類神經網路應用於渦電流檢測訊號分析之研究

Studies of the Neural Network Application on Signal Analysis of Eddy Current Testing

鄭錦智(Jin-Jhy Jeng) ； 曾良雄(Liang-Shyong Jeng) ； 周至宏(Jyh-Hong Chou)

《石油季刊》 40卷1期 (2004/03) Pp. 47-57

https://doi.org/10.30128/JP.200403.0005

摘要

於煉油及石化工廠中，換熱器是一重要且不可缺少之設備，但換熱器管束常因腐蝕而造成洩漏，進而導致非計畫性停炉或工安事故。所以歲修中換熱器管束的腐蝕檢測是很重要的一環，而目前的非磁性換熱器管束以渦電流檢測為主，但因訊號研判須仰賴有經驗的判讀人員為之，而此類人員少之又少。為解決上述困難，本研究首先針對渦電流訊號分析方法進行探討而找出重要影響因子，其次引進人工智慧之類神經網路來學習判讀人員之經驗並做成類神經網路判讀軟體。由此研究所建立的渦電流訊號分析類神經網路學習軟體可直接運用於渦電流檢測的分析，大大地改進以往缺陷型態對缺陷深度判定的準確度影響。利用此軟體的判別分析，可以不必顧慮缺陷型態的影響而仍能評估出正確的深度，因此對於往後的分析人員而言可以縮短判別分析的時間，增進判讀速度且降低人為誤判機會。

關鍵字

類神經網路 ； 渦電流檢測 ； 非破壞檢測 ； 換熱器

並列摘要

The heat exchangers are critical equipments in petrochemical plant. The Corrosion leakage of tube bundle in heat exchangers frequently causes unscheduled shutdown and industry safety accident. Therefore, the corrosion detection of tube bundle is very important during turn-around. Currently, the method used to inspect the non-magnetic tube of heat exchanger is dominated by eddy current testing method, but the correct signal analysis of eddy current testing is relied on the experienced technicians who are rarely available. In order to solve the above issue, the study identified the significant factors through in depth in vestigation of the eddy current signal analysis method, and then introduce the neural network of artificial intelligence to gain the experience exhibited by skilled technician in signal analysis. The studies intend to develop the software of neural network on eddy current signal analysis as well. Using this software will greatly reduce the impact of defect type on accuracy of defect-depth determination. Therefore, it can assist the analysis technician in shortening evaluation time, speeding up signal analysis and reducing judgemental mistakes.

並列關鍵字

Neural network ； Eddy current testing ； Nondestructive testing ； Heat exchanger