Content uploaded by Zhenyou Zhang
Author content
All content in this area was uploaded by Zhenyou Zhang on Nov 13, 2014
Content may be subject to copyright.
A preview of the PDF is not available
Intelligent fault diagnosis and prognosis approach for rotating machinery integrating wavelet transform, principal component analysis, and artificial neural networks
Abstract and Figures
This paper proposes a new approach for rotating machinery which integrates wavelet transform (WT), principal component analysis (PCA), and artificial neural networks (ANN) to classify the fault and predict the conditions of components, equipment, and machines. The standard deviation of wavelet coefficients are extracted from processed historical signals of manufacturing equipment as features. Then, the features are analyzed by PCA and several new principal features obtained from original features can be used as inputs to train ANN. After training, the conditions and degradations of components and machines can be predicted, and the fault of them can be classified if it exists, by the trained ANN using the same kinds of principal features extracted from real time signals. A case study is used to evaluate the proposed method and the result indicates its higher efficiency and effectiveness comparing to traditional methods.
Figures - uploaded by Zhenyou Zhang
Author content
All figure content in this area was uploaded by Zhenyou Zhang
Content may be subject to copyright.
... Mateus et al. [45] presented in his article predictive models using an LSTM network to predict future equipment status based on data from an industrial paper press. Zhang et al. [46] proposed an approach to perform a prognosis of rotating equipment's health using wavelet transform (WT), a principal component analysis (PCA) and artificial neural networks (ANN) to classify the failure and predict the condition of components, equipment and machines. Martins et al. [47] showed how it was possible to classify the health condition of equipment via an HMM with multivariate analysis. ...
... It analyses a data table in which observations are described by several intercorrelated quantitative dependent variables and is widely used due to its ability to extract interpretable information by efficiently removing redundancies [54,55]. It is typically used to perform the dimensional reduction of large sets of time series observations [56], moving from representing possibly correlated variables to a new set of orthogonal, uncorrelated variables and preserving the highest percentage of information [40,46,55]. In this way, it allows a rapid assessment of any relationships between variables [54]. ...
... In other words, it is a method of projecting large dimensional measurements towards a minimum dimensional space and preserving maximum variance [57] by compressing sensory data according to their spatial and temporal correlations [58]. Then, the PCA produces linear combinations of the original variables to generate new axes, known as principal components (PCs), with the first PC having as high a variance as possible, possessing the greatest variability in the data, and each subsequent component in turn having as high a variance as possible under the constraint that it is not correlated with the previous components [46,54]. In other words, the PCA is a linear transformation that rearranges the data into a new coordinate system, in which the first PC is defined as the coordinate that shows the greatest variation in the data when projected in that direction; the second PC is the coordinate that presents the second largest variation, and so on for the other components. ...
The maintenance paradigm has evolved over the last few years and companies that want to remain competitive in the market need to provide condition-based maintenance (CBM). The diagnosis and prognosis of the health status of equipment, predictive maintenance (PdM), are fundamental strategies to perform informed maintenance, increasing the company’s profit. This article aims to present a diagnosis and prognosis methodology using a hidden Markov model (HMM) classifier to recognise the equipment status in real time and a deep neural network (DNN), specifically a gated recurrent unit (GRU), to determine this same status in a future of one week. The data collected by the sensors go through several phases, starting by cleaning them. After that, temporal windows are created in order to generate statistical features of the time domain to better understand the equipment’s behaviour. These features go through a normalisation to produce inputs for a feature extraction process, via a principal component analysis (PCA). After the dimensional reduction and obtaining new features with more information, a clustering is performed by the K-means algorithm, in order to group similar data. These clusters enter the HMM classifier as observable states. After training using the Baum–Welch algorithm, the Viterbi algorithm is used to find the best path of hidden states that represent the diagnosis of the equipment, containing three states: state 1—“State of Good Operation”; state 2—“Warning State”; state 3—“Failure State”. Once the equipment diagnosis is complete, the GRU model is used to predict the future, both of the observable states as well as the hidden states coming out from the HMM. Thus, through this network, it is possible to directly obtain the health states 7 days ahead, without the necessity to run the whole methodology from scratch.
... ANN models emulate a biological neural network, i.e., the central nervous systems of animals, particularly the brain [13,14]. These models can deal with complex problems without sophisticated and specialized knowledge, provide an effective classification technique, and deal with nonlinear systems and low operational response time after the learning phase [15]. ANN models have been applied to a wide range of fields [12][13][14][15][16][17], but it is still of interest to explore ANN models into PdM and especially with sensor data as the main input. ...
... These models can deal with complex problems without sophisticated and specialized knowledge, provide an effective classification technique, and deal with nonlinear systems and low operational response time after the learning phase [15]. ANN models have been applied to a wide range of fields [12][13][14][15][16][17], but it is still of interest to explore ANN models into PdM and especially with sensor data as the main input. ...
Possessing an efficient production line relies heavily on the availability of the production equipment. Thus, to ensure that the required function for critical equipment is in compliance, and unplanned downtime is minimized, succeeding with the field of maintenance is essential for industrialists. With the emergence of advanced manufacturing processes, incorporating predictive maintenance capabilities is seen as a necessity. Another field of interest is how modern value chains can support the maintenance function in a company. Accessibility to data from processes, equipment and products have increased significantly with the introduction of sensors and Industry 4.0 technologies. However, how to gather and utilize these data for enabling improved decision making within maintenance and value chain is still a challenge. Thus, the aim of this paper is to investigate on how maintenance and value chain data can collectively be used to improve value chain performance through prediction. The research approach includes both theoretical testing and industrial testing. The paper presents a novel concept for a predictive maintenance platform, and an artificial neural network (ANN) model with sensor data input. Further, a case of a company that has chosen to apply the platform, with the implications and determinants of this decision, is also provided. Results show that the platform can be used as an entry-level solution to enable Industry 4.0 and sensor data based predictive maintenance.
... Step 1: Daubechies-4 (dB-4) wavelet was employed to decompose the three-level raw signals. Daubechies wavelets have gained significant popularity in the field of rotating machine fault diagnosis, specifically db4 [30]. ...
In large-scale manufacturing, ensuring the efficient operation of rotating machines is crucial to avoid breakdowns and failures during production. This article introduces a method for detecting gearbox faults by analyzing vibration signals and employing artificial intelligence techniques, with a particular emphasis on comparing these methods. The diagnostic process consists of three stages: extracting features using Wavelet Packet Transform (WPT) and statistical analysis, selecting optimal properties through the gain ratio method, and using Support Vector Machine (SVM) and Artificial Neural Network (ANN) models to distinguish between faults and assess their performance. The diagnostic outcomes demonstrate that both SVM and ANN models accurately identify various fault patterns depending on the operating conditions. Remarkably, the study highlights the ANN model's superiority over the SVM model in classifying gearbox faults, indicating its suitability for gearbox fault diagnosis. This research yields valuable insights into machine condition monitoring, showcasing the ANN model as a robust tool for gearbox fault detection. The findings advocate for the implementation of ANN-based approaches in real-world applications to enhance the reliability of fault detection and prevention in rotating machines. Furthermore, future research directions may explore additional enhancements and optimizations for ANN models, leading to more advanced machine health monitoring systems in the manufacturing industry.
... Studying their impact on vibrations to examine the impacts of several defects on the vibrations of the rotor-bearing system. The fault of rotary machinery elements was studied by Zhang et al. [10] using hybrid wavelet and integration of main component analysis and ANN to predict the situation of systems. RSM was suggested by Goyal et al. [11] for identifying the appropriate method for positioning non-contact sensors in order to successfully monitor the condition of rotating machine components. ...
A gearbox is vital for transmitting motion and power in rotating machinery systems, and the failure in a gearbox due to faults in gears will degrade the performance, even leading to significant losses. Therefore, condition monitoring of a gearbox is vital for maintaining system performance. In the present work, condition monitoring of a bevel gearbox using response surface methodology is presented. An experimental test rig is used for experimentation with varying speed and loading conditions to obtain the vibration signals under normal and simulated defective conditions. Three input parameters, such as radial load, operating speed, and pitting defect, are considered to study the parametric effect. Time-domain Kurtosis is used as the response parameter to monitor the effects on vibrations of the gearbox due to load, operating speed, and pitting defect size. Response surface methodology (RSM), a combination of statistical and mathematical techniques, is employed to describe the relationship between the inputs and outputs of the system. The analysis of the results demonstrates the significance of the size of the pitting defect, the interaction effect of the size and load of the fault, and the interaction effect of the size and speed of the defect on system performance. It further establishes that RSM with analysis of variance (ANOVA) is a reliable method for evaluating the vital factors associated with the vibration of the gearbox.
The current research aims to analyze the impact of multiagent technology (MAT) on the performance of manufacturing organizations by using a hybrid approach of analytical hierarchy process (AHP), decision making trial evaluation laboratory (DEMATEL), and technique for order preference by similarity to ideal solution (TOPSIS) to enhance the process quality while reducing maintenance costs. Extant literature helps to identify five factors and twenty-one variables of MAT, which are used to construct the conceptual domain of MAT. The article uses data collected from different manufacturing sectors. Then, integration of AHP, DEMATEL, and TOPSIS is applied to the collected data in which AHP is used to construct the hierarchical structure to quantify the relative relevance and ranking of variables of MAT (local and global rank); DEMATEL is used to detect direct and indirect interaction between these variables and determine the interrelationship among them; and TOPSIS is used for comparison of the global ranking of variables of MAT evaluated through AHP. The AHP approach assigns the most significant weight to “integration of manufacturing systems with maintenance,” followed by fault detection, manufacturing process, organizational productivity, and organizational control. DEMATEL, on the other hand, considers “integration of manufacturing system with maintenance” to be the most influential factor in its group, and causal and effect behavior demonstrates that organizational productivity and organizational control are influenced by others. Finally, the cross-examination of TOPSIS concludes that most variables obtained the same ranking as computed through AHP. The implication scope of the article indicates that the performance of manufacturing organizations can be enhanced by utilizing MAT and its dimensions in a customized manner.
As the concept of Industry 4.0 is introduced, artificial intelligence-based fault analysis is attracted the corresponding community to develop effective intelligent fault diagnosis and prognosis (IFDP) models for rotating machinery. Hence, various challenges arise regarding model assessment, suitability for real-world applications, fault-specific model development, compound fault existence, domain adaptability, data source, data acquisition, data fusion, algorithm selection, and optimization. It is essential to resolve those challenges for each component of the rotating machinery since each issue of each part has a unique impact on the vital indicators of a machine. Based on these major obstacles, this study proposes a comprehensive review regarding IFDP procedures of rotating machinery by minding all the challenges given above for the first time. In this study, the developed IFDP approaches are reviewed regarding the pursued fault analysis strategies, considered data sources, data types, data fusion techniques, machine learning techniques within the frame of the fault type, and compound faults that occurred in components such as bearings, gear, rotor, stator, shaft, and other parts. The challenges and future directions are presented from the perspective of recent literature and the necessities concerning the IFDP of rotating machinery.
The work aims to devise a three strata multivariate statistical process monitoring strategy for an integrated steel plant. A stratified process representation has been devised, and a new fault diagnostic statistic has been proposed. The three strata process representation took into consideration the stages, sub-stages and the process characteristics (PCs) within the sub-stages where each of these entities formed a strata or level. The process representation or the nominal model for the process monitoring strategy is based on Multi-Block Principal Component Analysis. Besides, the contribution of primary (PCs), the proposed diagnostic statistic also took into consideration the effect of associate PCs. PCs from the preceding sub-stages having significant correlation with the primary PCs are termed in the article as associate characteristics (ACs). The strategy thus devised, and the diagnostic statistic developed herein has been corroborated using an example of an Integrated Steel Plant (ISP). The proposed monitoring strategy when applied to the ISP related case study was proficient enough to detect the faults occurring in the process heats. The results obtained such as the model fit values (0.572); the faults detected, and the diagnosis of the fault was found to be in good synchronization with real findings pertaining to facility.
The current research aims to analyse the critical factors of Multi-Agent Technology (MAT) which influence manufacturing organizations to enhance quality and reduce maintenance costs by implementing a combined approach of Analytical Hierarchical Process (AHP) and Decision-Making Trial Evaluation Laboratory (DEMATEL). A framework has been developed in which AHP is used to quantify the relative importance of each MAT variable and DEMATEL is used to capture the MAT variable interaction. The hybrid approach of AHP and DEMATEL concludes that ‘integration of manufacturing systems with maintenance’ is the highest weight factor, followed by ‘manufacturing processes’, ‘fault detection’, ‘organizational productivity’, and ‘organizational control’. This suggests that efficient MAT deployment in industrial firms could improve product quality by decreasing maintenance efforts. This study provides a tested and validated framework to motivate the decision-makers, professionals, and researchers of a variety of manufacturing organizations to implement this framework in a customised way for improving quality and cutting maintenance costs.
A traditional difficulty in vehicle road testing is the nonstationary nature of the data. Frequency Response Functions (FRF) and other classical input/output relationships are difficult to obtain accurately due to the uncontrolled time and energy characteristics of the road input to the vehicle. Time-variant methods can be useful in this analysis. This paper presents a new FRF approach based on the wavelet transform. The procedure employs the continuos Grossman-Morlet wavelets calculated in the frequency domain. The method is used to analyze acceleration transmissibilities across an automobile seat in the vertical direction. The study involves transient road data from a speedbump and nonstationary road data from a pave' surface. The method provides new insights with respect to classical time-invariant approaches and would thus seem promising for a wide range of vehicle testing applications. Several observations are made regarding the dynamics of the person/seat system.
Modern sensors working on new principles and/or using new materials and technologies are more precise, faster, smaller, use less power and are cheaper. Given these advantages, it is vitally important for system developers, system integrators and decision makers to be familiar with the principles and properties of the new sensor types in order to make a qualified decision about which sensor type to use in which system and what behavior may be expected. This type of information is very difficult to acquire from existing sources, a situation this book aims to address by providing detailed coverage on this topic. In keeping with its practical theme, the discussion concentrates on sensor types used or having potential to be used in industrial applications.
Good progress has been made in the development of diagnostic systems for machines. The computer has been used effectively to shorten the time and reduce the skills required by maintenance personnel for solving problems. Particular progress has been made in the areas of the relatively infrequent but complex CNC problems and the more frequent but less complex machine cycle hangups. There is strong research interest in the use of vibration signatures to anticipate the failure of mechanical elements such as slideway and spindle bearings.
Good progress has been made in the development of diagnostic systems for machines. The computer has been used effectively to shorten the time and reduce the skills required by maintenance personnel for solving problems. Particular progress has been made in the areas of the relatively infrequent but complex CNC problems and the more frequent but less complex machine cycle hangups. There is strong research interest in the use of vibration signatures to anticipate the failure of mechanical elements such as slideway and spindle bearings.
In order to choose the rotating machinery fault diagnosis characteristics accurately, in this paper, a kind of choice method of fault diagnosis characteristics was put forward based on the time domain statistical analysis. Through analysis the probability distribution of the time domain dimensionless characteristic parameters, choose the parameters witch is obviously different from others and used as rotating machinery's fault diagnosis characteristic parameters. Through MATLAB simulation, we can draw a conclusion from the results that this method can improve the efficiency and accuracy of fault diagnosis effectively.
Models of transformer fault diagnosis were developed by using on-line data to improve the conventional testing method and physical law methods. The operation data of 7 variables that affect transformer fault had been studied by using principal component analysis method, 5 principal components had been obtained and the contributions of the principal components had been computed. Based on the factors, a three-layer RBF neural network is designed. It is proved by MATLAB experiment that RBF neural network is a strong classifier which can be used to diagnose transformer fault effectively.