Sensor Fault Detection in IoT System Using Machine Learning
0 likes65 views
This document summarizes research on using machine learning techniques for sensor fault detection in IoT systems. The researchers collected temperature and humidity data from a DHT22 sensor and injected drift faults using an Arduino microcontroller. They extracted time-domain features from the normal and faulty signals and used them to train classifiers like artificial neural networks, support vector machines, naive Bayes, k-nearest neighbors, and decision trees. The trained models detected drift faults in the sensor output in real-time on an ESP8266 device. Support vector machines and artificial neural networks achieved the best performance based on accuracy, recall, F1-score metrics. The lightweight system demonstrates potential for low-cost, real-time sensor fault detection using machine learning.
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 07 | July 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 2579
V. FUTURE SCOPE
For future work, an additional capable single-board pc is
often used rather than an esp8266, which might handle
additional complicated operations, and numerous sensors,
like measuring an instrument or a pressure device are often
used rather than a temperature device for various sorts of
alternative device faults. Also, a fault diagnosis and
prognosis are typically done following the data-driven
approach.
REFERENCES
[1] Z. Gao, C. Cecati, and S. X. Ding, “A survey of fault
diagnosis and fault- tolerant techniques-Part I: Fault
diagnosis with model-based and signal- based
approaches,” IEEE Trans. Ind. Electron., vol. 62, no. 6,
pp. 3757- 3767, Jun. 2015.
[2] D. Park, S. Kim, Y. An and J. Jung, “LiReD: A Light-
Weight Real-Time
Fault Detection System for Edge Computing Using LSTM
Recurrent Neural Networks,” MDPI Sensors., 18,2110;
DOI: 10.3390/s18072110, Jun. 2018.
[3] J. Tian, C. Morillo, M. H. Azarian, and M. Pecht, “Motor
bearing fault detection using spectral kurtosis-based
feature extraction coupled with K-nearest neighbor
distance analysis,” IEEE Trans. Ind. Electron., vol. 63,
no. 3, pp. 1793-1803, Apr. 2016.
[4] T. W. Rauber, F. De A. Boldt, and F. M. Varejao,
“Heterogeneous feature´ models and feature selection
applied to bearing fault diagnosis,” IEEE Trans. Ind.
Electron., vol. 62, no. 1, pp. 637-646, Sep. 2015.
[5] O. Castro, C. Sisamon, and J. Prada, “Bearing fault
diagnosis based´ on neural network classification and
wavelet transform,” in Proc. 6th WSEAS Int. Conf.
Wavelet Anal. Multirate Syst., Bucharest, Romania, pp.
22-29, Oct. 2006.
[6] B. Samanta, “Gear fault detection using artificial neural
networks and support vector machines with genetic
algorithms,” Mech. Syst. Signal Process., vol. 18, no. 3,
pp. 625-644, 2004.
[7] B. Sreejith, A. K. Verma, and A. Srividya, “Fault
diagnosis of rolling element bearing using time-
domain features and neural networks,” in Proc. IEEE
Region 10 3rd Int. Conf. Ind. Inf. Syst., vol. 1, pp. 1-6,
Sep. 2008.
[8] Y. Wang, J. Xiang, R. Markert, and M. Liang, “Spectral
kurtosis for fault detection, diagnosis and prognostics
of rotating machines: A review with applications,”
Mech. Syst. Signal Process., vols. 66-67, pp. 679-698,
Apr. 2016.
[9] Q. Xiao, Z. Luo, and J. Wu, “Fault detection and
diagnosis of bearing based on local wave time-
frequency feature analysis,” in Proc. 11th Int. Conf.
Natural Comput. (ICNC), pp. 808-812, 2015.
[10] J. L. Yang, Y. S. Chen, L. L. Zhang, and Z. Sun, “Fault
detection, isolation, and diagnosis of self-validating
multifunctional sensors,” Rev. Sci. Instrum., vol. 87, no.
6, p. 065004, 2016.
[11] R. Dunia, S. J. Qin, T. F. Edgar, and T. J. Mcavoy,
“Identification of faulty sensors using principal
component analysis,” Process Syst. Eng., vol. 42, no. 10,
pp. 2797-2812, 1996.
[12] J. Kullaa, “Detection, identification, and quantification
of sensor fault in a sensor network,” Mech. Syst. Signal
Process., vol. 40, no. 1, pp. 208221, Sep. 2013.
[13] Y. Yu, W. Li, D. Sheng, and J. Chen, “A novel sensor fault
diagnosis method based on modified ensemble
empirical mode decomposition and probabilistic
neural network,” Measurement, vol. 68, pp. 328-336,
May 2015.
[14] S. U. Jan, Y.-D. Lee, J. Shin, and I. Koo, “Sensor Fault
Classification Based on Support Vector Machine and
Statistical Time-Domain Features,” IEEE Access, vol. 5,
no. 1, pp. 8682–8690, 2017.
[15] S. Haykin, Neural Networks: A Comprehensive
Foundation, 2nd ed. Upper Saddle River, NJ, USA:
Prentice-Hall, 1994.
[16] A. Geron, Hands-On Machine Learning with Scikit-
Learn & TensorFlow. O’Reilly, 2017.
[17] Scikit-learn library. Available online scikit-learn org
(accessed on 5 Sep 2019).](https://image.slidesharecdn.com/irjet-v9i7532-221027070344-f9d3690a/85/Sensor-Fault-Detection-in-IoT-System-Using-Machine-Learning-4-320.jpg)
Sensor Fault Detection in IoT System Using Machine Learning
- 1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 09 Issue: 07 | July 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 2576 Sensor Fault Detection in IoT System Using Machine Learning Mahadev Lot1, Sachin Belekar2, Pranay Redekar3, Prof. Sejal Shah4 1,2,3 Students, Department of Electronics Engineering, KJSIEIT, Mumbai, India 4 Professor, Department of Electronics Engineering, KJSIEIT, Mumbai, India -------------------------------------------------------------------------***------------------------------------------------------------------------ Abstract—From good industries to good cities, sensors within the present plays a vital role by covering an oversized range of applications. However, sensors get faulty typically resulting in serious outcomes in terms of safety, economic price and dependability. This paper presents associate analysis and comparison of the performances achieved by machine learning techniques for real- time drift fault detection in sensors employing a low-computational installation, i.e., ESP8266. The machine learning algorithms underneath observation embrace artificial neural network, support vector machine, na¨ıve mathematician classifier, k- nearest neighbors and call tree classifier. the info was noninheritable for this analysis from digital relative temperature/humidity detector (DHT22). Drift fault was injected within the traditional information exploitation Arduino Uno microcontroller. The applied math time- domain options were extracted from traditional and faulty signals and pooled along in coaching information. Trained models were tested in a web manner, wherever the models were wont to sight drift fault within the detector output in period. The performance of algorithms was compared exploitation exactness, recall, f1-score, and total accuracy parameters. The results show that support vector machine (SVM) and artificial neural network (ANN) outmatch among the given classifiers. I. INTRODUCTION Modern technologies like Industrial systems or wireless sensing element networks (WSNs) typically comprises many sensors which will be deployed in comparatively harsh and complicated environments. Natural factors, magnetism interference, and lots of different factors will have an effect on the performance of the sensors. once the sensing element becomes faulty, it’s going to utterly stop generating signals or turn out incorrect signals. It are often jumping between traditional and faulty state unstably. to enhance safety, information quality, shorten reaction time, strengthen network security and prolong network time period, several studies have targeted on sensing element fault detection. A fault are often expressed as associate uncommon property or behavior of a system or machine. Studies are disbursed chiefly since the Eighties for the detection and identification of defects in industrial facilities, i.e., physical-based or mathematical. These approaches were restricted to specific environments and conditions. it’s tough to see variant model parameters thanks to system complexities. to beat these limitations, data-driven approaches victimization machine learning techniques are projected, that analyses information to develop the simplest models. The models essentially use historical information to seek out hidden patterns and determine expected outcomes. As fashionable systems have become complicated, previous approaches have become tough to implement. On the opposite hand, the information-driven models are often developed to adequately approximate real systems supported the collected data. The fault happens in actuators, sensors or the other mechanical systems. within the past, algorithms for fault detection in rolling components of machines are explored in an exceedingly large range of studies news economical results. However, sensors conjointly fault oftentimes resulting in serious consequences in terms of safety and operation. Therefore, sensing element fault detection is extremely vital to make sure the security and responsibleness of systems. many studies with time have mentioned variety of faults, which might presumably occur in sensors. However, in the present study the most occurred sensor fault is focused, i.e., drift fault, which can be defined as follows: A. Drift Fault The output of the sensing element keeps increasing or decreasing linearly from traditional state. associate example of traditional and faulty signal.
- 2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 09 Issue: 07 | July 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 2577 Lately, machine learning techniques like support vector machine (SVM) and neural network (NN) has gain eminence in fault detection and diagnosing for rolling parts and sensors. Techniques for bearing fault detection and sensing element fault detection ar uniform, however, the signal characteristics of sensing element faults ar totally different from the rolling parts. Hence, exploitation similar options for each doesn’t guarantee an equivalent accuracy in results. The data needed for this analysis is obtained from the temperature/humidity sensing element (DHT11). The signals obtained from the sensing element through Arduino Uno microcontroller would be sent to ESP8266 for coaching. The drift fault is simulated within the signaling from the sensing element. applied mathematics time-domain options ar extracted from the signal. knowledge is trained exploitation classifiers, elaborate mentioned in fault detection strategy section. For testing, arbitrarily drift fault is generated exploitation Arduino Uno microcontroller and is given to many classifiers on ESP8266 in an internet manner to look at the results for fault detection. Figure two shows the applied system model for fault detection within the gift study. 1) Light-weight System: Low procedure grid (esp8266) used for fault detection with a DHT11 temperature sensing element. ESP8266 is delineate as alittle all-purpose singleboard laptop running chiefly on Debian OS supported the UNIX operating system kernel. within the future, these little all-purpose computers is wide utilized in industries for AI applications. These systems area unit low-cost, simple to deploy, needs less area with good procedure powers. 2) Real-Time Fault Detection: : The proposed system adopted the machine learning approach, that learns from the collected information and detects detector faults. a sign from the temperature detector is given to ESP8266 in a web manner. Algorithms square measure trained victimization scikit-learn, that may be a far-famed machine learning library for Python artificial language. Trained classifiers in period of time square measure accustomed observe faults within the detector. II. FAULT DETECTION METHODOLOGY A. Data-Driven Approach The data-driven approach has been applied in several real- world applications to develop associate degree correct model. an oversized variety of techniques within the datadriven approach are applied to resolve fault detection issues. Statistically primarily based strategies and people supported AI techniques area unit completely different strategies within the data-driven approach. Figure four illustrates the approach towards fault detection, once information assortment and have extraction, intelligent detection are going to be used. B. Machine Learning for Classification Classification may be a supervised machine learning approach, which may be outlined as a way of categorizing some unknown things into a distinct set of categories. during this work, the binary classification approach is employed, that distinguishes between 2 categories, i.e., traditional and faulty. a number of the classification techniques employed in this work ar explained as follows: 1) Support Vector Machine (SVM): Developed within the Nineteen Seventies, SVM deals with the conception of applied mathematics learning theory and within the field of machine learning, exactly for fault detection and classification, SVM is one in all the good-performance
- 3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 09 Issue: 07 | July 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 2578 algorithms deals primarily with two-class classification issues. Linear line or hyperplane is generated as a call boundary for classification tasks between datasets of 2 categories. the closest knowledge points to the hyperplane, that impart construction of the hyperplane ar known as support vectors. In this analysis, binary-class SVMbased classifier with linear kernel perform is employed to research the results for detector fault classification. the value parameter C was set to default (C=1). III. SIMULATION RESULTS A. Data Acquisition and Feature Extraction The data were noninheritable from the digital relative Temperature/Humidity sensing element DHT11 developed by Adafruit Industries, on the market during a 4-pins package. the information were obtained serially from a sensing element victimisation Arduino Uno microcontroller through Arduino’s IDE and PLX-DAQ, that may be a optical phenomenon microcontroller information acquisition tool. The output of the sensing element was connected to 1 of the Arduino Uno’s I/O pins. A serial communication link was established between the Arduino Uno and therefore the digital computer. baud was set to 9600bps. Total of ten,000 traditional information components and fifty,000 faulty information components were obtained at temperature (approximately 24 26°C). Faulty information was generated through simulations. For each thought of drift fault worth, information was generated of one hundred twenty samples, every sample consisting of one hundred information parts, 1st fifty traditional and last fifty faulty information parts, as incontestable in Figure nine. Out of one hundred twenty samples, 1st sixty faulty and last sixty traditional samples were generated. The knowledgebased fault detection technique is adopted, which solely needs historical information for coaching. The received information from Arduino UNO was kept on ESP8266 for additional process and simulation functions. The data were divided into one hundred twenty samples, every sample consisting of one hundred information parts. Then, drift faults were simulated within the obtained information. For every thought of drift, we tend to get one hundred twenty samples. The resultant information set consisted of 5*120*100 data parts for the 5 drift categories. What is more, for feature extraction and to cut back the size, gamma-hydroxybutyrate and mean options were extracted from the conventional and faulty signal information and so pooled along to come up with coaching information? The mean and most worth is taken into account sensible to be calculated once the defect affects the mean and gammahydroxybutyrate of the signal amplitude. B. Training and Testing Classifiers were trained on esp8266 exploitation machine learning library scikit-learn for the Python programing language. For coaching SVM, inbuilt perform SVC supported the one-versus-rest manner with linear kernel perform was used. testing, Arduino microcontroller was code to arbitrarily generate binary range x. The temperature output, Vout wherever the fault was injected in traditional temperature T.For each thought of drift fault price, pickle files were generated and used any on for testing the performances. IV. CONCLUSION In this paper, the authors establish drift fault in detector fault detection downside. Low procedure facility (ESP8266) was projected, which may effectively be employed in sensible systems for showing intelligence fault detection in a period of time exploitation AI techniques. Many machine learning classification algorithms were accustomed classify knowledge as traditional and faulty. Experimental results show that SVM and ANN performed hugely well, even with the smallest amount options and while not requiring an outsized amount of knowledge.
- 4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 09 Issue: 07 | July 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 2579 V. FUTURE SCOPE For future work, an additional capable single-board pc is often used rather than an esp8266, which might handle additional complicated operations, and numerous sensors, like measuring an instrument or a pressure device are often used rather than a temperature device for various sorts of alternative device faults. Also, a fault diagnosis and prognosis are typically done following the data-driven approach. REFERENCES [1] Z. Gao, C. Cecati, and S. X. Ding, “A survey of fault diagnosis and fault- tolerant techniques-Part I: Fault diagnosis with model-based and signal- based approaches,” IEEE Trans. Ind. Electron., vol. 62, no. 6, pp. 3757- 3767, Jun. 2015. [2] D. Park, S. Kim, Y. An and J. Jung, “LiReD: A Light- Weight Real-Time Fault Detection System for Edge Computing Using LSTM Recurrent Neural Networks,” MDPI Sensors., 18,2110; DOI: 10.3390/s18072110, Jun. 2018. [3] J. Tian, C. Morillo, M. H. Azarian, and M. Pecht, “Motor bearing fault detection using spectral kurtosis-based feature extraction coupled with K-nearest neighbor distance analysis,” IEEE Trans. Ind. Electron., vol. 63, no. 3, pp. 1793-1803, Apr. 2016. [4] T. W. Rauber, F. De A. Boldt, and F. M. Varejao, “Heterogeneous feature´ models and feature selection applied to bearing fault diagnosis,” IEEE Trans. Ind. Electron., vol. 62, no. 1, pp. 637-646, Sep. 2015. [5] O. Castro, C. Sisamon, and J. Prada, “Bearing fault diagnosis based´ on neural network classification and wavelet transform,” in Proc. 6th WSEAS Int. Conf. Wavelet Anal. Multirate Syst., Bucharest, Romania, pp. 22-29, Oct. 2006. [6] B. Samanta, “Gear fault detection using artificial neural networks and support vector machines with genetic algorithms,” Mech. Syst. Signal Process., vol. 18, no. 3, pp. 625-644, 2004. [7] B. Sreejith, A. K. Verma, and A. Srividya, “Fault diagnosis of rolling element bearing using time- domain features and neural networks,” in Proc. IEEE Region 10 3rd Int. Conf. Ind. Inf. Syst., vol. 1, pp. 1-6, Sep. 2008. [8] Y. Wang, J. Xiang, R. Markert, and M. Liang, “Spectral kurtosis for fault detection, diagnosis and prognostics of rotating machines: A review with applications,” Mech. Syst. Signal Process., vols. 66-67, pp. 679-698, Apr. 2016. [9] Q. Xiao, Z. Luo, and J. Wu, “Fault detection and diagnosis of bearing based on local wave time- frequency feature analysis,” in Proc. 11th Int. Conf. Natural Comput. (ICNC), pp. 808-812, 2015. [10] J. L. Yang, Y. S. Chen, L. L. Zhang, and Z. Sun, “Fault detection, isolation, and diagnosis of self-validating multifunctional sensors,” Rev. Sci. Instrum., vol. 87, no. 6, p. 065004, 2016. [11] R. Dunia, S. J. Qin, T. F. Edgar, and T. J. Mcavoy, “Identification of faulty sensors using principal component analysis,” Process Syst. Eng., vol. 42, no. 10, pp. 2797-2812, 1996. [12] J. Kullaa, “Detection, identification, and quantification of sensor fault in a sensor network,” Mech. Syst. Signal Process., vol. 40, no. 1, pp. 208221, Sep. 2013. [13] Y. Yu, W. Li, D. Sheng, and J. Chen, “A novel sensor fault diagnosis method based on modified ensemble empirical mode decomposition and probabilistic neural network,” Measurement, vol. 68, pp. 328-336, May 2015. [14] S. U. Jan, Y.-D. Lee, J. Shin, and I. Koo, “Sensor Fault Classification Based on Support Vector Machine and Statistical Time-Domain Features,” IEEE Access, vol. 5, no. 1, pp. 8682–8690, 2017. [15] S. Haykin, Neural Networks: A Comprehensive Foundation, 2nd ed. Upper Saddle River, NJ, USA: Prentice-Hall, 1994. [16] A. Geron, Hands-On Machine Learning with Scikit- Learn & TensorFlow. O’Reilly, 2017. [17] Scikit-learn library. Available online scikit-learn org (accessed on 5 Sep 2019).