Intelligent diagnosis method for a centrifugal pump using features of vibration signals

In the field of machinery diagnosis, the utilization of vibration signals is effective in the detection of fault, because the signals carry dynamic information about the machine state. However, knowledge of a distinguishing fault is ambiguous because definite relationships between symptoms and fault types cannot be easily identified. This paper presents an intelligent diagnosis method for a centrifugal pump system using features of vibration signals at an early stage. The diagnosis algorithm is derived using wavelet transform, rough sets and a partially linearized neural network (PNN). ReverseBior wavelet function is used to extract fault features from measured vibration signals and to capture hidden fault information across optimum frequency regions. As the input parameters for the neural network, the non-dimensional symptom parameters that can reflect the characteristics of a signal are defined in the amplitude domain. The diagnosis knowledge for the training of the PNN can be acquired by using the rough sets. We also propose a diagnosis method based on the PNN, one which can deal with the ambiguity problem of condition diagnosis, and distinguish fault types on the basis of the possibility distributions of symptom parameters automatically. The decision method of optimum frequency region for extracting feature signals is also discussed using real plant data. Practical examples of diagnosis for a centrifugal pump system are shown in order to verify the efficiency of the method.     

基于邻域粗糙集的航空发电机健康诊断方法

提出一种基于邻域粗糙集和支持向量机相结合的航空发电机智能健康诊断方法.采用专业健康试验平台对某型战斗机的真实航空发电机进行试验,得到转速、负载、油压等大量表征发电机健康状态的监测数据.引入数据挖掘思想,采用邻域粗糙集理论对监测数据进行属性约简,将约简后的属性集输入给所设计的支持向量机健康诊断器,对航空发电机的健康状态进行了诊断研究.研究表明,该方法能够很好实现对某真实航空发电机的健康诊断,具有较高的推广应用价值.     

A comparative study of Naïve Bayes classifier and Bayes net classifier for fault diagnosis of monoblock centrifugal pump using wavelet analysis

In most of the industries related to mechanical engineering, the usage of pumps is high. Hence, the system which takes care of the continuous running of the pump becomes essential. In this paper, a vibration based condition monitoring system is presented for monoblock centrifugal pumps as it plays relatively critical role in most of the industries. This approach has mainly three steps namely feature extraction, classification and comparison of classification. In spite of availability of different efficient algorithms for fault detection, the wavelet analysis for feature extraction and Naïve Bayes algorithm and Bayes net algorithm for classification is taken and compared. This paper presents the use of Naïve Bayes algorithm and Bayes net algorithm for fault diagnosis through discrete wavelet features extracted from vibration signals of good and faulty conditions of the components of centrifugal pump. The classification accuracies of different discrete wavelet families were calculated and compared to find the best wavelet for the fault diagnosis of the centrifugal pump.     

Designing fuzzy inference systems from data: Aninterpretability-oriented review

Fuzzy inference systems (FIS) are widely used for process simulation or control. They can be designed either from expert knowledge or from data. For complex systems, FIS based on expert knowledge only may suffer from a loss of accuracy. This is the main incentive for using fuzzy rules inferred from data. Designing a FIS from data can be decomposed into two main phases: automatic rule generation and system optimization. Rule generation leads to a basic system with a given space partitioning and the corresponding set of rules. System optimization can be done at various levels. Variable selection can be an overall selection or it can be managed rule by rule. Rule base optimization aims to select the most useful rules and to optimize rule conclusions. Space partitioning can be improved by adding or removing fuzzy sets and by tuning membership function parameters. Structure optimization is of a major importance: selecting variables, reducing the rule base and optimizing the number of fuzzy sets. Over the years, many methods have become available for designing FIS from data. Their efficiency is usually characterized by a numerical performance index. However, for human-computer cooperation another criterion is needed: the rule interpretability. An implicit assumption states that fuzzy rules are by nature easy to be interpreted. This could be wrong when dealing with complex multivariable systems or when the generated partitioning is meaningless for experts. The paper analyzes the main methods for automatic rule generation and structure optimization. They are grouped into several families and compared according to the rule interpretability criterion. For this purpose, three conditions for a set of rules to be interpretable are defined     

Leak detection of pipeline: An integrated approach of rough set theory and artificial bee colony trained SVM

The generation of leak along the pipeline carrying crude oils and liquid fuels results enormous financial loss to the industry and also affects the public health. Hence, the leak detection and localization problem has always been a major concern for the companies. In spite of the various techniques developed, accuracy and time involved in the prediction is still a matter of concern. In this paper, a novel leak detection scheme based on rough set theory and support vector machine (SVM) is proposed to overcome the problem of false leak detection. In this approach, ‘rough set theory’ is explored to reduce the length of experimental data as well as generate rules. It is embedded to enhance the decision making process. Further, SVM classifier is employed to inspect the cases that could not be detected by applied rules. For the computational training of SVM, this paper uses swarm intelligence technique: artificial bee colony (ABC) algorithm, which imitates intelligent food searching behavior of honey bees. The results of proposed leak detection scheme with ABC are compared with those obtained by using particle swarm optimization (PSO) and one of its variants, so-called enhanced particle swarm optimization (EPSO). The experimental results advocate the use of propounded method for detecting leaks with maximum accuracy.     

Data mining for exploring hidden patterns between KM and its performance

A large volume of works have addressed the importance of Knowledge management (KM). However, there are increasingly numerous concerns about whether the KM efforts can be fairly reflected and transformed into the business performance. Even though the KM contribution is qualitative and hard to measure, some works using statistical methods declare that a specific KM style may produce a better corporate performance. Statistical methods attempt to summarize yesterday’s success rules, while data mining techniques aim to explore tomorrow’s success clues. This study challenges the issue of what the hidden patterns between KM and its performance are, and whereby identifies the reality of whether a better performance is resulted from a special KM style. The analysis results using Bayesian network classifier and rough set theory show that it is not easy to support that a special KM style would produce a similar performance.     

Change Detection Using Adaptive Fuzzy Neural Networks: Environmental Damage Assessment after the Gulf War

This article introduces an adaptive fuzzy neural network classifier for environmental change detection and classification applied to monitor landcover changes resulting from the Gulf War. In this study, landcover change is treated as a qualitative shift between landcover categories. The Change Detection Adaptive Fuzzy (CDAF) network learns fuzzy membership functions for each landcover class present at the first image date based on a sample of the image data. An image from a later date is then classified using this network to recognize change among familiar classes as well as change to unfamiliar landcover classes. The CDAF network predicts landcover change with 86% accuracy representing an improvement over both a standard multidate K-means technique which performed at 70% accuracy and a hybrid approach using a maximum likelihood classifier (MLC)/K-means which achieved 65% accuracy. In this study, we developed a hybrid classifier based on conventional statistical methods (MLC/K-means classifier) for comparison purposes to help evaluate the performance of the CDAF network. The CDAF compared with existing change detection methodology has two features that lead to significant performance improvements: 1) new landcover types created by a change event automatically lead to the establishment of new landcover categories through an unsupervised learning strategy, and 2) for each pixel the distribution of fuzzy membership values across possible categories are compared to determine whether a significant change has occurred.     

Fuzzy classifier design using genetic algorithms

A new method for design of a fuzzy-rule-based classifier using genetic algorithms (GAs) is discussed. The optimal parameters of the fuzzy classifier including fuzzy membership functions and the size and structure of fuzzy rules are extracted from the training data using GAs. This is done by introducing new representation schemes for fuzzy membership functions and fuzzy rules. An effectiveness measure for fuzzy rules is developed that allows for systematic addition or deletion of rules during the GA optimization process. A clustering method is utilized for generating new rules to be added when additions are required. The performance of the classifier is tested on two real-world databases (Iris and Wine) and a simulated Gaussian database. The results indicate that highly accurate classifiers could be designed with relatively few fuzzy rules. The performance is also compared to other fuzzy classifiers tested on the same databases.     

Investigation of an efficient shape optimization procedure for centrifugal pump impeller using eagle strategy algorithm and ANN (case study: slurry flow)

Investigation an efficient shape optimization method for centrifugal pump and other turbo-machine is significant to reduce time consumption of process and increase accuracy and modification. For analysis an efficient shape optimization procedure, slurry flow in centrifugal pump is investigated. Since a centrifugal water pump has been not designed to carry out slurry flows, its performance decreases and energy consumption of this kind of pump increases. Therefore, improvement of performance and reduction of energy consumed for these pumps are the major issues. Since the performance of a centrifugal pump strictly depends on its impeller shape, in this study, the shape of impeller was optimized in order to achieve a higher efficiency for slurry flow. To optimize the impeller geometry and to improve the performance of Berkeh 32–160 pump as for the case study, Artificial Neural Networks (ANN) and Eagle Strategy (ES) algorithms have been coupled with a validated 3D Navier–Stokes equations for two phase flow based on Eulerian-Eulerian model. In the next step, the pump experimentally tested in an established slurry flow test rig in laboratory. Measured data were used to verify the numerical results of initial pump with slurry flow. Finally, the complete numerical characteristic curves of the pump with the optimized impeller were compared to the validated numerical characteristic curves of that with the initial impeller to verify optimization. An efficiency improvement of 3.33% at only 9.9% increasing of head has been obtained for optimized geometry. The results indicated a reasonable improvement in the optimal design of pump impeller and a higher performance using the ES algorithm. Furthermore the ES and PSO algorithm was compared and results shows that ES is efficient than PSO algorithm in this application and this methodology is more efficient than other surrogate methods.     

A genetic algorithm-based rule extraction system

Individual classifiers predict unknown objects. Although, these are usually domain specific, and lack the property of scaling up prediction while handling data sets with huge size and high-dimensionality or imbalance class distribution. This article introduces an accuracy-based learning system called DTGA (decision tree and genetic algorithm) that aims to improve prediction accuracy over any classification problem irrespective to domain, size, dimensionality and class distribution. More specifically, the proposed system consists of two rule inducing phases. In the first phase, a base classifier, C4.5 (a decision tree based rule inducer) is used to produce rules from training data set, whereas GA (genetic algorithm) in the next phase refines them with the aim to provide more accurate and high-performance rules for prediction. The system has been compared with competent non-GA based systems: neural network, Naïve Bayes, rule-based classifier using rough set theory and C4.5 (i.e., the base classifier of DTGA), on a number of benchmark datasets collected from UCI (University of California at Irvine) machine learning repository. Empirical results demonstrate that the proposed hybrid approach provides marked improvement in a number of cases.     

The dominance-based rough set approach to cylindrical plunge grinding process diagnosis

The dominance-based rough set approach is proposed as a methodology for plunge grinding process diagnosis. The process is analyzed and next its diagnosis is considered as a multi-criteria decision making problem based on the modelling of relationships between different process states and their symptoms using a set of rules induced from measured process data. The development of the diagnostic system is characterized by three phases. Firstly, the process experimental data is prepared in the form of a decision table. Using selected methods of signal processing, each process running is described by 17 process state features (condition attributes) and 5 criteria evaluating process state and results (decision attributes). The semantic correlation between all the attributes is modelled. Next, the phase of condition attributes selection and knowledge extraction are strictly integrated with the phase of the model evaluation using an iterative approach. After each loop of the iterative feature selection procedure the induction of rules is conducted using the VC-DomLEM algorithm. The classification capability of the induced rules is carried out using the leave-one-out method and a set of measures. The classification accuracy of individual models is in the range of 80.77–98.72 %. The induced set of rules constitutes a classifier for an assessment of new process run cases.     

一种改善的基于支持向量机的图像边缘检测算法

本文给出一种基于支持向量机方法的边缘检测算法,用以改善传统边缘检测方法中存在的比如粗糙边缘、不准确边缘等缺点。支持向量机是建立在统计学理论基础上的一种新的机器学习方法。首先提出了边缘检测算法流程,然后使用支持向量机分类方法对图像进行边缘检测。用所得到的边缘检测算法与Prewitt算法的性能进行了比较。仿真结果表明本文给出的算法与Prewitt算法相比,边缘检测性能得到提高。     

Hypoglycaemia detection using fuzzy inference system with multi-objective double wavelet mutation Differential Evolution

In this paper, a fuzzy inference system (FIS) is developed to recognize hypoglycaemic episodes. Hypoglycaemia (low blood glucose level) is a common and serious side effect of insulin therapy for patients with diabetes. We measure some physiological parameters continuously to provide hypoglycaemia detection for Type 1 diabetes mellitus (TIDM) patients. The FIS captures the relationship between the inputs of heart rate (HR), corrected QT interval of the electrocardiogram (ECG) signal (QT_c), change of HR, change of QT_c and the output of hypoglycaemic episodes to perform the classification. An algorithm called Differential Evolution with Double Wavelet Mutation (DWM-DE) is introduced to optimize the FIS parameters that govern the membership functions and fuzzy rules. DWM-DE is an improved Differential Evolution algorithm that incorporates two wavelet-based operations to enhance the optimization performance. To prevent the phenomenon of overtraining (over-fitting), a validation approach is proposed. Moreover, in this problem, two targets of sensitivity and specificity should be met in order to achieve good performance. As a result, a multi-objective optimization using DWM-DE is introduced to perform the training of the FIS. Experiments using the data of 15 children with TIDM (569 data points) are studied. The data are randomly organized into a training set with 5 patients (l99 data points), a validation set with 5 patients (177 data points) and a testing set with 5 patients (193 data points). The result shows that the proposed FIS tuned by the multi-objective DWM-DE can offer good performance of doing classification.     

The effects of combining classifiers with the same training statistics using Bayesian decision rules

Multiple classifier systems (MCS) are attracting increasing interest in the field of pattern recognition and machine learning. Recently, MCS are also being introduced in the remote sensing field where the importance of classifier diversity for image classification problems has not been examined. In this article, Satellite Pour l'Observation de la Terre (SPOT) IV panchromatic and multispectral satellite images are classified into six land cover classes using five base classifiers: contextual classifier, k-nearest neighbour classifier, Mahalanobis classifier, maximum likelihood classifier and minimum distance classifier. The five base classifiers are trained with the same feature sets throughout the experiments and a posteriori probability, derived from the confusion matrix of these base classifiers, is applied to five Bayesian decision rules (product rule, sum rule, maximum rule, minimum rule and median rule) for constructing different combinations of classifier ensembles. The performance of these classifier ensembles is evaluated for overall accuracy and kappa statistics. Three statistical tests, the McNemar's test, the Cochran's Q test and the Looney's F-test, are used to examine the diversity of the classification results of the base classifiers compared to the results of the classifier ensembles. The experimental comparison reveals that (a) significant diversity amongst the base classifiers cannot enhance the performance of classifier ensembles; (b) accuracy improvement of classifier ensembles can only be found by using base classifiers with similar and low accuracy; (c) increasing the number of base classifiers cannot improve the overall accuracy of the MCS and (d) none of the Bayesian decision rules outperforms the others.     

Noninvasive detection of bladder cancer using mid-infrared spectra classification

In this paper, we focus on the detection of Bladder Cancer (BC) via mid infrared spectroscopy. Two main contributions, material and methods, are presented. In terms of material, a new minimally invasive technology, combining fiber evanescent wave spectroscopy and newly patented biosensors, is used for the first time to acquire mid-infrared spectra from voided urine/bladder wash. This new machine promises practicality, cheapness and high-quality of spectrum acquisition. As for classical systems, the data acquired using the new system was highly correlated, resulting in a poor classification performance using classical methods. Therefore, the second contribution consists in developing statistical methods that alleviate the problem. Three new statistical methods based on Partial Least Square Discriminant Analysis algorithm (PLSDA) are proposed. PLSDA is a supervised classifier well-known for its ability to process correlated data. The key point is the choice of the most discriminant latent variables in the training step. In this work, we propose three new decision rules in order to select the most relevant latent variables. These decision rules give rise to three algorithms, namely bayesian, joint and best model PLSDA. A comparative study between the proposed methods and standard ones, namely SVM, K-MEANS and classical PLSDA, confirms clearly the efficiency of the former. The best performance in terms of accuracy is achieved by joint and best model PLSDA (82.35%). Besides, by embedding the proposed statistical methods in the new machine, we are able to provide a new medical device that is very promising in terms of automatic bladder cancer detection.     

Wind turbine condition monitoring based on SCADA data using normal behavior models. Part 1: System description

This paper proposes a system for wind turbine condition monitoring using Adaptive Neuro-Fuzzy Interference Systems (ANFIS). For this purpose: (1) ANFIS normal behavior models for common Supervisory Control And Data Acquisition (SCADA) data are developed in order to detect abnormal behavior of the captured signals and indicate component malfunctions or faults using the prediction error. 33 different standard SCADA signals are used and described, for which 45 normal behavior models are developed. The performance of these models is evaluated in terms of the prediction error standard deviations to show the applicability of ANFIS models for monitoring wind turbine SCADA signals. The computational time needed for model training is compared to Neural Network (NN) models showing the strength of ANFIS in training speed. (2) For automation of fault diagnosis Fuzzy Interference Systems (FIS) are used to analyze the prediction errors for fault patterns. The outputs are both the condition of the component and a possible root cause for the anomaly. The output is generated by the aid of rules that capture the existing expert knowledge linking observed prediction error patterns to specific faults. The work is based on continuously measured wind turbine SCADA data from 18 turbines of the 2 MW class covering a period of 30 months.The system proposed in this paper shows a novelty approach with regard to the usage of ANFIS models in this context and the application of the proposed procedure to a wide range of SCADA signals. The applicability of the set up ANFIS models for anomaly detection is proved by the achieved performance of the models. In combination with the FIS the prediction errors can provide information about the condition of the monitored components.In this paper the condition monitoring system is described. Part two will entirely focus on application examples and further efficiency evaluation of the system.     

Forecasting PGR of the financial industry using a rough sets classifier based on attribute-granularity

In the financial industry, continually changing economic conditions and characteristics involving uncertainty and risk have made financial forecasts even more difficult, increasing the need for more reliable ways to forecast a bank’s operating performance. However, early related studies of performance analysis for using statistical methods usually become more complex when relationships in input/output data are nonlinear. Furthermore, strict data assumptions, such as linearity, normality, and independence, limit real-world applications often. Additionally, a drawback of traditional rough sets is that data must be discretized first for improving classification accuracy. To remedy the existing shortcomings above, the study proposes a hybrid procedure, which mixes professional knowledge, an attribute granularity, and a rough sets classifier, for automatically classifying profit growth rate (PGR) to solve real problems faced by investors. The proposed procedure is illustrated by examining a practical dataset for publicly traded financial holding stocks in Taiwan‘s stock markets. The experimental results reveal that the proposed procedure outperforms listing methods in terms of accuracy, and they provide useful insights in responsiveness to rapidly changing stock market conditions. Importantly, the output created by the rough sets LEM2 (Learning from Examples Module, version 2) algorithm is a set of comprehensible rules applied in a knowledge-based investment system for investors.     

Hybrid genetic algorithm and fuzzy clustering for bankruptcy prediction

In the design of a financial bankruptcy prediction model, financial ratio selection and classifier design play major roles. Methodology based on expert opinion, statistical theory and computational intelligence technique has been widely applied. In this study, a hybrid structure integrating statistical theory and computational intelligence technique was developed using genetic algorithm (GA) with statistical measurements and fuzzy logic based fitness functions for key ratio selection. A fuzzy clustering algorithm was used for the classifier design. In the experiments, two financial ratio sets, one extracted from the suggestions of other studies and the other obtained by using the GA toolbox in the SAS statistical software package, were applied to examine the proposed ratio selection schemes. For classifier design, the developed fuzzy classifier was compared with the well known BPNN classifier frequently used in other studies. Besides, comparison between the developed hybrid structure and other well applied structures was also given. Experimental results based on one to four years of financial data prior to the occurrence of bankruptcy were used to evaluate the performance of the proposed prediction model.     

Comparison of fuzzy inference system (FIS), FIS with artificial neural networks (FIS + ANN) and FIS with adaptive neuro-fuzzy inference system (FIS + ANFIS) for inventory control

Conventional inventory models mostly cope with a known demand and adequate supply, but are not realistic for many industries. In this research, the fuzzy inference system (FIS) model, FIS with artificial neural network (ANN) model and FIS with adaptive neuro-fuzzy inference system (ANFIS) model in which both supply and demand are uncertain were applied for the inventory system. For FIS model, the generated fuzzy rules were applied to draw out the fuzzy order quantity continuously. The order quantity was adjusted according to the FIS model with the evaluation algorithm for the inventory model. The output of FIS model was also used as data for FIS + ANN and FIS + ANFIS models. The FIS + ANFIS model was studied with three membership functions; trapezoidal and triangular (Trap), Gaussian and bell shape. Inventory costs of the proposed models were compared with the stochastic economic order quantity (EOQ) models based on previous data of a case study factory. The results showed that the FIS + ANFIS_Gauss model gave the best performance of total inventory cost saving by more than 75 % compared to stochastic EOQ model.