基于小波相关分析框架的大气腐蚀数据挖掘

目的针对大气腐蚀中腐蚀数据复杂多变且不易于分析的问题,提出一种基于小波的相关分析框架,挖掘腐蚀的变化特征及其影响因素。方法首先采用小波功率谱确定腐蚀数据的周期波动特征,然后应用小波相干谱分析腐蚀和环境因素的相关强度、超前滞后及正反相关系,最后利用平均小波相干和显著相干百分比筛选出与腐蚀相关程度较高的环境因素。然后以青岛碳钢腐蚀数据为具体对象,挖掘了温度、相对湿度、降雨和污染物对碳钢初期腐蚀行为的影响。结果在腐蚀初期,碳钢的腐蚀具有很强的日波动特征,温度和相对湿度是导致其波动的主要环境因素;腐蚀在降雨时刻呈现间歇性波动特征,且降雨超前于腐蚀;污染物浓度较低使得其对碳钢的腐蚀影响较弱。结论基于小波的相关分析框架有效展现了复杂腐蚀数据的变化特征及其与环境因素的关系,为深入理解大气腐蚀规律提供了一种新的途径。

Data Mining of Atmospheric Corrosion Based on Wavelet Correlation Analysis Framework

Aimed at the problem that the corrosion data in atmospheric corrosion is typically complex and not easy to analyze, a wavelet-based correlation analysis framework is proposed to mine the changing characteristics of corrosion and its influenc factors. First, the wavelet power spectrum is used to determine the periodic fluctuation characteristics of the corrosion data. Then, the correlation strength, lead-lag, and phase relationship of corrosion and environmental factors are analyzed using the wavelet coherence spectrum. Finally, the average wavelet coherence and the percent area of significant coherence are used to select the environmental factors of high degree relevance to corrosion. Moreover, corrosion data of carbon steel in Qingdao is taken as the specified object for mining the influence of temperature, relative humidity, rainfall and pollutants on the initial corrosion behavior. Corrosion of carbon steel has obvious daily fluctuation characteristics, and temperature and relative humidity are the main environmental factors that cause its fluctuations; corrosion exhibits intermittent fluctuations at the time of rainfall and the rainfall leads corrosion; the lower concentration of pollutants makes its impact on the corrosion of carbon steel weaker. The wavelet-based correlation analysis framework effectively shows the changing characteristics of complex corrosion data and its relationship with environmental factors. It provides a new way for a deep understanding of atmospheric corrosion laws.