Building of Multibody Dynamics and Machine Learning Hybrid Model and Application to Drivetrain Tortional Vibration Development

Ryu Chikasue･Roberto Gonzalez Flores･Seijun Morita･Kazuhiro Tanaka･Kenji Kudo (Mazda)

In order to reduce CAE simulation time using machine learning, one important issue is how to build a machine learning model that can be used without re-collecting data and retraining when the system’s design changes. In this study, we suggested a hybrid approach to use machine learning model (LSTM) representing non-linear components that cause simulation difficulties (e.g. DMF including contact behavior) in a drivetrain multibody dynamics model. A maximum 90% simulation time reduction was confirmed when we applied this approach to the development of drivetrain tortional vibration.

Analysis of the Mechanism of Vibration During THS Engine Start-up Due to HEV Damper Characteristics

Masaki Sagawa･So Shimizu･Yuki Notake (Toyota Motor)

The THS system connects the engine and driveshafts through a planetary gear, causing vibrations to be transmitted not only through the engine mounts but also via the driveshafts to the vehicle body during engine start-up. This study reports that the torsional characteristics of the damper, particularly around 0Nm, significantly impact engine start-up vibration.

Application of PT-VRS in a non-echo environment for vehicle vibration and noise development

Yoshihide Narita (Toyota Motor)

In recent prototype-less vehicle development, PT-VRS has been utilized during platform development, primarily focusing on the dynamic performance in low-frequency ranges .
This report introduces case studies on the application of PT-VRS in a non-echo environment for developing vehicle vibrations and noise from low to mid-frequency ranges.

Study on Amplitude Dependent Friction Damping at Joints and Development of a Simple Estimation Method

Sotaro Tajiri (NewtonWorks)

Dissipation due to friction at the joint interface contributes to the vibration response of a structure, but it is difficult to evaluate quantitatively. In this study, vibration tests and nonlinear dynamic contact analysis were performed on a simple bolted joint structure to evaluate the amplitude dependence of damping caused by friction. In addition, a simple method was developed to estimate the friction damping of a structure from the static contact state and natural vibration mode.

The Method to Break Down NVH Performance Targets from Vehicle Sound Pressure to Component Eigenvalues

Takuro Sonoda･Koji Saito (Altair Engineering)･Mari Nakagami (DAIHATSU MOTOR)

In order to define target for each component in early stage of development, the method to break down NVH performance target from vehicle sound pressure to component eigenvalues are proposed in this paper. At first, the degree of freedom of each component which consist of full vehicle is reduced by modal reduction method, and then, sound pressure target values are broken down by understanding relationship between component eigenvalues and sound pressure with using that modal reduction model. By applying this method to product development, the availability of this method with a realistic man hour within the product studying duration was shown.