行政院農業委員會台南區農業改良場   研究彙報第67號

 

適應性卡爾曼濾波器於感測器訊號雜訊消除及錯誤偵測之應用

李健、鍾瑞永、楊清富

摘  要

李健、鍾瑞永、楊清富。2016。適應性卡爾曼濾波器於訊號雜訊消除及錯誤偵測之應用。臺南區農業改良場研究彙報67:62-72。

本研究首先介紹卡爾曼濾波器疊代演算的基本矩陣式及其直覺上容易理解的含意,並將一般通用的矩陣式卡爾曼濾波器推導至適合濾除訊號雜訊的純量演算式。此簡單演算式除了可以調整消除雜訊的頻帶外,亦能藉由卡爾曼增益值的突增來偵測訊號的異常。本文將演算式實際應用於溫度感測器TMP37 上,並探討實驗訊號的取樣時間間距及調整濾波器參數對運算處理後輸出訊號品質的影響,此外亦以真實方式產生訊號錯誤來測試演算式錯誤偵測的性能。最後探討感測器訊號品質及可靠度對農業環境控制系統的潛在重要性與本研究在此方面的未來應用。

現有技術:目前處理訊號量測雜訊問題,一般皆使用常見的移動平均法,或者重複量測多次後再進行多筆數據平均合併成一筆,前者演算處理效率低且不適合微控制器現場即時演算,後者取樣時間長無法處理快速變動的數據。

創新內容:將通用的矩陣式卡爾曼濾波器推導至專門濾除訊號雜訊的純量演算式,使其可以在一般微控制器上即時執行感測訊號雜訊消除及錯誤偵測。

對產業影響:農業環控系統的控制決策需要依賴感測器來提供資訊,雜訊干擾會使控制決策搖擺不定,造成系統機械輸出啟動關閉頻繁使致動元件壽命耗損,或者在感測器訊號錯誤的情況下作出不合理決策而導致農業生產上的嚴重損失。應用本技術將可提供可靠的資訊給農業環控系統,可提高生產效率及延長機械原件壽命。

關鍵字:適應性、卡爾曼濾波、感測器、雜訊、錯誤偵測
接受日期:2016年1月21日


An Application of Adaptive Kalman Filtering
on Noise Elimination and Fault Detection of Sensor Signal

Li, C., J. Y. Chung and C. F. Yang

Abstract

An introduction to iterative algorithm of matrix-form Kalman filter and its intuitive meaning are first given and then a scaler Kalman filter algorithm suited for noise elimination of sensor signal is derived from the matrix-form one. The derived algorithm provides the advantage of convenient adjusting of filtering band and it can also be used to detect the signal anomalies by monitoring the sudden rise of Kalman gain. The algorithm is applied to the temperature sensor TMP37 and the effects of sampling interval and adjustable parameter on the response and filtering of optimal estimate are discussed. In addition, the signal anomalies are intentionally generated to demonstrate the fault detection capability. Finally the applications to environmental control for agricultural productions and their potential improvements are discussed.

What is already known on this subject?
The traditional ways to eliminate noises in measurements mostly rely on the moving average or the average of multiple samples, the former lacks efficiency and cannot work well on generic microcontroller, the later responses sluggishly and is hard-pressed to tackle fast-changing data.

What are the new findings?
A scaler Kalman filter algorithm suited for noise elimination of sensor signal is derived from the matrix-form one. The derived algorithm provides the advantage of convenient adjusting of filtering band, and it can also be used to detect the signal anomalies.

What is the expected impact on this field?
The environmental control system for agricultural productions relies on the information provided by sensor signal to make decisions. If the signal is severely interfered by noise, their actuators will frequently turn on and off and the life of device will be reduced. On the other hand, the sensor faults can make the wrong decision for control go on all the time and cause irreversible economic cost. The above problems can be effectively solved or relieved by applying the technique described in this research.

Key words: Adaptive, Kalman filter, Sensor, Noise, Fault detection
Accepted for publication: January 21, 2016


    回研究彙報目錄