Abstract
Semi-active suspensions can control damping force and provide a good ride comfort. However, it requires dedicated sensors on four wheels, which increases cost and tuning man-hours. Therefore, development of sensor value prediction technology based on CAN data has been promoting to achieve low cost sensorless semi-active suspension system. This paper describes the suspension control method to replace the sensor function to the neural network by getting data of the vehicle with the sensors as a teacher data, and training. As a result of verifying with the data in the test course, the neural network could almost predict the sensor value.
