A study of using back-propagation neural model in automatic lubrication installation for the feeding system of computer numerical control machine tool

Abstract

Nowadays, the feeding systems of computer numerical control machine tools are lubricated by periodic oil supply or fixed stroke, the lubrication is insufficient in the case of high load and high-speed movement, and the lubrication is excessive during finishing and low feed rate. This study discusses the optimum lubrication timing of the feeding system. When the feeding system is moving, the servomotor torque value and current, accuracy, and oil film thickness are measured by sensors. Moreover, the lubrication characteristic model is validated and built by using the sensor values, and the optimal lubrication state estimation is obtained by using the back-propagation neural model. Then analyzed and feedback to the machine tool controller, to intelligent the lubrication system. According to the test, when the feed rate is increased by 5 m/min, the friction coefficient increases with rate, increasing the output of the frictional value of the work table by 6.90%. When the load is increased by 175 kg, the friction coefficient decreases with load, reducing the frictional value output of table movement by 6.71%. In the relationship between oil film thickness and current, the accuracy difference between the prediction and actual test results is less than 10%; in the case of the same accuracy, the oil supply frequency is reduced by 80%, and environmentally friendly machine tool has been achieved.