25 th Iranian Conference on Electrical Engineering (ICEE2017) 978-1-5090-5963-8/17/$31.00 ©2017 IEEE (Extended Version with more explainations) A Novel Extended Adaptive Thresholding for Industrial Alarm Systems Mahdi Bahar-Gogani Member, IEEE Electrical Engineering Shahid Beheshdi University Tehran, Iran mahdi_bahar@ace.sbu.ac.ir Koorosh Aslansefat CTS-UNINOVA, FCT Campus Universidade Nova de Lisboa (UNL) Lisbon, Portugal k_aslansefat@uninova.pt Mahdi Aliyari Shoorehdeli Senior Member, IEEE Electrical and Computer Engineering K. N. Toosi University of Technology Tehran, Iran aliyari@eetd.kntu.ac.ir Abstract— Decision-making systems are known as the main pillar of industrial alarm systems, and they can directly effect on system’s performance. Meanwhile, thresholding system as a part of the decision-making system becomes a challenging issue for system designers. An excellent thresholding system should be separate normal and abnormal data from sensor measurements completely. It is evident that because of hidden attributes in the measurements such as correlation and nonlinearity, thresholding systems faced wrong separation defining by Missed Alarm Rate (MAR) and False Alarm Rate (FAR). This study introduced a novel extended adaptive thresholding based on mean-change point detection algorithm and shows that it is more efficient than other existing thresholding algorithm in the literature. Number hypothetical and industrial examples are given to delineate the capabilities and limitation of proposed method and prove its effectiveness in an industrial alarm system. Keywords- Adaptive threshold, industrial alarm systems; False alarm probability; Missed alarm probability; Performance assessment; Optimal thresholding I. INTRODUCTION Nowadays, decision-making systems are somehow inseparably attached to industrial plants to improve process safety, efficient operation and reduce cost and energy. One of the most important responsibilities of decision-making systems is notifying to indicate abnormal behavior in industrial systems. As a definition, Alarm Management consists of methods, tools, standards (such as ISA-18.2 and EEMUA-191 [1,2]), and activities that improve system performance by improving the effectiveness of alarm systems. In practice, there are lots of measurements in different areas to monitor physical or environmental conditions of the plant. Then an immense amount configured alarms often make the alarm system complicated; therefore the operators dealt with excessively different alarms that they couldn't handle them immediately. For instance, in the nuclear power plant accident occurred at Three Mile Island in 1979 [3,4]. Furthermore, some troubles exist in the designing path of alarm system like the existence of noise and illegal constructions in the modeling of systems which are caused to alarm nuisance and chattering in notified signals. As a result, we need proper alarm management systems to overcome these drawbacks. The performance of an alarm system usually can be specified by three indices, namely, the false alarm rate (FAR), missed alarm rate (MAR), and averaged alarm delay (AAD) [5]. In practice, because of uncertainty and noisy situation, in normal conditions, sometimes an alarm is raised while no fault occurs in the system, which is called a false alarm. Furthermore, in abnormal conditions, maybe some alarms are not recognized by an alarm system, are called missed alarms [6]. Besides, the alarm delay is measured by AAD. The main purpose of alarm management is the reduction of these three parameters. Some ways to improve the performance of alarm system and reduce nuisance alarm have been investigated in [7], such as deadband, delay-timers and filtering methods. In [8] trade- offs between the detection delay and the false alarm, and between the false alarm rate and the missed alarm rate are discussed by considering EWMA as a filtering method to suggest a proper design of alarm system. In [9] deadband in time and frequency domain is used to reduce chattering. In [10] deadband is used to reduce chattering which is created around fixed set point. This work could not be reduced false alarm as much as missed alarm. Reference [6] represented expected detection delay as a new index to evaluate the performance of alarm systems. In this paper, delay timer and deadband are used to reduce alarm nuisance and false alarm. In [5] performance assessment and design for univariate alarm systems were considered based on FAR, MAR, and AAD. In this paper, mean change detection as a new method to separate normal and abnormal conditions by the use of a simple threshold was examined. This work could not apply on non- Gaussian distribution. Reference [3] represented generalized delay timer to improve the performance of alarm systems comparison with earlier version i.e. conventional delay timer. The results of this paper show that this method could not be reduced FAR and MAR simultaneously. Reference [11] introduced an optimal filter for finding the best alarm accuracy and minimizing false and missed alarm rates in comparison with conventional moving average filters. Reference [4] explained the improved alarm delay-timer method by the use of