The Recognition Of Working State Of Oil-Well And The Development Of Its Sensor

In the oil field of our country, there is a problem of low productive efficiency of oil extraction commonly. To prevent pumping system stopping working, working without efficiency and being destroyed , so it is very important to monitor oil-well.The thesis dissertates the process of developing displacement-load sensor which collects displacement data and load data to form the indicator diagrams, researches the means based on RBF Neural Network of recognizing fault types, designs parts of hardware and software of the mainframe of the oil-well monitor and control system ,and debugs the entire oil-well monitor and control system.Based on the basic requirements and function of the displacement-load sensor, the design adopts Single-Chip 8-Bit Microcontroller-P89C51RC2 manufactured in an advanced CMOS process and belonged to the 80C51 microcontroller family to make the P89C51RC2's circuit less, simple and more reliable.Monitoring electric parameters of oil well is the basic function of the monitor and control system. The design adopts AC sample method to get 3-phase voltage and current samples. The design computes three basic electric parameters according the sample values and monitors the voltage and current of pumping machine with keyboard and LCD.Using displacement-load sensor to collect displacement and load data and adopting micro-power wireless communication module to transmit the load and displacement data to main part of the oil-well monitor and control system by serial port to form the indicator diagrams.The key to give the oil well effective monitor and control is monitoring the data of the indicator diagrams. According to this we can get fault types. Whereas the disfigurement and limitation of the fault recognition methods at present and the application background of Neural Network, the design uses the RBF Neural Network to recognize fault types. To solve the problems of memory of F149 and training speed or real time recognitions, the design adopts several previous coefficient of DCT to describe the input samples of the Neural Network.