Neural Network Prediction Model Of Film Thickness Of Micro-arc Oxidation On Magnesium Alloys

Magnesium alloy micro-arc oxidation (MAO) is a complicated process generating ceramic layer under the common function of the thermal chemical, plasma chemical and electrochemical. The parameters of the MAO coatings on magnesium effect is not clear. The film thickness as a measure of micro-arc oxidation film layer, one of the most intuitive and important indicators, it directly affects the performance of micro-arc oxidation film layers on hardness, wear and corrosion resistant etc. But, the quality of the MAO coatings can only be to ensure by the stable various parameters and finished inspection, its limitations are obvious. And because of the complexity of micro-arc oxidation, leading to the study about the effects of various parameters on the quality is difficult. Based on the artificial neural network theory, establish the mapping model of the main parameters to the film thickness, using of some easily measured quantity (current, voltage, temperature, etc.) to reflect the changes in film thickness to achieve the monitoring of film thickness, thus providing a way for the application of the neural network theory in dynamic monitoring of the micro-arc oxidation coating quality. The paper mainly includes the following:The micro-arc oxidation experiments are carried out based on the experimental platform for micro-arc oxidation. By analyzing the characteristics, the law of the spark discharge and the surface morphology , the structure of the film of the different stages of micro-arc oxidation, and combining with the theory, the mechanism of micro-arc discharge and the growth mechanism of the MAO coatings will be proposed, that provide the theoretical basis to explain the impact of law the micro-arc oxidation process parameters on the thickness.With the improvement of power supply voltage, micro-arc oxidation process always showed obvious characteristics of the 3 stages: anodic oxidation, micro-arc oxidation and large arc discharge stage.The initial film will be formated during anodic oxidation, but the large arc discharge will lead to surface ablation, surface treatment failure. In the micro-arc oxidation stage, with the increases of time and voltage, the growth of micro-arc oxidation film is divided into three periods that are early, middle and the late period.which can be defined by the thickness of the film or deposition rate. There are a lot of shallow conductive channels in the early stage, so it puts many small arc spots, the deposition rate is lower. In the medium , shallow channels gradually closed, it puts arc spots increase gradually and thin, when deposition rate is the highest.In the late stage , the discharge of deep conductive channels will be the main part,the arc spot will be larger and less , the deposition rate will also decreases.The Micro-arc discharge mechanism and model of micro-arc oxidation which was proposed based on the arc physical theory , divided an independent micro-arc discharge into four processes: electrolysis, discharge, oxidation and cooling. Based on the proposed mechanism of micro-arc discharge and the growth mechanism of the MAO coatings, the paper analyzes the factors of the thickness of MAO. It is mainly affected by the electrolyte, power output characteristic and the processing time .On the basis of a large number of micro-arc oxidation experiment, it discuss the law of each factor on the thickness, and based on the correlation analysis results it also determine the eigenvector of neural network.Using the MATLAB neural network toolbox as a platform, the micro-arc oxidation film thickness mapping model was established,which was based on the BP network and RBF network. Through the analysis of the disadvantages of each network model, a suitable network modle was finally confirmed for real-time monitoring of the micro-arc oxidation film thickness.Results show that using the method of neural network to realize the dynamic monitoring of film thickness is feasible.