Deep Seabed Mining Machine Fault Diagnosis And Sensor Fault-tolerant Systems

Deep-seabed mining robot is a quite complex mechanical system with hydraulic drives. The traditional fault diagnosis methods generally need relative accurate mathematics model, so their applications in mining robot fault diagnosis come up against many drawbacks. By adopting intelligent diagnosis theory to this fault diagnosis system, sensor fault-tolerant technique and integrated neural network are combined into an implement for mining robot conditions monitoring and fault diagnosis. It has the count for much meaning to stable deep-sea mining process and raise to the whole dependability of the deep-sea mining system.In this paper, the mining robot is composed of three parts, including mining system, running outfit and hydraulic system. Considering the characteristics of each part's fault, the deep-seabed mining robot fault diagnosis system is studied and designed. Firstly, the structure and functions of the system are presented. Secondly, a method of transducer fault tolerance for deep seabed mining robot is proposed in this paper. By combining the hardware redundant technology with weighted fusion algorithm and gray forecast model, the transducer fault-tolerant unit is constructed, which is with sensor fault detection, fault isolation and redundancy, and it ensures that the sensor's output is always valid when sensor's faults are not happened simultaneously. Thirdly, the fault diagnosis on mining robot is realized by founding multi-child neural network and syncretizing diagnosis results of multi-child neural network. The simulation results shows that the system is with high practicability and reliability, it is well prepared for the coming "sea test".Finally, the system software is developed with VISUAL C++ 6.0 and SQL Server 7.0 in the environment of Windows2000. The functions, such as imformation inputing, data query, real-time fault diagnosis and fault simulation, are implemented.