Alarm based root cause diagnostics
Koivumäki, Edward (2015)
Koivumäki, Edward
2015
Kuvaus
Opinnäytetyö kokotekstinä PDF-muodossa.
Tiivistelmä
The increasing requirements regarding reduced emissions and improved engine efficiency contribute in more complex engine systems. To date, the engine troubleshooting has been done mostly manually and is a difficult and time consuming process. The development of more advanced and automated fault diagnostic systems would facilitate the troubleshooting procedure.
The aim of the thesis was to investigate the potential of determining the most probable root causes based on existing engine alarms. Alarms are triggered due to many different causes, and the intention was by identifying which alarms that are sensitive to which root causes, create a type of decision structure. When one alarm triggers no further conclusions may be taken, but with several active alarms decisions are made through the created combinations and subsets of alarms. The purpose was not to determine the exact root cause, which would be far too difficult, but to list the most likely root causes.
In the automation system Wärtsilä Unified Controls (UNIC), the structure of the diagnostic system consists of diagnostic tests and diagnostic cores. The diagnostic core was developed to handle the type of dependencies and combinations between alarms that are proposed in this research. Diagnostic tests may be periodically executed and trigger alarms according to the test results.
The thesis includes a theoretical study of fault diagnostics approaches and a case study describing the investigated dual fuel engine. The results origin from an engine laboratory test as well as from alarm knowledge gathered through interviews. The faults simulated in the engine tests indicated that the proposed alarm combinations are accurate, but if the engine is operated at lower loads all alarms will not react. By setting up sophisticated alarm combinations the root causes are possible to determine mainly on existing alarms. The work is limited to common engine problems which made it possible to accomplish, the creation of combinations for the whole engine is a time consuming but viable process.
The aim of the thesis was to investigate the potential of determining the most probable root causes based on existing engine alarms. Alarms are triggered due to many different causes, and the intention was by identifying which alarms that are sensitive to which root causes, create a type of decision structure. When one alarm triggers no further conclusions may be taken, but with several active alarms decisions are made through the created combinations and subsets of alarms. The purpose was not to determine the exact root cause, which would be far too difficult, but to list the most likely root causes.
In the automation system Wärtsilä Unified Controls (UNIC), the structure of the diagnostic system consists of diagnostic tests and diagnostic cores. The diagnostic core was developed to handle the type of dependencies and combinations between alarms that are proposed in this research. Diagnostic tests may be periodically executed and trigger alarms according to the test results.
The thesis includes a theoretical study of fault diagnostics approaches and a case study describing the investigated dual fuel engine. The results origin from an engine laboratory test as well as from alarm knowledge gathered through interviews. The faults simulated in the engine tests indicated that the proposed alarm combinations are accurate, but if the engine is operated at lower loads all alarms will not react. By setting up sophisticated alarm combinations the root causes are possible to determine mainly on existing alarms. The work is limited to common engine problems which made it possible to accomplish, the creation of combinations for the whole engine is a time consuming but viable process.