In recent decades, articial intelligence (AI) has played a significant role in the creation of machines that function as closely as possible to human brains as well as researching in some dynamic topics. Humans solve intuitive tasks easily, but describing that intuitive pro- cess is difficult. Therefore, AIs main applications include cognition and machine learning abilities. The machine is an intelligent computer that collects data from experience, learns complicated concepts and then makes an accurate decision. Deep learning is a subset of machine learning, which itself falls under the category of AI . AI takes input data from the environment, and processes it for the purpose of decision making. The main goal of AI is simulating and understanding of human behavior. AI has a variety of applications, including robotics, natural language recognition, computer games, economics, behaviour recognition, and faultdetection and diagnosis.
Distinguishing faults from disturbances is often controver- sial. Therefore, we define anything that can be rejected by the controller eventually as a disturbance and anything else as a fault. Furthermore, we assume that the plastic film is divided into 10 lanes in this paper (Hur et al., 2008). Fig. 2 depicts the results for all these scenarios with one figure for comparison. The upper and lower plots are the same but have different y-axes. The results show that the residual is sensitive to the fault but insensitive to both disturbances, which is the desired property of a FD system. Although, this FD system is capable of detecting faults successfully, faultdiagnosis has not been addressed. A FDD system conducts faultdiagnosis after faultdetection and is presented in the following section.
The combination of Statistical Process Control (SPC) charts and multivariate analysis approach is used for FaultDetection and Diagnosis (FDD) in the chemical process. Principal Component Analysis (PCA) and Partial Correlation Analysis (PCorrA) are the techniques used in this study. A precut multicomponent distillation column that has been installed with controllers is used as the study unit operation. Improved Statistical Process Control method is implemented to detect and diagnose various kinds of faults, which occur in the process.
The development of model-based faultdiagnosis began in the early 1970s. This method of faultdetection in dynamic systems has been receiving more and more attention over the last two decades. Generally ‘fault’ is to be understood as an unexpected change of the system functionality. It may not, however, represent the failure of physical components. Such malfunctions may occur either in the sensors or actuators, or in the components of the process itself. However, the same difference signal can correspond to model-plant mismatches or noise in real measurements, which are erroneously detected as a fault. The availability of a good model of the monitored system can significantly improve the performance of diagnostic tools, minimizing the probability of false alarms. In all but the most trivial cases the existence of a fault may lead to situations related to safety, health, environmental, financial or legal implications.
The switches that make up a matrix converter can either fail open circuit or short circuit, this work focuses on open circuit switch faults. Common causes of open circuit switch faults include gate drive faults, wire bond lift-off and cracking of solder layers . Research has been carried out on the continued operation of a matrix converters during an open-circuit switch failure . These methods require a faultdetection and diagnosis method that is both fast and reliable. There are several existing methods in the literature for detecting open circuit switch faults in AC-AC converters. One method for open circuit faultdetection in AC-AC converters is the error voltage method, in these methods the node voltages of an inverter  or matrix converter  are compared to a set of reference voltages. The differences between the estimated and actual voltages are then used for faultdetection and diagnosis. The problem with the output voltage methods is that the voltages sensors are not normally required for operation of the converter. These methods add to the cost and reduce the reliability of the converter. Spectral methods have also been applied to matrix converters but these methods cannot diagnose the faulty device in direct matrix converters . Another method uses a low frequency estimate of the output current of the converter and compares this to the actual output current of the AC-AC converter . These low frequency methods use the existing output current sensors and do not alter the cost or reliability of the converter. These methods do not detect the fault when it poses the largest risk to the converter; when the output current is close to the nominal value. This is because at this point the error signal used in the low frequency methods is zero or close to zero. Another drawback of the low frequency methods is that they require a load model, to estimate the output currents. If the load is not well defined then this requirement can decrease the performance of the low frequency methods.
The Induction motor is a three phase AC motor and is the most widely used machine. Its characteristic features are: simple and rugged construction, low cost and minimum maintenance, high reliability and sufficiently high efficiency, it needs no extra starting motor and need not be synchronized. During the operation; however, there are several types of faults frequently happening such as: bearing faults, gear faults, rotor bar eccentricity and stator winding failures and misalignment. The faultdetection and diagnosis methods for induction motors are wide such as current spectrum analysis, vibration analysis and acoustic analysis for different types of motor fault identification  to . Since previous research showed that more than 40 percent of faults in induction motors are related to bearing faults , a number of research works have been done with bearing faultdetection using wavelet technique for vibration data  to . Gear faults are also common in the induction motor. In , the authors used an adaptive method for vibration to detect the gear tooth faults. Neural networks and classification techniques also widely used in faultdiagnosis  to .
This paper has reported on the design of a FDD sys- tem and its application to a first-principles model of a plastic film extrusion process. This model has been used to simulate the plant throughout this paper. The FDD system is based on the parity relations and thus requires solving a multi-objective optimisation problem. A genetic algorithm, which is an evolutionary algorithm, has been utilised for solving this multi-objective optimisation prob- lem. This is a novel approach as an analytical method is usually utilised for solving this multi-objective opti- misation problem. The simulation results in Section 4 demonstrated that the FDD system is sensitive to faults but insensitive to disturbances at the same time, which is the desired faultdetection property. In addition to faultdetection, the simulation results in Section 5 showed that faultdiagnosis could also be conducted successfully. The design can be used for other kinds of processes, such as papermaking and steel-rolling as long as a reference model can be identified in the state space form.
ABSTRACT: An Controller Area Network(CAN) bus is used in sending and receiving messages between devices in automobiles. During the transmission of messages through the nodes, there are possibilities of errors. To detect those errors an algorithm called Adaptive FaultDiagnosis algorithm for Controller Area Network(AFDCAN) is used to detects all faulty nodes on the CAN. The algorithm uses single-channel communication expanding the bus-based classic standard CAN protocol like single path communication. Faults at nodes can arise due to failures in the system, in the memory. Bugged node or faulty node doesn’t send or receive packets to neighbour nodes and send wrong packets to neighbour nodes. CAN centrate uses an active hub to connect the CAN based nodes and prevents the propagation of errors from one port to others. The individual test results are exchanged among these processors and the fault-free processors accurately diagnoses the actuator fault. Test rounds continue until the last node in the system is tested. The last fault-free node sends the second result frame to the earlier fault-free node after all of the test rounds are completed. One node can be tested multiple times by another node, and tests are conducted asynchronously. The number of test rounds required for a hierarchical adaptive distributed system level diagnosis algorithm Hi-ADSD is less than that for adaptive DSD. Parameters are faultdetection time, bus load, no of faults occurred.
Theoretically, any difference between the real measurement and the model prediction can be referred to as a fault. However, it is impossible to develop such an accurate model and model error is inevitable in practice. Besides, the measurement is often corrupted by noise. In order to cover these error and noise, a threshold is usually necessary. The faultdetection and diagnosis are then carried out with reference to given thresholds including warning levels and fault alarm levels. All response signals and residual signals are displayed in the scope. Where and when a fault occurred as well as its severity can be established from the scope as shown latter in conjunction with fault cases.
I N the last three decades process control and automation area had a tremendous improvement due to advances on computational tools. Many of regulatory control actions that were performed by human operators are now performed automatically with aid of computers. Nonetheless, in a pro- cess with hundreds of variables, instruments and actuators it is impossible that a person or a group can manage every and any alarm triggered by an abnormal event. Therefore the FaultDetection and Diagnosis (FDD) field had received extensive attention. According to , the current challenge for control engineers is the automation of Abnormal Event Management (AEM) using intelligent control systems. Inside this field, Instrument FaultDetection and Diagnosis is a potential tool to prevent process performance degradation, false alarms, missing actions, process shutdown and even safety problems. A well-known strategy related to this pro- blem is preventive maintenance. In that, periodical tests and calibration are made in instruments. This is a cumbersome task where instruments are dismantled, cleaned, reassembled and calibrated. Even so, this is not a guarantee that faults will not occur . This paper presents an Interval Type- 2 Fuzzy Logic (IT2FL) classifier to detect and diagnose temperature sensor faults in an alternative compressor, named Sales Gas Compressor (SGC), operating in a Gas Processing Unit (GPU).
energy system photovoltaic (PV) electrical power generation has become the most important and promising research tycoon. In order to increase the efficiency of the PV array used further research are being carried out for the localization and detection of faults. In this paper, we present different methods for faultdetection. Method 1: faultdetection by signal response in PV module strings. Method 2: faultdiagnosis in PV system using discrete wavelet transform (DWT). Method 3: fault location in underground PV system using wavelets and artificial neural network (ANN). Method 4: faultdetection and localization using current voltage sensing framework.
The diagnoses in industrial systems represent an important economic objective in process industrial automation area. To guarantee the safety and the continuity in production exploitation and to record the useful events with the feedback ex- perience for the curative maintenance. We propose in this work to examine and illustrate the application ability of the spectral analysis approach, in the area of faultdetection and isolation industrial systems. In this work, we use a com- bined analysis diagram of time-frequency, in order to make this approach exploitable in the proposed supervision strat- egy with decision making module. The obtained results, show clearly how to guarantee a reliable and sure exploitation in industrial system, thus allowing better performances at the time of its exploitation on the supervision strategy. Keywords: Diagnosis; Spectral Analyzes; Faults Detection and Isolation; Condition Monitoring
Although there is a self test in ABS, this test usually examines the electrical circuit instead of mechanical performance. It will be a disaster if the ABS passes the self-test but failure for mechanical reasons because the ABS is only used in emergency and the driver never realises the failure is existing in this case. In order to provide fault prediction and prognosis, different types of ABS faults are modelled in this report. They are pump efficiency loss, fluid leakage, oil air blister inclusion and brake pad efficiency loss, all of which exclude electrical failure assuming that those failures can be detected in self-test procedure.
B = and B 26 ( = B 62 ) observed to remain comparable with the amplitudes for the healthy condition (Fig. 6-7). However, in case of the faulty rotor case the peak, B 22 , reduced significantly from the healthy condition (nearly 0.10 times) and the faulty stator cases, but other peaks, B 11 , B 12 ( = B 21 ) increased significantly (nearly 8-10 times) compared to the healthy and the stator fault cases. These observations are also summarized in Table I. Hence the based on the observation, it can be concluded that the bispectrum of the phase current signal can identify and distinguish the rotor fault and stator fault of the electric motor. It has also been observed that the amplitude of the peaks can show the severity of the stator and the rotor faults. The tests have also been conducted on different load levels of the motor. From the observation, the bispectrum for the healthy, the stator fault and the rotor fault cases was consistent with the different load conditions.
Two induced fault conditions were investigated; fuel starvation through disconnection of the fuel- feed pipe to an individual cylinder, and reduced injection discharge pressure. Under fuel starvation conditions a simple comparison revealed that the events purported to relate to the combustion process in normal operation were absent. To characterise a progressive reduction in injector discharge pressure the authors used a simple technique whereby the energy in a time-windowed section of the signal was calculated and compared as conditions varied. It was also found practical to use the signal averaged over 10 cycles so that a time-averaged, time-domain signal was obtained, which provided a better overall representation of the AE signature at a particular running condition. The signal energy in a window corresponding to the combustion process was found to increase with a reduction in discharge pressure. It was reasoned that this was due to an increase in combustion harshness as a result of poorer fuel atomisation. Analysis of event timing was also found to be useful for faultdetection. The initial event, postulated to be combustion-related occurred earlier in the cycle with decreasing injection pressure. This was consistent with the premise that the event was injection related since the lower delivery pressures would be achieved by the fuel pump earlier in the cycle.
Abstract:- This paper presents an novel approach for faultdetection and diagnosis (FDD) of sensor as well as process faults for Electro-Hydraulic Actuators (EHA) using a bank of residual generators, each of which employs an Extended Kalman Filter (EKF)-based parameter estimator. In traditional sensor faultdetection schemes, actual sensor measurements are compared with measurements reconstructed using state estimators following an analytical redundancy approach. In contrast, we propose detection of sensor faults by comparing estimated values of plant parameters, which deviate under fault, with their nominal values. Since process faults usually manifest themselves in deviation of process parameters, this leads to a unified approach to faultdetection using parameter estimators. Fault isolation is then achieved by using the set of detection flags, obtained by thresholding each of the residuals, in a so-called diagnosis matrix (D-Matrix). Unlike several earlier works on FDI for electro- hydraulic actuator systems, which do not address sensor faults, the present approach is capable of detection and identification of both sensor and process faults. Numerical simulation results for an EHA of a rocket demonstrate the efficacy of the method.
Because of the increasing demand for the operational safety refer to switched systems, the research about faultdiagnosis of switched systems has been a hot top- ic . Hwang et al.  introduced three types of fault in the field of model-based systems, including actua- tor fault, sensor fault and component fault or loss of effectiveness fault. The actuator fault and sensor fault are also called the additive fault, because they have an additional relationship with the original system state. There are many excellent faultdiagnosis methods for additive faults in switched systems. In , a multiple Lyapunov function approach was employed for faultdetection and diagnosis in switched systems. Then the switched Lyapunov function approach was first applied in stability analysis of switched systems . Wang et al.  designed a fault filter for the actuator fault and sensor fault based on H ∞ performance index.
Abstract. The motor failure of precooled air conditioning unit (PAU) affects the operation of HVAC system directly. Traditional faultdetection methods based on frequency analysis of vibration signals need high sampling frequency. However, in some actual operation and maintenance, the sampling frequency of related data is lower, it is difficult to meet the needs. In this paper, a PAU motor faultdiagnosis model is constructed based on long short-term memory neural network (LSTM) combined with in-depth learning technology. The temperature of motor shell is an important symbol of motor fault. Therefore, effective features are extracted by analyzing the data characteristics. LSTM method is used to predict the motor shell temperature, and the motor faultdetection and diagnosis are carried out according to the predicted residual threshold. According to the computation, the diagnostic accuracy of fault data is 100%, and the false alarm rate of fault-free data is 0.3%. The results show that the model has stronger generalization ability, higher prediction accuracy.
Existing Self-configuring Algorithm- In WSN sensor nodes are in a virtual grid structure in which the network nodes are divided into several cells. One node in each cell is selected as cell manager. Upper level nodes of the grid are cell managers and the remaining nodes will be in lower level grid. A large virtual group can be formed by several virtual cells and these cells can have hundreds to thousands sensor nodes. A group manager is appointed for each virtual group. This group manager is responsible for managing and organizing sensor nodes in its group. Another virtual grid structure is created by the group managers from different groups. This structure is shown in Figure 5. Top level of the management hierarchy is the sink, which is above the group manager. We are referring the algorithm of M. Asim et al  as existing algorithm. This self-configuring algorithm follows cellular approach [31, 32]. In self-detection mechanism, sensor nodes monitor their residual battery energy periodically
It is well known that Induction Motor monitoring has been studied by many researchers and reviewed in a number of works –. Reviews about various stator faults and their causes, and detection techniques, latest trends, and diagnosis methods supported by the artificial intelligence, the microprocessor, the computer, and other techniques in monitoring and protection technologies have been proposed. The major intricacy is the lack of an accurate model that describes a fault in the motor. Moreover, experienced Engineers are often called upon to interpret measured and/or observed data that are frequently inconclusive or unconvincing. A Fuzzy Logic approach help diagnose induction motor faults under such jittery situations. In fact, Fuzzy Logic is reminiscent of human thinking process and natural language enabling decisions to be made based on vague information. Therefore, fuzzy logic technique can adequately be extended to the Induction Motor FaultDetection and Diagnosis. The motor condition is described using linguistic terms like „good‟, „bad‟, „overload‟, „performance declined‟ etc. Thus health interpretation of induction motor turns out to be a Fuzzy Concepts .