Analysis And Research The Parameters Of Spark Of Wire Cut Electrical Discharge Machining

Wire cut electrical discharge Machining (WEDM) is using the moving wire (copper or clamp) as an electrode. Impulse voltage loading into the workpiece and metal wire generate the spark and cut forming. In the wire cut electrical discharge machining, the process parameters are interrelated constraints, even contradictory, coupling strongly. It is difficult to describe the processing model using mathematical formulas. WEDM main processing stage is to generate the spark between the wire and wokpiece throw pulse power turned on. But in this stage it appeared generating no spark phenomenon. This paper focuses on the amount of processing time spark area and its variation and collecting data to analyze and predict learning. The predicted results can developed as a process control strategy experts system.Data clustering analysis is studying the degree of similarity data and finding large amounts of data implicit, unknown, potential value of the information or pattern. It is the starting point of the data processing or analysis. Traditional machine learning data classified according to the presence or absence determines the classification of the type of the basic division of supervised learning and unsupervised learning. Cluster analysis is an unsupervised learning process. Semi-supervised clustering algorithm can effectively solve the problem of clustering into local optimization. In this paper using metric pairwise constraint optimization K-Means clustering, that using part of the marked data and attribute matrix to improve the cluster center search. Selecting the UCI machine learning data sets compare with the existing clustering analysis. The experimental results show that the algorithm polymerizing ability is excellent while in low dimensional data set.The purpose of using semi-supervised clustering is to optimize searching cluster center in Redial Basis Function (RBF) neural network centric vector selection. RBF neural network is expert in complex function or model which is difficult to determine the effect of having a higher approximation. Theoretically it can approximate continuous function with arbitrary precision. RBF neural network learning center vectors flaw is difficult to determine. The traditional way is determining by output. The Output model has high fitting accuracy and the low prediction accuracy.This paper focus on cutting sparks area and processing parameters affecting the establishment of a two-step prediction neural network learning model for WEDM. We proposed semi-supervised algorithm optimization RBF network. Semi-supervised cluster using data analysis the influence of system parameters and optimize the center vector and width selection. RBF neural network sparks area and processing efficiency prediction learning. In this paper, semi-supervised clustering RBF neural network is better in terms of the ability of fit or predictive ability than traditional neural network. The prediction error is relatively small, the higher the overall amount of anticipation spark success rate.