Study Of The Diagnosis Technology Of Electric Power Steering

Electric power steering (EPS) is an advanced steering system that uses an electric motor to provide steering assist torque which is controlled by ECU according to signals from torque sensor and vehicle speed sensor. For the benefits of EPS such as improvement in safety,remarkably less fuel consumption etc, it is going to replace Hydraulic Power Steering (HPS) in the following years. People have paid closer attention to the EPS technology all over the world day by day. This paper talks about the fail-safe and diagnosis studies about EPS. In this paper, a Fault Tree (FT) has been developed as a model for EPS, through Fault Tree Analysis (FTA), system faults have been identified and the process of the development of diagnostics has been analyzed. After that, a failure diagnosis program based on Artificial Neural Network (ANN) with error Back Propagation is designed and a test on the EPS bench is carried out to verify it. In chapter one, the components, control strategies, benefits and development of EPS are described. To compare EPS with HPS, EPS has the merits such as ability to provide proper assist force under different situations, saving energy, saving space, assembly easier, short period of development etc., so it is a future development direction of power steering. Then key issues of EPS such as assist characteristic, motor and sensor technology, control strategy, fail-safe and diagnosis, are raised and discussed. This chapter also describes the diagnosis technology and diagnosis method for automotive electronic control system. In chapter two, the development process and application of FTA is described. To the model of fault tree, its basic theory and meaning of special symbol are illustrated in detail. The concept of reliability chart and minimum cut set are raised, and the methods how to get them are explained. And the necessity and importance of FTA used in EPS system are discussed also. In chapter three, FTA is described first. FTA is a method to analyze the system from the failure or the hazard event to the root cause, and to express the whole process in a tree shape chart. Then the structure and basic theory of EPS are studied. Based on the relation of all the key parts, a general fault tree of EPS is constructed, through the established fault tree the main failure parts are defined for the diagnosis model. To the different type of EPS, failure mode analysis of electric control unit, electric motor, torque sensor, gear box etc. are carried out. In the case of the EPS system of column type developed by us, a specific fault tree is constructed. Based on the specific fault tree, the main failure components are defined and system failure modes such as pilot damp circuit failure, torque sensor failure, vehicle speed sensor failure, electric motor and clutch failure are diagnosed and analyzed. As the same time, to the mechanical steering part failure of EPS, the minimum cut set is found for the reliability design. In chapter four, the overall concept of Artificial Neural Network (ANN), structure of neuron and ANN, activation function, feed-forward network, learning process of neuron, Hebb learning rule and Delta learning rule are introduced. Fundamental of Error Back Propagation (BP) learning rule, process of multilayer BP neural network are described concretely. The improved BP learning rule with momentum is illustrated and the attention problem and design of BP neural network is discussed in addition. Based on BP learning rule with momentum, a two-hidden-layer neural network is designed for the diagnosis of EPS, and software design about EPS diagnosis system is introduced. In chapter five, characteristic of EPS working in different situations including normal status, torque sensor failure status, electric motor circuit open status, clutch circuit open status, are tested on EPS test bench. In the experiment, torque sensor failure can be done by replacing the original torque sensor by a bad one; electric motor open situation can be done by opening the cable between the electric motor and ECU; by opening the cable between the clutch and ECU, clutch open situation will happen. The characteristic and system data from above test can be used in the diagnosis system. The test data including torque value ,electric motor current, electric motor voltage and vehicle speed ,should be classified in two kinds, one is for training set, another is for testing set. After standardizing of the training set and the testing set, the training time of the diagnosis network is greatly shortened and the precision of the system is improved at the same time. By adding momentum to...