Dynamic Characteristics Prediction Model for Diesel Engine Valve Train Design Parameters Based on Deep Learning

Abstract

This paper presents a comprehensive study on the utilization of machine learning and deep learning techniques to predict the dynamic characteristics of design parameters, exemplified by a diesel engine valve train. The research aims to address the challenging and time-consuming analysis required to optimize the performance and durability of valve train components, which are influenced by numerous factors. To this end, dynamic analyses data have been collected for diesel engine specifications and used to construct a regression prediction model using a gradient boosting regressor tree (GBRT), a deep neural network (DNN), a one-dimensional convolution neural network (1D-CNN), and long short-term memory (LSTM). The prediction model was utilized to estimate the force and valve seating velocity values of the valve train system. The dynamic characteristics of the case were evaluated by comparing the actual and predicted values. The results showed that the GBRT model had an R2 value of 0.90 for the valve train force and 0.97 for the valve seating velocity, while the 1D-CNN model had an R2 value of 0.89 for the valve train force and 0.98 for the valve seating velocity. The results of this study have important implications for advancing the design and development of efficient and reliable diesel engines.

A Multi-Input Convolutional Neural Network Model for Electric Motor Mechanical Fault Classification Using Multiple Image Transformation and Merging Methods

“…To address these challenges, we suggest a two-pronged approach. Firstly, the integration of simulation-based methods [24,25] into the fault diagnosis process presents a viable solution. By utilizing simulated data, we can artificially augment the dataset with a wider range of fault conditions, including those not commonly encountered in real-world operations.…”