divide the analog signal into. An n-bit resolution means that the ADC divides the measurement range into 2n discrete steps (Measurement Computing, 2012). Therefore, the resolution of the digital value de- pends on the resolution of the ADC and the measurement range (Smith, 2009). Martyr and Plint (2012) note that the required resolution of ADC should be selected according to the accuracy and significance of the signal. If the inaccuracy of the signal is an order of magnitude greater than the value of a discrete step in the digital output of the ADC, it is not cost-effective and may even give a wrong picture about the accuracy of the signal. According to Asad et al. (2011), ADCs with 16-bit resolution are common in high-fre- quency data acquisition systems that are used in engine testing. In measurement applications, the accuracy of the AD conversion is a critical factor. There are many possible errors during the AD conversion that can happen, some of which are practically unavoidable (Measurement Computing, 2012). In Figure 13 below, the most common types of ADC errors are presented. 45 Figure 13. Most common types of errors resulting from ADCs (Measurement Computing, 2012). In Figure 13, the straight line in each graph represent the analog output that is given to ADC and the step function represents the digital output to the acquisition system. Graph A represents the linear, therefore ideal relationship between the analog and digital out- puts. Even in a situation where the relationship would be ideal, there will be restrictions in accuracy caused by the resolution. Regardless of the ADCs resolution, there will always exist small gaps between the discrete steps that the real value will most likely be in reality (Measurement Computing, 2012). Gain and offset errors, represented in graphs B and E respectfully, can be eliminated with proper calibration of the ADC and related acquisition software. Linearity error presented in graph C means that the digital output deviates from the actual analog input nonlinearly. Graph D represents a missing code error where ADC is unable to produce a digital output for a certain analog value. Linearity error and missing codes are not possible to eliminate with calibration (Measurement Computing, 2012). 2.3.4 Test cell data communication technologies Modern engine test cells have several different equipment to be controlled and receive data from. At the same time as the need for more control has increased, also the need 46 for high-speed data transmission has increased. Engine test cell data transmission sys- tem has to be able to handle a two-way flow of data with high transfer rates (Martyr & Plint, 2012). Additionally, many devices used in test cell measurement and control sys- tems have their specific requirements for data transfer protocols and standards that must be satisfied. Controller area network (CAN) is a commonly used standard automotive engineering se- rial bus used to transmit data packets to every device that is connected to the network. Electronic devices connected to the network are usually sensors and actuators. Accord- ing to Martyr and Plint (2012), the CAN bus is able to operate in a high-noise environ- ment enabled by twisted pair physical media carrying a 5 V differential signal. Resistance to noise is a useful characteristic in engine test cell environment due to many possible sources of high-frequency noise. Serial communication provides a communication link between a transmitter and a digital system. These two systems must be equipped with serial communications cards that provide a communication point for the device. According to Smith (2009), common types of protocols used in serial communications are RS-232 and RS-485. They are hardware standards, meaning that they determine for example the topology of wires in the cable and voltage levels. RS-232 is more commonly used and has a maximum distance of around 15 meters. A significant drawback of RS-232 is that it requires a common ground for a transmitter and a receiver due to a lack of electrical insulation. RS-485 is better for industrial applications because it enables longer communication distances and provides electrical isolation (Smith, 2009). Another typical industrial connection standard found in most engine testing facilities is Universal Serial Bus (USB). It determines specifications for cables, connectors and con- nection protocols. Characteristics of USB are high data transfer rate and a wide range of applications with a maximum cable length of approximately 5 meters. USB has been re- 47 vised many times to gain faster data transfer, but all new revisions are backwards com- patible with older revisions (Compaq et al., 2000). According to Morris and Langari (2012), most computer-based data acquisition systems are connected via USB connec- tion. Modbus is a commonly used messaging protocol for industrial applications. It was devel- oped to exchange information between programmable logic controllers (PLCs) and a host computer (Smith, 2009). Modbus provides a client-server communication over transmis- sion control protocol/internet protocol (TCP/IP) network between devices that are able to read and write Modbus messages. In Modbus messaging, the Modbus client sends a request to the Modbus server. After receiving an indication, the Modbus server sends back a response to a Modbus client, which then receives a confirmation. In a standard Modbus network, there is a possibility to have one client and a maximum of 247 servers (Modbus Organization, 2006). 2.3.5 Sampling As ADCs convert a continuous analog signal into digital value, they take samples from it several times per second reducing it into a discrete signal. The required sampling rate depends on the characteristics of the analog signal. The sampling rate must be sufficient in order to reconstruct the form of the original signal. Variables that have little variations over time or are practically in a steady state usually require sampling rates of only a few hertz (Measurement Computing, 2012). Some parameters in engine testing requires high-frequency acquisition, especially signals that are required to be recorded in the crank angle domain. The need for high sampling rates also sets the requirements for the sensor bandwidth, which describes the ability of a sensor to produce a rapidly varying signal (Osman & Massoud, 2013). According to Zhang et al. (2018), the required sampling rate for signals, that are recorded in the crank angle domain, depends on the engine speed and the resolution of the crank angle intervals. The sampling frequency of crank angle-based signals can be calculated with Equation (1) 48 𝑓 = 360 ∙ 𝑁 𝑛 ∙ 60 , (1) where 𝑁 is the engine speed in rpm (revolutions per minute) and n is the resolution of the crank angle intervals. In order to record all the information present in the signal, the Nyquist sampling theorem must be applied. According to the Nyquist sampling theorem, a signal must be sampled with a sampling rate of at least two times higher than the maximum operating frequency of the sensor. If the Nyquist theorem is not applied in sampling, an aliasing effect can occur. This means that the signal reconstructed from samples differs from the original continuous signal (Measurement Computing, 2012). Even if the signal is sampled at the frequency set by Equation 1, the effective resolution of the data also depends on the sensor bandwidth. In practical engine testing applications, it is good to have even higher sampling rate for crank angle-based data acquisition than the theoretical minimum stated in the Nyquist theorem. This enables confident reconstruction of the original sig- nal that accurately represents the measured phenomenon (Zhang et al., 2014). The downside of a high sampling frequency is that it leads to large amounts of data. There- fore, sampling rates should not be unnecessarily high despite the recent developments in data storing technologies. 2.4 Combustion analysis and post-processing ACT concepts like RCCI require precise shaping of the combustion process in order to provide superior efficiency and emission characteristics. Therefore, combustion analysis providing insight into in-cylinder phenomena is crucial for both, the fundamental under- standing of the concept and its control on an applied level. The following section will 49 focus on the requirements of the measured parameters in order to perform the combus- tion analysis and the main results that are obtained. Particular attention is put towards special requirements set by the RCCI concept in question. 2.4.1 Combustion measurements Conducting research on an internal combustion engine gives certain requirements for accuracies, resolution, and post-processing methods of measured parameters. As stated by Rogers (2010), cylinder pressure and crank angle are key parameters in measure- ments aiming to understand the combustion phenomenon occurring inside the cylinder. Depending on the purpose of the measurement, other parameters can also be measured. As discussed before, the piezoelectric pressure transducer is the most widely used in- strument for measuring in-cylinder pressure due to its advantages and always requires proper amplification (Rogers, 2010). An angular encoder is used to measure crank angle reference at multiple points during a crankshaft revolution. When cylinder dimensions are known, the cylinder volume required in certain calculations can be derived from the angular position of the crankshaft (Heywood, 2018). After proper signal processing, sen- sor signals together with the encoder signal are sampled and digitized by the data acqui- sition system for further processing, displaying, and storing the data in the crank angle domain. The measurement task defines the required resolution for the crank angle-based meas- urements. According to Rogers (2010) and Atkins (2009), basic calculations related to the combustion process and the direct analysis of a pressure curve are possible with a reso- lution close to 1 CAD (crank angle degree). On the other hand, some measurements like knock detection and combustion noise require resolutions down to 0.1 CAD. Martyr and Plint (2012) divide the results of combustion analysis into two categories, direct and indirect results. Direct results, unlike indirect results, are acquired by analyz- 50 ing the raw data coming from measurement devices and do not require further calcula- tions and post-processing. Results obtained specifically from the raw pressure curve in- clude maximum pressure and pressure rise rate with respective angular positions. Addi- tionally, abnormal conditions like knocking, misfiring, and combustion noise can be de- tected. In case of direct results, errors are similar in magnitude to the errors in measured signals. For calculated results, errors in measurements may lead to deviations by several orders of magnitude greater (Rogers, 2010). The measurement uncertainty for indirect results can be determined using the partial derivative method originally proposed by Kline and McClintock (1953). The uncertainty can be calculated with Equation (2) ∆𝑌 = [∑ ( 𝜕𝑌 𝜕𝑥𝑖 ∙ ∆𝑥𝑖) 2𝑛 𝑖=1 ] 1 2 , (2) where Y is the calculated variable, xi is the measured variable, ∆xi represents the error of the measured variable and n is the number of independent measurements used to calculate Y. The application of this method for engine uncertainty analysis can be found for instance in Mikulski et al. (2016). 2.4.2 Top dead center detection Accurate alignment of the cylinder pressure curve in relation to the crank angle is im- portant to mitigate errors in results and calculations. Alignment of the encoder is done by determining the angle difference between the top dead center (TDC) of the piston and the trigger mark of the angular encoder (Rogers, 2010). This difference is known as TDC offset and it defines how much the cylinder pressure curve has to be shifted in re- lation to crank angle. Rogers (2010) states that for some calculations, the determination of TDC position requires an accuracy of 0.1 CAD. 51 TDC can be determined with various methods. The manual way of TDC detection can be time-consuming and prone to errors (Rogers, 2010). The procedure requires a dial gauge to measure the distance of the piston from the TDC. The crankshaft is rotated approxi- mately 90 degrees away from the TDC and the reading from the dial gauge is recorded. After this, the crankshaft is rotated in opposite direction as long as the piston distance indicated by the dial gauge equals the reading recorded previously. True TDC will be ex- actly in the middle of these two positions where piston height was equal according to the dial gauge (Rogers, 2010). After the TDC position has been determined manually, the encoder can be calibrated to trigger exactly at the TDC. According to Rogers (2010), the disadvantage of this method is that it does not consider that the actual TDC position may vary because some parts can change shape while the engine is running. TDC can also be determined by reading the maximum value of the cylinder pressure curve (Martyr & Plint, 2012). According to Rogers (2010), assuming that cylinder peak pressure indicates TDC is not accurate enough, because thermodynamic loss angle spe- cific to engine type and conditions needs to be considered. The main weakness of this method is the difficulty to accurately determine the thermodynamic loss angle. Typically, combustion measurement devices contain a specific built-in TDC detection procedure based on the method of pressure curve determination. In addition to the methods described above, there is a possibility to use a specific type of capacitance sensor to accurately determine the TDC. The sensor needs to be mounted to a deactivated cylinder while the engine is running on the remaining cylinders. The sensor probe is located inside the cylinder and it measures the distance to piston crown. By using the data from the capacitive sensor and encoder, TDC can be determined with great accuracy (Rogers, 2010). 2.4.3 Zero-level correction 52 Because piezoelectric transducers are unable to measure the absolute cylinder pressure due to their dynamic response, zero-level correction (pegging) needs to be applied for further calculations and processing. In zero-level correction, the pressure curve obtained directly from the transducer is shifted on the Y axis according to a correction offset value (∆p) to have the curve represent the absolute cylinder pressure (Rogers, 2010). The sim- plest method of zero-level correction is to shift the whole cylinder pressure curve until it reaches the atmospheric pressure value at a reference point. According to Merker et al. (2012), this method has significant limitations because actual intake port pressure is not near atmospheric in the majority of modern engines. One sufficiently accurate and fast method to determine the offset value is to calculate the thermodynamic zero-line correction value. In this method, two points from the pres- sure curve, are chosen before TDC and offset calculated with Equation (3) ∆𝑝 = ( 𝑉1 𝑉2 ) 𝛾 ∙ 𝑝1 − 𝑝2 1 − ( 𝑉1 𝑉2 ) 𝛾 , (3) where Vn and pn represent cylinder volume and measured pressure at respective points of the curve, and exponent γ is the polytropic coefficient (Merker et al., 2012). The meas- ured cylinder pressure curve is then shifted accordingly cycle by cycle. Calculation points which are not significantly affected by noise resulting from vibrations caused by intake valve motion must be chosen. Points with angle values from 100 CAD to 70 CAD before TDC are typically used (Rogers, 2010). The zero-level correction that is based on a pressure value measured directly from the intake port is an accurate correction method. Correction offset is determined as a differ- ence between the measured absolute intake port pressure and in-cylinder pressure value at a reference angle (Merker et al., 2012). The reference crank angle is usually the point at the bottom dead center (BDC) during intake valve opening where the piston speeds are the lowest, resulting in the lowest pressure gradients and oscillations. Care 53 needs to be taken that the reference point is outside the valve overlap or the valve clos- ing phase. The two above requirements can be conflicting while advanced variable valve actuation strategies are incorporated. In such a case, more careful consideration of the reference is required. To mitigate the error resulting from noise, averaging the measured pressure values around the reference point should be used. To achieve the best accuracy, intake port pressure should be measured from every cylinder individually (Rogers, 2010). 2.4.4 Mean effective pressure Mean effective pressure (MEP) is an important performance indicator of an internal combustion engine, which can be derived from the in-cylinder pressure curve plotted against cylinder volume, called as p-V diagram (Merker et al., 2012). Net indicated work (Wi) is a quantity that describes the work done by in-cylinder gas on a piston during the whole engine cycle and it can be calculated by integrating the enclosed area on the p-V diagram over the whole engine cycle, as shown in Equation (4) 𝑊i = ∮ 𝑝 d𝑉, (4) where p is the cylinder pressure, and dV is the incremental swept volume (Heywood, 2018). Net indicated mean effective pressure (IMEPn) describes the ratio between the net indicated work and swept volume of the cylinder, as expressed in Equation (5) IMEPn = 𝑊i 𝑉 , (5) where V is the cylinder swept volume (Heywood, 2018). Gross work is done on a piston during the working cycle and is calculated by taking the integration in Equation 4 over the compression and expansion strokes instead of the whole cycle (Martyr & Plint, 2012). Gross indicated mean effective pressure (IMEPg) is calculated by dividing the gross work by swept volume. As stated by Martyr and Plint (2012), pumping mean effective pressure 54 (PMEP) indicating the pumping losses during the gas exchange cycle can be derived from IMEPn and IMEPg with Equation (6) PMEP = IMEPg − IMEPn, (6) or by taking the integration in Equation 4 over the gas exchange cycle. Due to mechanical losses and work necessary to drive auxiliary systems on a test bench, work delivered by the engine is always lower than indicated. Brake work of the engine can be expressed by using the brake power. According to Heywood (2018), brake work for four stroke engines can be calculated with Equation (7) 𝑊b = 2 ∙ 𝑃b 𝑁 , (7) where Pb is the brake power that the engine generates. From the calculated brake work, the brake mean effective pressure (BMEP) of the engine can be calculated using the Equation (8) BMEP = 𝑊b 𝑉 . (8) BMEP is an accurate indicator of engine’s performance because it considers the losses. The friction mean effective pressure (FMEP) indicates the amount of work that is lost to friction. It can be derived from IMEPn and BMEP of the whole engine with Equation (9) FMEP = IMEPn − BMEP. (9) According to Martyr and Plint (2012), accurate TDC detection is crucial for calculating the IMEP because even a 1 CAD difference between the encoder and true position of 55 the crankshaft will result to approximately 5% error in IMEP. For IMEP calculations, res- olution for crank angle-based cylinder pressure data is required to be at least 1 CAD (Rogers, 2010). 2.4.5 Heat release rate Heat release rate is a quantity in engine testing that indicates the energy released from the combustion (Heywood, 2018). Fundamentally, the cylinder pressure curve displays the in-cylinder pressure variations resulting from combustion, changing cylinder volume and heat transfer to surfaces (Rogers, 2010). Pressure variations with respect to volume can be derived from the polytropic coefficient and a change in volume. When the pres- sure variation due to changing volume is subtracted from the in-cylinder pressure curve, a curve resembling the net heat release can be derived with respect to crank angle de- grees. Due to the strong correlation between heat release rate and in-cylinder pressure curve, accurate zero-level correction and TDC position are crucial in order to derive ac- curate heat release curve (Rogers, 2010). Heat release rate representing the rate of re- leased energy during combustion minus heat absorbed to the walls and crevice volume is known as apparent or net heat release rate (Heywood, 2018). Multiple sources deter- mine the net heat release rate based on the first law of thermodynamics with Equation (10) d𝑄net d𝜃 = 𝛾 𝛾 − 1 ∙ 𝑝 ∙ d𝑉 d𝜃 + 𝛾 𝛾 − 1 ∙ 𝑉 ∙ d𝑝 d𝜃 , (10) where θ is the crank angle and γ describes the composition of gas at a specific crank angle (Sogbesan, 2016; Willems et al, 2021). The gross heat release rate is obtained when heat losses are added to the net heat release rate (Heywood, 2018). Heat loss can be derived by calculating the rate of heat transfer between in-cylinder gasses and the cylinder wall with Equation (11) 56 d𝑄ht d𝑡 ∙ 1 𝐴cyl = ℎc ∙ (𝑇g 4 − 𝑇w 4) + 𝜎 ∙ 𝜀 ∙ (𝑇g 4 − 𝑇w 4), (11) where Acyl is the cylinder wall area, hc is the heat transfer coefficient, σ is the Stefan- Boltzmann constant, ε is the emissivity, Tg is the temperature of the in-cylinder gas, and Tw is the temperature of the cylinder wall (Heywood, 2018). Therefore, the gross heat release rate is determined by Equation (12) d𝑄gross d𝜃 = 𝛾 𝛾 − 1 ∙ 𝑝 ∙ d𝑉 d𝜃 + 𝛾 𝛾 − 1 ∙ 𝑉 ∙ d𝑝 d𝜃 + d𝑄ℎ𝑡 d𝜃 . (12) The cumulative heat release rate is calculated by integrating the gross heat release with respect to the crank angle from the closing of the intake valve to the opening of the exhaust valve (Willems et al., 2021). The curve of cumulative heat release indicates the amount of energy that has been released in relation to crank angle degrees. It enables one to derive the phasing of the 5%, 50%, and 90% of the total energy released, known as CA05, CA50, and CA90 respectively. Values of these parameters are important in RCCI control applications (Li et al., 2017). They are used in assessing ignition delay, and com- bustion phasing, as well as in monitoring the effect of input variables on the rate of com- bustion (Sogbesan, 2016). Combustion duration can also be derived from the cumulative heat release curve (Heywood, 2018). During heat release rate calculations, the polytropic coefficient, γ, is often assumed con- stant, while in reality it varies due to several factors such as heat transfer to the walls, in-cylinder gas composition, and valve timing (Lee & Min, 2019). The problem with using a constant polytropic coefficient in RCCI is that it can vary significantly between different operating points due to above-mentioned reasons. More accurate estimation of the pol- ytropic coefficient is crucial when developing and using advanced combustion concepts, such as RCCI (Rogers, 2010). 57 Rogers (2010) states that the calculation of heat release rate in commercial combustion measurement systems is typically computed with a simplified process resulting in net heat release. The calculated value is approximately 20% lower than the gross heat re- lease due to ignored surface and blow-by losses. Error in the result is more significant towards the end of the combustion process. Simplified calculation may be required if repeatability is more important than the high level of accuracy and when processing ca- pacity is of concern. As discussed above, accurate determination of heat release rate and thus cumulative heat release rate in RCCI applications are particularly important in un- derstanding the combustion phenomenon and enabling its efficient control. Therefore, the use of any simplified process should be critically assessed. For heat release calcula- tions, a resolution of at least 1 CAD should be used, and although finer resolution im- proves the accuracy, it can significantly increase processing time (Rogers, 2010). 2.4.6 Gas exchange analysis The gas exchange process is part of the engine cycle where exhaust gases are forced out of the cylinder and a fresh mixture of air is drawn from the intake port. Gas exchange analysis offers a thorough understanding of the phenomena during this event and has significant importance in modern engine research and development. According to Rog- ers (2010), pressure measurements from intake and exhaust manifolds provide infor- mation to optimize gas exchange processes and accurately calculate the PMEP, which together with BMEP allows a more precise evaluation of friction losses. Measuring absolute intake port pressure during the gas exchange process also allows accurate estimation of in-cylinder air flow. According to Ferguson & Kirkpatrick (2015), the mass flow rate through the valve is determined with Equation (13) d𝑚 d𝑡 = 𝜌v ∙ 𝐴eff ∙ 𝑈is , (13)