【作者】 黄家怿；

【导师】 郭兴蓬； 邱于兵；

【作者基本信息】 华中科技大学 ， 材料物理与化学， 2008， 博士

【摘要】 本文以研究电化学噪声(Electrochemical Noise,EN)分析方法为主线,从预处理方法入手,首先探讨了直流漂移对EN时频分析的影响。然后,通过分析典型点蚀体系产生的噪声电位和电流的分布特征,提出用模式识别(Pattern Recognition,PR)研究数据点(E,log|I|)与不同点蚀状态之间的关系。当待分析数据量很大,或EN时域谱特征随时间变化不显著时,尝试用时域统计量代替E、log|I|作为样本变量,研究局部腐蚀的变化规律。在上述研究的基础上,探讨利用EN研究金属的土壤腐蚀特征,建立土壤腐蚀EN特征谱的识别分析技术,重点在于讨论多种EN解析方法在土壤腐蚀研究方面的优缺点。最后,将EN分析结果与电化学阻抗谱(ElectrochemicalImpedance Spectroscopy,EIS)、极化曲线、失重法、腐蚀形貌和产物分析结果相结合,为研究土壤腐蚀这个复杂系统发生、发展的本征机制提供依据。主要研究成果有:选择m次多项式拟合(Polynomial Fitting,PF)和小波变换(Wavelet Transform,WT)消除EN直流漂移,比较了消除前后以及两种消除方法对EN统计量、功率谱密度(Power Spectral Density,PSD)、绝对能量谱(Absolute Energy Distribution Plot,AEDP)和相对能量谱(Energy Distribution Plot,EDP)的影响,并提出选择合适消除方法的判别标准。结果表明,消除前后各统计量变化不显著(均值除外)。由PSD和AEDP可知,较高方次的多项式和WT会对低频数据产生过度消除,因此推荐选择方次较低(m=1～3)的多项式消除漂移。根据PSD低频平台出现与否或EDP可以判断拟合结果的正确性并选择合适的m值,两种判断结果具有良好的一致性。针对噪声数据量大的特点,提出多段m次多项式拟合法消除漂移。窗口大小p和m值共同影响消除结果和EDP特征,其中p由腐蚀特征决定,以1024≤p≤4096,m≤3为宜。选择Q235碳钢在典型点蚀体系中产生的EN,研究点蚀不同阶段的E、log|I|分布特征。尝试应用k-均值聚类分析(k-means Cluster Analysis,k-means CA)对点蚀过程进行定量分类评价。根据聚类结果,采用判别分析(Discriminant Analysis,DA)建立判别分析方程,对电极在含不同氯离子浓度的混合溶液中产生的EN进行分类,从而达到区别不同点蚀状态的目的。结果显示,k-means CA可将点蚀样本(E,log|I|)分为2类,其中类1表示亚稳态,类2表示稳态。当样本在类1和类2之间随机分布时,说明电极处于从亚稳态向出现稳态宏观蚀点过渡的腐蚀状态。DA不仅能验证CA结果的正确性,更重要的是可以很好地判定来自于同种腐蚀体系的未知EN数据所属类别,从而达到确定其所代表的腐蚀状态的目的。为了有效减少CA/DA样本容量,在对点蚀过程进行定量分类前,先采用主成分分析(Principal Component Analysis,PCA)从基于噪声电位和电流的常用统计量中挑选出了具有显著性分类作用的统计量(?)、(?)、σ_E和σ_I为聚类变量,然后选择分层聚类(Hierarchical Agglomerative Cluster Analysis,HACA)同样可将该过程分为亚稳态、过渡态和稳态三个阶段。最后DA可对来自于同种点蚀体系的未知样本所属类别进行判别。结果表明,HACA与k-means CA结果具有良好一致性,且HACA结果更直观、清晰,无需预先指定分类数目,而k-means CA对点蚀过渡态的出现更为敏感。由于溃疡状腐蚀体系EN时、频谱随时间变化不明显,因此难于通过时域统计量、PSD或EDP清楚划分腐蚀发展的不同阶段以及确定合适的k值。因此,尝试应用基于统计量的PCA/HACA解析EN信号。结果表明,由PCA可确定聚类变量为(?)、σ_E、σ_I,和σ_I/(?),并根据HACA可将溃疡状腐蚀过程分为高速萌发期、横向发展期和纵向发展期三个阶段,该分类结果与电极表面腐蚀形貌具有较好的一致性。采用EN技术研究X70钢在30℃库尔勒原土(含水量1.04%)及35℃低含水量库尔勒土壤(含水量1.04～3.12%)中初期(0～7d)的腐蚀行为,其中PR、PSD和WT用于研究EN与土壤腐蚀状态之间的相关性,并将解析结果与EIS、极化曲线、失重法和腐蚀形貌与产物分析结果进行了对比。结果表明,基于EN统计量的HACA可用于解析土壤EN数据。根据PCA/HACA,可将30℃库尔勒原土的初期局部腐蚀过程分为不稳定萌发期、快速发展期和稳定发展期三个阶段:将35℃低含水量库尔勒土的腐蚀过程分为快速发展期和稳定发展期两个阶段;并对35℃不同含水量库尔勒土的腐蚀程度进行了分类和比较。聚类分析结果与EN时域谱、电流EDP、EIS和腐蚀形貌特征相吻合,且更加精确。更多还原

【Abstract】 In this paper,researching the methods of analyzing electrochemical noise（EN） was considered as the main line.The pretreatment for EN raw data before other time/frequency domain analysis were compared firstly,and then the influence of DC drifts on the analysis results was researched.After analyzing the distribution characteristics of the potential and current noise（E and log|I|） from a typical pitting corrosion system,pattern recognition（PR） was proposed to study the relationship between data points（E,log|I|） and different pitting states.In order to reduce the sample number for the PR,some statistical parameters were used as variables in place of E and log|I|,and its validity was also verified in the same system.For ulcerous corrosion which has not characteristic EN time records,the PR with some EN statistical parameters as variables was applied to study the relationship between EN and its corresponding different corrosion states.Based on the above analysis,EN measurement was proposed to research the characteristics of the soil corrosion to establish its recognition technology.It focused on the EN analysis methods in soil corrosion research. Finally,the characteristics and mechanism of the complex soil corrosion were studied by EN,electrochemical impedance spectroscopy（EIS）,polarization curve,loss-weight method and corrosion morphology observation,and good agreement was obtained.The main research results were shown as following:Polynomial fitting（PF） and wavelet transform（WT） were chosen to remove trends of EN from typical pitting system.The effects of trends on statistical parameters,power spectral density（PSD）,absolute energy distribution plot（AEDP） and relative energy distribution Plot（EDP） were discussed.Furthermore,the indication of selecting the appropriate method was put forward.The results showed that the common statistical parameters except mean value would not change distinctly.Based on PSD and WEDPs,it can be found that the higher order polynomial and WT would reduce the low frequency signals more heavily,so lower order（m=1～3） polynomial was recommended.The appearance of plateau in low-frequency part of PSD plots or EDP could be regarded as the indication of the best selection of m.For large numbers of EN data,PF using window technique has been applied to remove trend.The order and window size influenced corporately the removing results and the EDP characteristics.In order to attenuate the low-frequency components without damage the useful information,the lower polynomial order（no bigger than 3） and appropriate size（between 1024 and 4096） which was determined by characteristics of EN fluctuations should be selected.In a typical pitting system of Q235 low carbon steel in 0.50mol/L NaHCO₃+NaCl solutions,a new method was presented to analyze EN data and identify its corresponding pitting states.The proposed method is based on k-means cluster analysis（k-means CA） and discriminant analysis（DA）.Firstly,E and log|I| were determined as the variables for clustering.Then,data points（E,log|I|） of the EN groups from different pitting states were classified by k-means CA to two clusters,which relate to the metastable state（Cluster 1） and stable state（Cluster 2） respectively.When a group of（E,log|I|） data points were dispersed stochastically into two clusters,it relates to the intermediate state that was defined to describe the transformation from the metastable pitting to stable pitting.Based on the obtained clustering results,a discriminant function（s） was established to discriminate the ungrouped EN data from the similar pitting processes and thus its corresponding pitting state could be determined by the cluster distribution result.In order to reduce data number of the cases reasonably without loss of useful information,principal component analysis（PCA）,hierarchical agglomerative cluster analysis（HACA） and DA were applied for analyzing EN statistical parameters from the similar pitting system as described as before.Firstly,according to the PCA results,the EN mean value（（?） and（?） ） as well as standard deviation（σ_E andσ_I） were determined as the descriptors for clustering.Then,using the selected four statistical parameters as variables, the cases from different pitting states were classified by the HACA to three clusters,which relates to the metastable state,intermediate state and stable state,respectively.It shows a good agreement with the classification obtained from k-means CA with E and log|I| as variables.Based on the cluster results,the pitting states of the ungrouped data points from the similar pitting processes also can be distinguished according to the established discriminant function（s）.The time and frequency domains of the EN from 0.5mol/L NaCl ulcerous corrosion would not change with the development of the corrosion time,so it is difficult to distinguish the different corrosion states by statistical parameters,PSD or EDP and then to determine the proper k value.Thus,the PCA/HACA was proposed to analyze this kind of EN signals.Based on the PCA,（?）,σ_E,σ_I andσ_I/（?） were determined as the descriptors for clustering.And then,using them as variables,the cases from the ulcerous corrosion would be classified by the HACA to three clusters,which relates to the quick germination state,horizontal development state and vertical development state.Good agreement between classification and corrosion morphology was gained.According to the above analyses,EN technique was applied to study the corrosion behaviors of X70 steel in 30℃Xinjiang Ku’erle saline soil（moisture content 1.04%） and 35℃Ku’erle soil with low moisture（moisture content 1.04～3.12%） during the initial corrosion period（0～7d）.The obtained EN data were analyzed by the PCA,HACA and WT. After the PCA applied,σ_E,（?） and（?） were considered as the descriptors to characterize EN distributions.Then,using them as variables,for the 30℃Ku’erle soil with 1.04% moisture,three different local corrosion states can be classified clearly by the HACA, including instable germination state,quick development state and stable development state. For the 35℃Ku’erle soil with low moisture,only two corrosion states were classified,i.e. quick development state and stable development state.Furthermore,the extent of the local corrosion in different moisture soil also can be differentiated.According to the current EDP, EIS and corrosion morphology,the characteristics of the different corrosion stages were further proved.The results showed that the PCA/HACA method should be successful for the analysis of the EN signals from the soil corrosion system.更多还原