Intelligent Stress Estimation in Hydraulic Crane Structures Using Deep Sequential Models

annif.suggestionsmachine learning|simulation|time series|deep learning|neural networks (information technology)|modelling (representation)|hydraulics|dynamics|artificial intelligence|hydraulic systems|enen
annif.suggestions.linkshttp://www.yso.fi/onto/yso/p21846|http://www.yso.fi/onto/yso/p4787|http://www.yso.fi/onto/yso/p12290|http://www.yso.fi/onto/yso/p39324|http://www.yso.fi/onto/yso/p7292|http://www.yso.fi/onto/yso/p3533|http://www.yso.fi/onto/yso/p10545|http://www.yso.fi/onto/yso/p4095|http://www.yso.fi/onto/yso/p2616|http://www.yso.fi/onto/yso/p25132en
dc.contributor.authorSAEED, SOHAIB
dc.contributor.facultyfi=Tekniikan ja innovaatiojohtamisen yksikkö|en=School of Technology and Innovations|-
dc.contributor.orcidhttps://orcid.org/0009-0002-6433-0183-
dc.contributor.organizationfi=Vaasan yliopisto|en=University of Vaasa|
dc.date.accessioned2025-06-19T10:11:11Z
dc.date.accessioned2025-06-25T18:04:10Z
dc.date.available2025-06-19T10:11:11Z
dc.date.issued2025-05-30
dc.description.abstractThe accurate estimation of the structural stress in the hydraulically crane system is very im-portant to ensure the safety and reliability, as it leads to integration of Real time Control. This study focuses specifically on the data driven approach where the stress prediction is carried out using the deep sequence learning, trained on the simulated operational data from the Flexible Hydraulic Crane Model. The mechanical system under investigation, a Flexible Multibody repre-sentation of PATU 655 crane, was subjected to different dynamic simulations where real time sensor signals such as joint torques, joint forces, accelerations were recorded across various payload and motion scenarios. To learn the complex temporal patterns governing the stress evo-lution, two deep learning architectures were developed and compared. The first was the CNN+BiLSTM Model, where convolutional layers act as feature extractors and bidirectional LSTMs models learned the pattern of temporal dependencies. The second model in comparison was taken as Transformer based sequence model, utilizing the positional encoders and attention mechanisms to capture long range relationships in the multivariate input sequences. The key point to be noted was that both models were trained to predict the stress profile of movement of the crane directly from the sequence of the mechanical and kinematic features. Results showed that both architectures achieved strong performance in the unseen test cases. The CNN+BiLSTM model reached a test MAE of 0.0348 abd MSE of 0.0032, while transformer outper-formed it with a significantly lower MAE of 0.0054 and MSE of 0.0000687. This thesis demon-strates that the deep learning models can serve as the intelligent surrogates for physical simula-tion, enabling fast and accurate stress estimation in real time applications. The findings are clear pathway for the future application of Artificial Intelligence based structural monitoring and intel-ligent control system in Flexible multibody systems.-
dc.format.bitstreamtrue
dc.format.contentfi=kokoteksti|en=fulltext|-
dc.format.extent61-
dc.identifier.olddbid23907
dc.identifier.oldhandle10024/19857
dc.identifier.urihttps://osuva.uwasa.fi/handle/11111/12522
dc.identifier.urnURN:NBN:fi-fe2025053056428-
dc.language.isoeng-
dc.rightsCC BY 4.0-
dc.source.identifierhttps://osuva.uwasa.fi/handle/10024/19857
dc.subject.degreeprogrammeMaster's Programme in Sustainable and Autonomus Systems (SAS)-
dc.subject.disciplinefi=Tietojärjestelmätiede ja automaatiotekniikka|en=Information Systems Science and Automation Technology|-
dc.subject.specializationIn person-Urgent-
dc.subject.ysomachine learning-
dc.subject.ysosimulation-
dc.subject.ysotime series-
dc.subject.ysodeep learning-
dc.subject.ysoneural networks (information technology)-
dc.subject.ysomodelling (representation)-
dc.subject.ysohydraulics-
dc.subject.ysodynamics-
dc.subject.ysoartificial intelligence-
dc.subject.ysohydraulic systems-
dc.titleIntelligent Stress Estimation in Hydraulic Crane Structures Using Deep Sequential Models-
dc.type.ontasotfi=Pro gradu -tutkielma|en=Master's thesis|sv=Pro gradu -avhandling|-

Tiedostot

Näytetään 1 - 1 / 1
Ladataan...
Name:
Intelligent Stress Estimation in Hydraulic Crane Structures Using Deep Sequential Models.pdf
Size:
1.72 MB
Format:
Adobe Portable Document Format
Description:
Intelligent Stress Estimation in Hydraulic Crane Structures Using Deep Sequential Models