Fault location measurement of sensor nodes based on fuzzy control algorithm

The sensor nodes are devices with extra capabilities of sensing, communicating and processing of the recorded data. The sensor-based wireless and IoT-enabled networks can contribute a precarious part of army command, defense control, defense communications, computerized based intelligence and surveillance, and above all the targeting systems. Examples of defense applications include monitoring of own army forces and enemy forces; monitoring of ammunition, monitoring of equipment, targeting at certain regions; and nuclear as well as biotechnology-based attack detection. With the advent of AI and latest technologies, a technical paradigm shift can be seen in monitoring the faults in sensor nodes that are capable to collect the data and entire decisions are inferred on the basis of this data. By deploying sensors in critical areas, all the movements can be followed in detail. Hence, instant fault location measurement of sensor nodes with the least complex approaches is the need of the hour. Therefore, a novel soft computing-based approach is proposed in this paper for measuring the fault in locations of sensor-based networks using fuzzy logic and neural networks with high accuracy. The proposed technique allows lower consumption of energy while locating the faults. The results demonstrate the superior performance of the proposed fault location scheme with respect to the network node loss, measurement error, and measurement time taken for detecting faults in locations of sensor nodes. The proposed fuzzy and neural network-based method outperforms the existing benchmarked techniques considered for comparative study.


Introduction
With the advent of the latest technologies, emergence of 6G, and evolution of AI (artificial intelligence), and IoT (Internet of Things), the research in sensors and allied areas is growing rapidly as the sensor nodes are used to capture the real-time data, processing of the data and then providing useful information from the inferences of the gathered data. Especially in military services, sensor nodes are automated with specified programs and deployed to send the sensitive information to the headquarters if any movement in sensitive region is noticed. Sensor nodes are deployed in peaceful regions can also be used as alert systems to provide early warning if any suspicious movement or any toxic substance is identified.
At the initial stage of deployment, wireless sensor network nodes are in a dormitory state where no monitoring and inspection take place. The execution of running state of sensor network nodes is unknown most of the times, and it is not feasible to carry out real-time monitoring or frequent inspection of these nodes. Due to emergence of IoT, the data are captured by diverse devices and sensor nodes to generate the useful information; generation of relevant information is only feasible if the processing of data, mining of data, and analysis of data are completed in an appropriate manner. In case the sensor node fails, it may affect the processing of data, analysis of data and also affects the monitoring system. There are soft and hard realtime requirements for processing the data and if the data belong to hard real-time application then failure in sensor nodes could be very expensive affair. It may cause huge damage to the end users and intermediator users such as scientists, industrialists and academicians. Therefore, Communicated by Suresh Chandra Satapathy.
& Nuo Yu yunuopaper@163.com accurate and timely diagnosis of faulty sensor nodes is inevitable and troubleshooting mechanism should be devised at the earliest to provide seamless services to the users in wireless sensor networks, IoT-and cloud-based environment (Kaur and Kadam 2019). Timely detection and correction of faulty sensor nodes can certainly improve the operation reliability of sensor-based network (wireless sensor networks, IoT environment and heterogeneous networks) to ensure smooth functioning of these networks (Jia et al. 2018). The wireless networks which carry IoT-and cloudbased data are composed of a large number of sensor nodes which are deployed in diverse geographical areas. There is hardly any central node for controlling the other sensor nodes in the network. The transmission of data takes place by the intermediate nodes in a multi-hop forwarding fashion (Guoyan and Qian 2019). The sensor nodes are often attacked by various intruders, and it is very difficult to segregate the faulty sensor nodes from the normal sensor nodes. In addition to it, sensor nodes have quality issues with respect to their manufacturing and problems with protocol designs (Kaur and Kadam 2017). All these factors cause failure to the sensor nodes and thus make the wireless networks vulnerable to the intruders. Therefore, the timely detection of faults in sensor nodes is very important to ensure seamless connectivity and smooth functioning of wireless networks. The fault detection techniques for sensor nodes in IoT environment and wireless networks can be categorized into centralized or client-server fault detection methods and distributed or P2P fault detection methods (Yang et al. 2021).
The client-server fault detection techniques usually require the data being processed by a specific sensor node and then regulate the status of the other nodes in the same network. These techniques have certain limitations, such as failure in single sensing node can halt the entire network if no replacement is found for central sensing node, loss of information, and consumption of more energy (Han et al. 2019). The distributed or P2P fault detection techniques are more flexible, and these techniques enable each node to find faults in itself and in the nodes located nearby since these nodes have adequate information about the surrounding nodes (Zhang 2020). Irrespective of the distributed or centralized fault detection methods, the present techniques have limitations and new mechanisms can be devised to overcome these limitations. There are many state-of-the-art techniques proposed by researchers for fault location measurement of sensor nodes but still there is a need to explore better techniques for finding faults in sensor nodes for seamless connectivity on 6G-enabled wireless networks.
There are certain research gaps in the existing techniques which are mentioned as follows: • The existing techniques consume high energy for detection of faults with greater accuracy. The sensor node has to communicate with the neighboring nodes that cause higher consumption of energy. Nowadays energy saving is an important concern to be considered before devising any mechanism. • The detection techniques either focus on client-server or distributed layout for detection of faulty nodes. Therefore, there is a need for detection mechanism that can work precisely irrespective of the underlying layout of sensor nodes. • The sensor nodes can be judged on the basis of data collected in the past, and the judgment of network health is not the only factor to be considered during designing the fault detection mechanism.
In order to overcome the problems of the existing literature, a novel fuzzy and neural network-based model is proposed for fault location measurement of sensor nodes. The major contributions of the paper are discussed below: • A novel fault location method based on fuzzy login in a combination with RBF neural network for sensor nodes is designed. • The parameters such as residual energy, node centrality, node degree, energy threshold, data collection before failure are considered for decision making on sensor node failure. • Node fault information is captured through smart devices, and preprocessing of fault information is performed. Then, the fault diagnosis information reduction algorithm of rough set is used to reduce the fault symptom set, to remove the redundant and unimportant symptoms. Finally, the reduced symptom set is used as the input of neural network to measure the fault location of sensor nodes. • Then in the fault diagnosis of wireless sensor network (WSN), a fault location method based on fuzzy control algorithm has been proposed, which uses radial basis function (RBF) neural network. This manuscript is structured into five sections. The initial section provides background details in the area of fault finding in sensor nodes, also explains the requirement of further research in this area and states the major contributions of the paper. The next part of this paper provides detailed study on existing fault detection schemes. Third section provides detailed study on the proposed fuzzy method-based approach for measuring the faults in sensor nodes. The next section explains experimental results and analysis. Finally, the last section concludes the research work.

Related work
In this section, an attempt is made to provide insights into the research contributions made by others in our field of study.
WSNs can face multiple failures such as software and hardware failures, and connectivity failures (Zidi et al. 2017). SVM is used for identifying faults in sensor nodes placed in WSNs (Zidi et al. 2017). The decision function is based on the statistical learning theory and is deployed at cluster heads to determine defective sensors. An experimental comparative study with the current methodologies demonstrates the efficiency of SVM for fault detection in WSNs. The state-of-the-art fault detection approaches in WSN are evaluated in the study presented in Muhammed and Shaikh (2017), and taxonomy of the approaches has been offered to categorize the existing studies. The authors have offered a qualitative and quantitative assessment of various strategies based on their proposed taxonomy. Median absolute deviation (MADN) and trust matrix are found to be effective approaches for reducing the error probability. The research work develops a distributed online one class-support vector machine (OCSVM) for network anomaly detection (Miao et al. 2018) and derives a decentralized cost function. The kernel function is substituted with a random approximation function to obtain the distributed deployment without transferring the original data. Additionally, a sparse constraint is applied to the decentralized cost function to obtain an additional approximation dimension. Then, using stochastic gradient descent, these two cost functions are minimized, yielding two distributed methods. To demonstrate the usefulness of the proposed algorithms, several theoretical analyses and tests are carried out. The proposed distributed algorithms not only outperform the existing anomaly detection methods in terms of quick operating time and minimal consumption of CPU memory.
In (Gharghan et al. 2016), the goal of the research study is to measure the distance between a movable sensor node (such as a bicycle) and an anchor node (such as a coach) in both outdoor and interior settings. To calculate such a distance, two methods are examined. The first approach is based on the hybrid approach which makes use of PSO and ANN algorithm to optimize the mobile node's distance estimation accuracy and the second approach is based on the log-normal shadowing model (LNSM). The distance estimate accuracy of the hybrid PSO-ANN algorithm is found to be higher than the classic LNSM technique. The work presented in Swain et al. 2018a investigates link failures in WSNs from a theoretical and analytical standpoint. In testbed trials, the suggested neural network-based model has been thoroughly assessed. The simulation-based experiment is carried out to demonstrate the viability of the model for dense networks with up to 1000 interconnections. The results of the simulated environment, as well as the testbed testing in both indoor and outdoor situations, indicate that the method is proficient for detecting link failures with a greater detection rate in consistent manner.
Due to the limited precision and high complexity of existing algorithms for diagnosing sensor networks, paper (Cheng et al. 2018) proposes a new defect detection technique. A failure prediction model is formed using SVR approach. The node state is determined through mutual testing among trustworthy neighbors. Simulations reveal that the proposed SVR approach has achieved a detection accuracy of 87%. In (Yuan et al. 2018), machine learning (ML) techniques are compared and analyzed for anomaly detection such as SVM, Naive Bayes, and Gradient Lifting. The simulated results reveal that the Gradient Lifting has a higher accuracy in comparison with other ML methods for fault identification. In (Abdullah et al. 2018), the intrusion detection mechanism for WSNs is presented. The random forest (RF) algorithm is used to detect various DDOS attacks in sensor-based networks such as flooding, blackhole, grayhole and scheduling attacks. The accuracy is measured using F1 scores.
The data-driven fault detection model is proposed using an effective ML algorithm namely RF in combination with XGBoost algorithm . RF is exploited to prioritize the features, which are either direct sensor signals or indirect sensor signals. XGBoost then trains the classifier for each precise defect based on the high-ranking features. Furthermore, while dealing with multi-dimensional data, the proposed ensemble classifier protects against overfitting and produces superior fault detection results. The research work presented in Gu et al. (2018) provides an enhanced sensor fault diagnosis methodology that covers fault detection as well as fault identification. To detect distinct sensor failure modes, an improved errorcorrecting output coding (ECOC) SVM-based fault identification system is suggested. The experimental results show that the suggested defect diagnosis methodology works on the real-time data and provides better the accuracy for identification of faults. The defect detection problem of nonlinear stochastic systems with WSNs is addressed in the research work presented in Gao et al. (2018) using a distributed filtering approach. Fuzzy models are used to depict nonlinear stochastic systems with discrete temporal form. Simulation results satisfactorily confirm the effectiveness and applicability of the distributed fault detection technique.
A heterogeneous defect diagnostic protocol for WSNs is presented (Swain et al. 2018b). To diagnose the heterogeneous problematic nodes in WSN, the protocol with three stages such as clustering, location of faults, and categorization of faults is proposed. For identification of the different types of defective nodes, the feed-forward neural network is used. The testbed experiment is also performed to evaluate the suggested model in simulated and real-time environment. The study in Jan et al. (2017) considers diverse problems when diagnosing defects in sensor-based networks. In some situations, the fault incidence event in fault samples is picked at random basis to mimic a realworld scenario. To solve this problem of overfitting, k-fold cross-validation is used. The efficiency of SVM and the neural network-based classifier is demonstrated by the ROC curve and F1-score. The goal of the research presented in Kullaa (2013) is to use structural response data captured by sensor nodes to identify the sensor defects. The approach can detect a sensor fault, quantify the type of fault, and rectify the faulty sensor. The research work introduces a unique hybrid methodology for failure identification using a RF classifier (Wang et al. 2017). Several fault classes are considered as the methodology is tested on simulated environment and real-world data. The RF-based classifier has achieved 88.23% accuracy according to the result findings.
With the expansion in the size of WSN, the problematic nodes and the traffic overhead also grow. To address this problem, a cluster-oriented fault tolerance methodology is designed and implemented (Rajeswari and Neduncheliyan 2017). The network is clustered in this instance using a low-energy distance-based clustering technique. A group of backup nodes is chosen for each cluster head using a residual energy metrics. This aids in detection of defects in cluster members and cluster head. The proposed methodology reduces energy loss and overhead, according to simulation results. A multi-class classification in WSNs is proposed to address the decentralized defect detection problems (Zhang and Zhang 2015). Trial results show that the proposed fusion rule outperforms other existing works in terms of fusion accuracy. The new distributed GT technique is presented which predicts the set of spread faulty sensors from a limited number of linearly independent binary messages sent by the sensors using a lowcomplexity distance decoder (Tošić et al. 2013).
An extensive study has been carried out for fault location assessment of sensor nodes. It is found that WSNs and IoT-enabled networks need more dynamic methods to detect faults in sensor nodes. For sensor nodes, the traditional fault localization measuring approach required more energy and a lengthier measurement time. As a result, a novel fault location measurement approach based on fuzzy control algorithms for sensor nodes has been proposed in this paper.

Proposed method
This section explains the proposed fuzzy-based approach for measuring faults in sensor nodes.

Fuzzy control algorithm
In order to apply the algorithm, a few assumptions are made as described as follows: (1) The network is a square network, each sensing node s in the network has a unique id, assuming that there are S sensor nodes in the entire network; the node set in the network is expressed in Eq. 1.
(2) The nodes have been deployed in a random manner but the location of the nodes is fixed. (3) Each node has the same initial energy, memory, energy consumption for processing information, and the same ability to send and receive data. (4) The distance can be ascertained by using the received signal strength indication between the sensor nodes. (5) After the nodes are deployed, the distance among the sensor nodes and the sink node and the distance between the sensor nodes and the neighbor nodes are analyzed, and the number of neighbor nodes is determined.
The distance between the nodes is expressed as d, the energy consumption of the sensor node transmitting l-bit data is expressed in Eq. 2.
Here, E elec is the energy consumed by the transmitting and receiving circuits of the node; e fs and e mp are the energy consumption of the circuit amplifier in the short-distance and long-distance data transmission, respectively. When the broadcast distance d\d 0 , it is defined as the shortdistance; when the communication distance d ! d 0 , it is defined as the long-distance. Then, d 0 is defined as shown in Eq. 3.
The consumption of energy by the sensor node receiving l-bit data is mentioned in Eq. 4.
The consumption of energy by l-bit data fusion is mentioned below in Eq. 5.
Here, E pDh is the energy consumed by l-bit data fusion. Firstly, the input and output variables of the fuzzy controller are finalized. The input variables of the controller are the residual energy of the sensor node and the node centrality, the node degree is given as e, the distance from the converging node and the output variables are defined as the probability of the selection of node as the cluster head and the competition radius.
(1) Residual energy: The amount of residual energy directly determines the data collection, data fusion and data transmission capacity of the sensor node. Sufficient residual energy certifies that the operation of the node is normal.
(2) Node centrality: The node centrality indicates the importance of the node in the network. The smaller the average shortest distance between the node and the neighbor node is, the higher the node centrality value is, indicating that the node is more important in the given region. The value of node centrality can be calculated by Eq. 6.
where N i j j represents the no. of neighboring nodes of node i and S are represents the area of the sensor.
(3) Node degree: Node degree indicates the number of neighboring nodes in the range of node communication from the sensor node. In the process of clustering, the higher the node degree of cluster head is, the lower the transmission cost is. (4) Distance from sink node: The closer position of the cluster head to sink node accelerates the chances for the network ''hot spot'' problem. Therefore, reduction in the cluster size closer to the sink node can effectively reduce the impact of ''hot spots'' on the whole network life. (5) Probability of Cluster Head node selection: The probability P of the selection of a node as the cluster head is the first output variable of the fuzzy controller, which is determined by three inputs: the residual energy, the node centrality and the distance from the convergence node. This value indicates the possibility of node becoming cluster head. After comparison of probability values, the node with the largest value is selected as cluster head. (6) Competition radius: Competition radius is the second output variable of the fuzzy controller; it is determined by three inputs: residual energy, degree of node and distance from the convergence node. After the first output variable determines the cluster head, the value of the output variable represents the competitive radius of the cluster head that helps in determining the size of the cluster.
In sensor networks, in order to extend the life of nodes with low residual energy, it is often expected to reduce their node degree and to reduce their communication workload. On the contrary, nodes with higher residual energy are often expected to increase their node degree and improve communication efficiency. The node degree can be adjusted by controlling the transmission power. The operation process of the whole algorithm is explained in Table 1, for determination of sensor node energy.

Fault location measurement method
According to the characteristics of large-scale WSNs, the fault location measurement method based on clustering mechanism is proposed, and clustering mechanism is considered to be a very effective mechanism to increase the accuracy of fault location measurement in sensor nodes. However, these clustering mechanisms do not consider the diagnosis of cluster heads most of the time. Therefore, a fuzzy control approach is proposed for fault location measurement of the sensor nodes.
Assuming that sensor nodes are randomly positioned in a rectangular area, all nodes maintain the same wireless transmission distance d, and the area can be completely covered by the transmission range of sensor network. The assumption also considers that a large number of nodes are deployed in this area, so there are at least three neighbor nodes in each node's hop range, which can be achieved by properly setting the transmission distance d of the nodes. Each node can communicate with its neighbors in the transmission range and locate its neighbors through broadcast confirmation protocol.
The model of fault location measurement is presented in Fig. 1 by using the method of fuzzy control algorithm and neural network. First of all, wireless sensor network collects fault information of sensor nodes and preprocesses the collected fault information by using relevant signal processing methods, so that the original fault symptom set can be obtained. Then, the fault diagnosis information reduction algorithm of rough set is used to reduce the fault symptom set and remove the redundant and unimportant symptoms. Finally, the reduced feature set is used as the input to the neural network to measure the fault location of sensor nodes.
In the fault diagnosis of WSNs, a fault location measurement technique based on fuzzy approach is devised, which uses radial basis function (RBF) neural network. RBF network comprises three-layered architecture where the input layer is comprised of signal source nodes; the second layer is constituted by the hidden layer, the number of neurons in the hidden layer is determined by the needs of the described problem, and the transfer function of neurons in the hidden layer is radial basis function; the last layer exhibits as the output layer. The communication from input layer space to hidden layer space is nonlinear, and from hidden layer space to output layer space is linear. The input layer is used to transmit signals, and the hidden layer is used to adjust the parameters of the transfer function. The output layer adjusts the linear weights to provide output in a relatively fast mode then the hidden layer.
A radial basis cell model with m inputs is set. The transfer function of RBF takes the distance dist k k between the input vector p and the weight vector w as the independent variable. The transfer function of the radial basis function neuron has various forms, but the most commonly used form is Gaussian function. Therefore, the transfer function of the RBF neural network can be expressed as shown in Eq. 7.
The basic neuron model has m inputs, each of which is connected to the next layer by an appropriate weight w. The network output can be expressed as shown in Eq. 8.
where f is the transfer function representing the input/ output relationship, a is the output vector, p is the input vector, w is the weight, and b is the threshold. To limit the network output (such as between 0 and 1), the S-type transfer function can be used. RBF neural network needs to consider three parameters: the center of the basis function, variance and the weight from the hidden layer to the output layer. The activation function of RBF neural network is expressed as shown in Eq. 9.
where x p 2 c 1 is the Euclidean distance, c is the center of the Gaussian function, and r is the variance of the Gaussian function. The output of the network is expressed in Eq. 10. where represents the pth input sample, p represents the samples, c 1 represents the center of the hidden layer, w ij represents the connection weight from the hidden layer to the output layer, h represents the number of nodes in the hidden layer, and y j represents the actual output of the j th output node of the network. Table 1 Algorithm for determination of sensor node energy Algorithm-1 Determination of sensor node energy Step 1-First of all, determine the actual energy of the node Step 2-Determine the actual node degree of the node Step 3-Then find out the energy threshold Step 4-State the expected degree of nodes in the network; Step 5-Determine the Initial transmitting power of the node; Step 6-Node's energy difference is calculated as = node's actual energy-energy threshold; Step 7-Deviation of the expected node degree is determined; Step 8-Target node degree is determined; Step 9-Adjustment of node degree is done as per problem statement, Step 10-The next hop node determined by the transmission power p is added to the node set of the topology sub network; Step 11-The links of all nodes in the connection node set are added to the link set Step 12-Again calculate the energy of the sensor node Let d be the expected output value of the sample, then the variance of the base function is expressed in Eq. 11.
The specific operation steps of the learning algorithm are as follows: (1) By using K-means clustering, the center c of basis function is obtained.
À Network initialization h Training samples are selected randomly as cluster center c i ; The input training sample set is grouped according to the nearest neighbor rule: According to the Euclidean distance between x p and center c i , x p is allocated to the average value of training samples in each cluster set # p of input samples; if the new cluster center does not change any more, then c i is obtained, which is the final basis function center of RBF neural network, otherwise`, and the next round of center solution is carried out as shown in Eq. 12 where variance calculation is performed.
(2) Solving the variance r i as expressed in Eq. 12.
where c max is the maximum distance between the selected centers.
(3) The weight between the hidden layer and the output layer is calculated as shown in Eq. 13.
The basic idea of BP (back propagation) network learning rule is as follows: the correction of network weight and threshold value should follow the direction of the fastest decline of performance function, negative gradient direction as shown in Eq. 14.
where x k shows the current weight and closed value matrix, g k denotes the gradient of the current representation function, and a k is the learning rate.
In the process of measuring the possible states of each node, the probability of the normal node being wrongly measured as expressed in Eq. 15.
On the basis of the above analysis, combined with the fuzzy control algorithm, in each cluster, the cluster head node is used to directly measure the probability of the fault location of its member nodes as shown in Eq. 16.
4 Experimental results and discussions In order to check the viability of the proposed fault location method for sensor node the simulation-based experiments are performed. The experimental setting is made as follows: Intel (R) Pentium (R) 4, CPU2.30 GHz, 2 GB memory, Microsoft Windows 10platform (64 bit). The parameters considered for performance evaluation are network node loss, measurement error in percentage, and measurement time for fault detection.

Network node loss/(mJ)
In the following experiments, two state-of-the-art measurement methods are selected for comparative study on simulation experiments. The following experiments mainly compare the network node loss of different methods, and the results are revealed in Fig. 2. From Fig. 2, the inference can be drawn that the proposed method performs well while we compare the proposed method with the latest state-of-the-art techniques with respect to network node loss. The network node loss is maximum in case of SVR-based technique (Gharghan et al. 2016), moderate in case of fuzzy-based stochastic approach ) and minimum in case of our novel fuzzyand RBF-based neural network approach. Table 2 depicts that the proposed fuzzy-based approach of fault detection produces results with more accuracy and reduces errors in measurement as more iterations are performed. The performance improves with more number of iterations. Analysis of the above experimental data from Table 1 shows that the measurement errors of the existing techniques are more and among the three methods considered for the experimental study, the proposed method has high measurement accuracy.

Measurement time/(seconds)
The measurement time for fault detection of the three methods is compared in Fig. 3. According to results shown in Fig. 3, the measurement time of the proposed method is significantly lower than that of the other two techniques considered for experimental study, which shows that the performance of the proposed method is improved to a certain extent compared to the benchmarked state-of-theart methods. The SVR-based technique takes more time as number of iterations increase, the fuzzy-based stochastic method takes moderate time and the proposed method shows considerable improvement in the time taken for detecting faults in sensor-based networks. Figure 4 shows the energy consumed by various approaches while detecting the faults in the sensor network. It can be observed as the payload bits increase, the energy consumption is reducing and showing inversely proportion mechanism. The payload represents the data to be transmitted in any kind of encapsulated frame formats as defined by the network protocols.
In defense oriented services, sensor nodes are automated with explicitly programming to send the sensitive information. The faults in the sensors should be identified at the earliest to keep the pulse of the military networks healthy. Failure in sensor nodes and delay in sending sensitive information are not affordable affairs in the defense oriented networks. Though there exists many fault detection mechanisms, they have certain limitations. The energy consumption is imperative parameter for selection fault detection techniques, and most of the existing techniques consume high energy while detecting the faults with high accuracy. Secondly, the detection techniques are dependent upon the underlying layout of the nodes while our method is independent of the underlying architecture of the nodes. Another factors such as consumption of time for detecting faults, error in fault measurement and node loss are also important parameters to be considered while designing and deploying any fault detection mechanism and it can be observed from this section that the proposed method is very efficient with respect to the node loss, measurement error (%), time taken for detection of faults and consumption of energy for locating faults.

Conclusion
Sensor nodes are the important component of modern era whether it is the field of mobile computing, cloud networks, IoT-or defense-based networks. However, the latest software defined networks are performing well but the data are collected by sensing nodes and locating faults in sensor nodes is inevitable for the better performances of all types of networks. Hence, the research on fault detection novel approach is presented in this paper which uses fuzzy logic in a combination of neural networks to measure the faults  in sensor nodes. This paper focuses on the research of fault location measurement of sensor nodes. The results of the proposed technique have been measured with respect to network node loss, measurement error, energy consumption and measurement time taken by the fault detection methods with the two latest state-of-the-art techniques. The results are satisfactory as the proposed fuzzy-based fault detection technique shows minimal measurement error, minimal energy consumption and minimum time for detecting faults. The proposed fuzzy-based fault detection method outperforms the existing techniques against all the performance evaluation parameters. The proposed method is tested for distributed environment and is flexible enough to integrate in client-server architecture as well. In future, more parameters can be considered for the research study and the proposed mechanism can be tested on different networking environment to prove the viability of the proposed method.