Predictive Model Development Using Machine Learning for Engine Cranktrain System
Highly competitive automotive market demands shorter product development cycle while maintaining higher standards of performance in terms of durability and Noise Vibration & Harness (NVH). Engine cranktrain system is one of the major vibration sources in engine and first torsional mode frequency is a key parameter which influences vibration characteristics. Current CAE (Computer Aided Engineering) workflow for evaluating cranktrain system performance is time-consuming and takes around 55 Hrs. It involves crankshaft geometry cleanup, stiffness calculation, 1D model building and post processing. Over the time, significant historical data has been created while performing this virtual simulation during the product development cycle. Having a trained Machine Learning (ML) model based on this historical data, which can predict first torsional mode frequency accelerates the virtual validation. In this paper, prediction of first torsional frequency of cranktrain system using ML is presented. Altair’s Knowledge Studio is used for ML model building, and predictions are compared with CAE results. Based on domain expertise, feature selection and data cleanup are performed. Predictions with linear regression and deep learning (DL) algorithms are compared and hyperparameters are manually optimized for obtaining right balance between a reliable model and its predictions within acceptable error. The ML model based on DL algorithm delivered more than 90% variance-explained on training and test database. Also, more than 85% predictions on test database predicted performance within 15% error. Torsional Vibration Damper (TVD) ring inertia, pulley inertia and crankpin mass per conrod are found to be the most influencing parameters for the first torsional mode. Prediction using DL model is in good agreement (< 10 % error) with CAE results. Overall, the developed predictive model using DL algorithm helped in achieving up to 70%-time reduction with respect to current CAE cycle.
- Record URL:
-
Availability:
- Find a library where document is available. Order URL: http://worldcat.org/issn/01487191
-
Supplemental Notes:
- Abstract reprinted with permission of SAE International.
-
Authors:
- S, Aravamuthan
- Sharma, Ayush
- K, Karthikeyan
- S Kangde, Suhas
-
Conference:
- WCX SAE World Congress Experience
- Location: Detroit Michigan, United States
- Date: 2023-4-18 to 2023-4-20
- Publication Date: 2023-4-11
Language
- English
Media Info
- Media Type: Web
- Features: References;
-
Serial:
- SAE Technical Paper
- Publisher: Society of Automotive Engineers (SAE)
- ISSN: 0148-7191
- EISSN: 2688-3627
- Serial URL: http://papers.sae.org/
Subject/Index Terms
- TRT Terms: Computer aided design; Computer aided manufacturing; Machine learning; Product development
- Subject Areas: Highways; Vehicles and Equipment;
Filing Info
- Accession Number: 01879612
- Record Type: Publication
- Source Agency: SAE International
- Report/Paper Numbers: 2023-01-0150
- Files: TRIS, SAE
- Created Date: Apr 19 2023 4:34PM