The Research On Diagnose And Expert System Of Coke Pusher's Hydraulic System

This research is based on the project "Diagnose and Expert System of Coke Pusher's Hydraulic System" which is from Shanghai Baosteel. It aims to solve some problems of the hydraulic system of the coke pusher. The project which is designed to shorten the diagnose time and improve the production efficiency is based on a monitor system and combined with expert system. This method will be used on other important equipments in Baosteel.Three of the five coke pushers in the first-stage construction are imported from Japan and two are designed and set up by Dalian Heavy Machinery Plant. They are all build in the 80's last century. Due to the design defects and wear down, one of them can not work well after several times repairmen. This situation slows down the production efficiency.In order to find out the malfunctions of the system, we use AMESim to build the models of the door-lift system and the electro hydraulic directional valve and change the key parameters during simulation. The research shows that the low pilot pressure of the electro hydraulic directional valve is the main reason why the door-lift system can not work well.There are 16 pressure acquisition points were set to get the data of the working situation. The simulation model has been proved reliable and exact by the monitor system.The hardware system is mainly composed by sensor DAQ card and industrial control unit (ICU). The virtual instrument technology is used to actualize the functions such as collecting and showing data in real time, checking one special channel and querying the history data. The monitor system will provide data for the expert system.The method which is used to build the expert system is combining the ANS and expert system. The neural network is trained to be the inference machine of the expert system. Characteristic parameters which show the working situation are the inputs of the neural network and the malfunctions are the outputs. The knowledge base is composed by expert knowledge and history data. The expert system can make judgment as human expert and learn nonlinear relation between the characteristic parameters and malfunctions, at the same time it can learn new rules and knowledge.