Techniques And Methods Of Fault Diagnosis Of Hoist Braking System Based On Neural Network

Hoist is the "throat" and a big key equipment of mine. So far, many major serious accidents have happened on mining production in China because of brake system faults, which have caused a huge economic losses and casualties. An important characteristic of the fault diagnosis technique is that combination which of with concrete engineering technology is much closer than ever, it pays more attention to solving practical problems and adhere to the principle of putting prevention first in order to ensure the security and reliability of engineering system.Modern diagnostic techniques and methods have two development directions in solving practical engineering problems. One aspect is computer aided decision support system which is a technical combination of neural network, computer, FTA (Fault Tree Analysis), etc. The other aspect is using real-time monitoring system or achieving high performance of the control system which namely "braking torque NNSOC(Neural Network Fuzzy Self-organizing Controller)". It can make braking deceleration constant and the mount of the impact minimum and reduce brake failure rate significantly when the load changes or other unexpected event occurs.Firstly, analyse the mechanism of brake failure system with FAT. Then present the on-line monitoring and diagnosis methods in braking system and establish a mine hoist test bedstand according to braking system FTA. The braking system working conditions parameters were monitored and diagnoses were given by the hoist bedstand Monitoring results provide the simulation data of fault data samples and the control accuracy for braking system NNEDS fault diagnosis, constant deceleration NNSOC design. Make an analysis of the reliability of braking system circuit contact with MCSFT(Minimum Cut Set Fault Tree) and computers, find out the weaknesses of structural fault in circuit contact and set the fault monitoring point to ascertain the reason of electrical fault rapidly which happened frequently in mine hoist so as to enhance the hoisting efficiency.Secondly, we obtained the neural network optimal intelligent algorithm of neural network fault diagnosis and neural fuzzy control(Intelligent algorithm trainlm and traingdx are optimal algorithm of fault diagnosis and neural fuzzy control separately.) and the theoretical methods of structural optimal design through a comparative experiment on performance of fault diagnosis, neural fuzzy control with fault diagnosis samples and braking torque neural fuzzy control samples using the way of neural network multi-algorithm and network structure respectively. Neural network algorithm and structure design theory methods are not only the basic research of this thesis, but also supply the theoretical basis for other types of fault diagnosis and self-adaption fault control in finding the optimal network algorithm and optimal structure.Thirdly, because variables characterized by failures are more and its diagnosis is more difficult. Soft failures are often a precursor to hard failures, so the real-time forecasts of soft fault diagnosis are the characteristics and difficult of large complex systems. Combine with the experience and knowledge of experts in the field and base on a fault tree and on-line monitoring data of hoist brake system, the fault samples were designed.The seven kinds of brake system failure modes were classified with SOM network, so the first level of diagnosis was successfully achieved. Meanwhile, the second level of diagnosis was completed for the hydraulic station with BP and Elman network, and the reasons and the degree of the failures were determined. Results of hydraulic station fault diagnosis test show that:BP network, Elman network structure, the input and output of them meet the requirements of fault diagnosis and prediction.Finally, the braking torque NNSOC constant deceleration control theory and design methods were discussed, and the simulation of hoist constant deceleration NNSOC and FC control accuracy were conducted. NNSOC using BP networks optimal intelligent algorithm and optimal network structure emulate to control the accuracy of testing hoist constant deceleration. The simulation accuracy test shows that:response feature of NNSOC intelligently ensure that hoist implement constant deceleration, and the control accuracy of NNSOC constant deceleration is higher than that of the traditional closed-loop, PID, FC.