離心式冰水機組自動化故障偵測與診斷策略之研究

Study of an Automated Fault Detection and Diagnosis (FDD) Strategy for Centrifugal Chillers

陳俊宏(Chun-Hung Chen)

指導教授 ： 李宗興

國立臺北科技大學/機電學院/能源與冷凍空調工程系碩士班/碩士(2008年)

摘要

離心式冰水機故障種類，可分成急性(hard)及慢性(soft)故障二大類。其中慢性故障是指系統於運轉過程及不當維修保養所造成性能降低之故障，通常這類型故障前期並不容易辨別，造成故障發生後未能及時偵測診斷與通報，使故障被忽略一段時間，進而影響自動控制及性能監視系統的正常作動，更將造成大量能源消耗。有鑑於此，為了提高離心式冰水機長期穩定運轉之考量，發展自動化離心式冰水機自動化故障偵測與診斷之相關研究，就顯得十分重要。
本研究主要目的為發展創新離心式冰水機自動化故障偵測診斷策略。其研究內容如下，首先藉由文獻回顧尋求自動化FDD策略中，具有代表性之性能指標迴歸模式，並利用實驗室數據進行分析與綜合評估。再者藉由離心式冰水機故障模擬實驗，對於本文及其他文獻提出之故障分類器以圖解進行綜合分析，並以數值量化綜合評估本文所提出之創新離心式冰水機自動化FDD策略，於各種故障程度下之故障偵測率與故障診斷率。
研究結果發現僅有複變數多項式迴歸模式對於十一種性能指標之CVRMSE都在5%標準內，為四種模式之中對於此十一種性能指標之預測能力最佳。此外以圖解法結果發現，Wang and Cui故障分類器僅能診斷出三種故障，Reddy故障分類器可診斷四種慢性故障，而本文所提出之故障分類器能成功診斷出七種過程中慢性故障。最後以數值量化表示本文所提出策略之故障偵測與診斷率，結果發現於故障程度4時，其故障偵測率為67%~100%，故障診斷率為67%~100%。

關鍵字

迴歸模式 ； 故障偵測 ； 故障診斷 ； 離心式冰水機

並列摘要

There are two kinds of centrifugal chiller faults, “Hard faults”and“Soft faults”. Soft faults means system operating process and unsuitable maintaining cause low performance fault, it is usually hard to find in the early stage. The fault can’t be detected and promulgated immediately, effecting the automatic control and performance monitor normal operation, causing large energy consumption. In order to raise the operating stability of centrifugal chiller in a long term, studying on an Automated Fault Detection and Diagnosis (FDD) Strategy research appears very important.
The present paper studied on developing innovative Automated Fault Detection and Diagnosis (FDD) Strategy with centrifugal chiller. First, finding representative performance index regression model for Automated FDD Strategy from references and analyzing and estimating with experiment data. Second, by the experiment on the fault of centrifugal chiller, analyzing this paper and other references with chart and quantize the innovative Automated FDD Strategy with all kinds of faults in Fault detection rate and Fault Diagnosis rate.
The result revealed that only multivariate polynomial regression model eleven performance indexes are in 5% of CVRMSE standard. Four models have the best anticipative ability in eleven performance indexes. Otherwise, Wang’s fault classifier can only diagnose three kinds of faults and Reddy’s is four, but this paper is seven. Finally, when the fault degree reaches four, both the fault detection rate and the fault diagnosis rate are between 67~100%.

並列關鍵字

Regression model ； Fault Detection ； Fault Diagnosis ； Centrifugal Chillers

參考文獻

[26] 潘南飛，”工程統計”，全威圖書，2003。

[5] Wang, S.W. and J.T. Cui. 2005b. A model-based online fault detection and diagnosis strategy for centrifugal chiller systems. International Journal of Thermal Sciences. 44(10):986-999.

[9] Braun, J.E. 2003. Automated fault detection and diagnostics for vapor compression cooling equipment. Transactions of the ASME. 125 (3):266-274.

[10] Jia, Y. and T.A. Reddy. 2003. Characteristic physical parameter approach to modeling chillers suitable for fault detection, diagnosis. Transactions of the ASME. 125(3):258-265.

[11] Katipamula, S. and M.R. Brambley. 2005a. Methods for fault detection diagnostics and prognostics for building systems-a review, part I. HVAC and R Research. 11(1):3-25.