The On-line Monitoring System Of Micro-hole Drilling With Fuzzy-Neural Network

Nowadays, with the fast development of technology, many advanced products are developing toward the direction of micromation and integration, so, the application of micro-holes in various products are more and more comprehensive, and machining techniques are more and more important. In the micro-hole processing methods, the drilling process can obtain the good processing quality and the high production efficiency. However, there is some fatal weakness in micro-drilling. The micro-drill has low intensity, break off easily when it wore out or it is difficult to remove cutting scrap; and the life of micro-drill is too dispersive to estimate it. Once micro-drill aiguilles breaks off, it is hard to take out it from the work pieces, thereby usually causing the work piece waste. So, during the course of micro-drill's working, this fatal weakness of micro-drill is easy to break off and severely influences machining efficiency and the increase of automatization level.The force increase in micro-drilling is basic reason to break off of the micro-drill. But the force signal of the micro-hole drilling is extremely difficult to gain and it needs to carry on the transformation to the existing processing equipment structure. This not only bring wastes much, moreover, also possibly affects the Mechanics performance of the processing system. Therefore, this article proposed the plan that the on-line monitoring system of micro-hole drilling with fuzzy-neural network, and has carried on the correlation theories and the technical research.This article developed an on-line monitoring system of micro-hole drilling with fuzzy-neural network. This system is composed by the hardware system and the software system. The hardware system includes the automatic drilling machine which are changed from a manual system, the Hall current sensor, the data acquisition card NI 6013, SCM control system; The software system is the programs based on the environment of the LabVIEW software, achieves to voltage signal gathering, the filter, the memory, the normalization processing, the monitoring, the enactment of monitor value and the serial communication to the SCM control system.In this paper, the increments of tri-phase current of the main axle electromotor are collected during micro-hole drilling. Drill state after drilling and the increment's maximum of tri-phase current are recorded every time. Then, these data are normalized and input to network in order to train fuzzy-neural network, consequently gained three matrixes of monitoring system.The fuzzy-neural network is an improved network relative to ecumenical fuzzy-neural network. If the ecumenical fuzzy-neural network has three inputs and every input has five fuzzy aggregations, the network's second layer must have fifteen crunodes and the third layer must have one hundred and twenty five crunodes. The network structure will spend much time to calculate network's output. In this experiment, there are three inputs and one output, and the acquired data are calculated timely, so the ecumenical fuzzy-neural network is not the same with this experiment. It needs a simpler and faster improved network. Through lots of experiments, we draw a conclusion that the tri-phase current augment of the main axle electromotor goes with the augment of drill-hole number and abrasion of micro-drill at the same time. So, we choose the improved fuzzy-neural network where the crunodes in the third layer are fewer than those in the ecumenical fuzzy-neural network. Moreover, the improved fuzzy-neural network has obvious physical signification. That is that all the increment's maximum of tri-phase current of the main axle electromotor is in the same fuzzy aggregation.Then, after inputting the increment's maximum of tri-phase current of the main axle electromotor into network, the output of this monitoring system can be obtained. The output is compared with the datum in the sample of training network that reflects drill state, so the monitoring threshold value can be attained. In this experiment, the monitoring threshold value is 0.8. Accordingly, a complete on-line monitoring system of micro-drilling with fuzzy-neural network is built up.On the base of the above work, the experiment of micro-hole drilling is carried on, which is based on the on-line monitoring system of micro-hole drilling with fuzzy-neural network. The results indicated this method achieves the anticipated effect to avoid the drills breaking off, and enhance the drills use rate basically.