Practical Analysis of Neural Network for Machine Tool Structure Control System

In the thermal displacement detection, the high-frequency reflective eddy current sensor is used, and the frequency is generally in the order of MHz. Under normal circumstances, the first-stage frequency divider is added to directly detect the frequency by the single-chip microcomputer. The control system takes the single-chip microcomputer 89C51 as the core, and the heat source and the cold source are controlled by P113 and P114. The purpose of using photoelectric isolation is to improve the anti-interference ability of the whole system. In the field control, there may be interference caused by the MCU program. In this case, in order to continue the normal operation of the microcontroller, a watchdog circuit composed of the CD4060 chip is added. The eddy current sensor is a non-contact sensor. It has many advantages, such as large measuring range, small volume, strong anti-interference ability, and is not affected by medium such as oil and water between the measured object and the sensor, but it also has disadvantages, and the output frequency is not only related to distance, but also The material, shape and surrounding environment of the measured object are related. Therefore, when measuring accurately, it must be subjected to online calibration. The data obtained by the calibration is processed and input by the keyboard and placed in the E2PROM 93C46. The 93C46 has both an online rewriting function and a feature that the power down information is not lost. Therefore, after the entire system is scaled, it is not required to reset the parameters due to shutdown. In order to further verify the effectiveness of the neural network algorithm, the experimental calibration values ​​of the other three sets of sensors are given. It is compared with the estimated values ​​obtained by the neural network algorithm. Similarly, the experimental data is scaled in the y direction and continuously learned by the learning algorithm. When the number of iterations is 50000, it is found that W0=2170345, W1=-2110629, W2=4135136, W3=-3119801, ERMS=01005095. After the proposal of the spindle thermal displacement monitoring system is put forward, the performance is stable and reliable after on-site debugging. The computational processing and comparison of the results of the neural network algorithm by the computer show that the neural network algorithm has application value in the online calibration of the eddy current sensor, and it is an extremely effective online calibration processing method.

(Finish)

Brass Machining

Cnc Machining,Plastic Machining,Heavy Machining

SOMAX ENTERPRISE CO., LTD. , http://www.tzdiecast.com