Reasoning of atmospheric corrosion level under missing data based on CMAC and Bayesian network

摘要

According to Standard ISO9223, the level of atmospheric corrosion is determined by three factors including chloride ion, SO2 and time of wetness. In practice, missing of one or more of these data is very common, increasing the difficulty in accurate determination of the atmospheric corrosion level. In order to overcome such problem, we used Cerebellar Model Articulation Controller (CMAC) for missing partial data and Bayesian network for missing all data occasion. By obtaining the relationship between the missing parts and the other attributes of data, the correlation model was established to complement the missing data. Consequently, the level of atmospheric corrosive factors could be determined. Simulation results show that by using the method, the reasoning accuracy rate under the conditions of missing the data of chloride ion, SO2 and all three categories respectively reached 93.3%, 83.3% and 80%. The problem of missing atmospheric corrosive factors was thereby solved to some extent.