Crash Pulse Prediction Using Regression Algorithm with Gradient Descent Optimization Method for Integrated Safety Systems
- Gerald Joy Alphonso Sequeira - Technische Hochschule Ingolstadt, CARISSMMA, Germany ,
- Anudeep Reddy Konda - Technische Hochschule Ingolstadt, Germany ,
- Robert Lugner - Technische Hochschule Ingolstadt, Germany ,
- Ulrich Jumar - Otto-von-Guericke-Universität, Germany ,
- Thomas Brandmeier - Technische Hochschule Ingolstadt, Germany
ISSN: 2327-5626, e-ISSN: 2327-5634
Published March 28, 2022 by SAE International in United States
Citation: Sequeira, G., Konda, A., Lugner, R., Jumar, U. et al., "Crash Pulse Prediction Using Regression Algorithm with Gradient Descent Optimization Method for Integrated Safety Systems," SAE Int. J. Trans. Safety 10(2):2022, https://doi.org/10.4271/09-10-02-0009.
Crash pulse prediction is one of the crucial factors in developing integrated (a unified active and passive) safety systems in the vehicle. As discussed by many researchers, the crash pulse can have considerable effects on seat occupant response. So, thereby predicting the crash pulse can help to decrease the severity of the injury in many cases. In this research article, we propose a machine-learning-based model to predict the crash pulse during head-on collisions. The model uses a regression algorithm and gradient descent optimization method for accurate predictions. After introducing the topic, the related works from different researchers highlight the need to predict the crash pulse. The following section will discuss the method adopted for generating the data and selecting the training and testing data. The training data is used to learn the algorithm, and the testing data is used for validating the prediction model. A detailed methodology for developing and optimizing the hypothesis for the prediction model is also explained. Subsequently, a section discussing the evaluation of the model in three steps and corresponding results are presented. The results show a good similarity between the prediction and the true (FEM data set) values. The prime outcome of this research is the possibility of using outputs from the prediction model to make decisions on the safety action in the pre-crash phase.