Search

This paper presents a multivariate linear contribution analysis in the context of fault detection, isolation and diagnosis. The usually univariate contribution analysis in fault isolation is improved by the use of feature selection. The fault index and the individual contributions of the variables are calculated by Probabilistic Principal Component Analysis. A new and more efficient method is proposed to select the most decisive variables that contribute to the fault. Experiments are conducted with illustrative synthetic benchmarks and the Tennessee Eastman chemical plant simulator. Among the multivariate selection searches, the Sequential Backward and Forward search shows an optimized equilibrium between the quality of the selected set of contributing variables and the computational burden, compared to an exhaustive and Branch & Bound search.

In multi-label learning a single object can be associated with multiple labels simultaneously. In a context where labels follow a random distribution, every labelling has a probability of occurrence. Thus, any prediction is associated with an expected error measured by a predefined loss function. From an exponential number of possible labellings, an algorithm should choose the prediction that minimizes the expected error. This is known as loss minimization. This work shows a proof of the NP-completeness , with respect to the number of labels, of a specific case of the loss minimization of the Coverage loss function, which allows to conclude that the general case is NP-hard .

This work proposes a classification model for fault diagnosis. In a first stage, an unsupervised clustering algorithm discovers groups of potential fault classes. Subsequently, a specialized classifier is trained for each of the clusters, thus diminishing complexity and augmenting performance. The quality of the diagnosis is further improved by combining the classifiers in an ensemble. As a benchmark, data provided by the Tennessee Eastman chemical plant simulator was used, with promising results.

In the Oil and Gas industry, estimating a subsurface velocity field is an essential step in seismic processing, reservoir characterization, and hydrocarbon volume calculation. Full-waveform inversion (FWI) velocity modeling is an iterative advanced technique that provides an accurate and detailed velocity field model, although at a very high computational cost due to the physics-based numerical simulations required at each FWI iteration. In this study, we propose a method of generating velocity field models, as detailed as those obtained through FWI, using a conditional generative adversarial network (cGAN) with multiple inputs. The primary motivation of this approach is to circumvent the extremely high cost of full-waveform inversion velocity modeling. Real-world data were used to train and test the proposed network architecture, and three evaluation metrics (percent error, structural similarity index measure, and visual analysis) were adopted as quality criteria. Based on these metrics, the results evaluated upon the test set suggest that the GAN was able to accurately match real FWI generated outputs, enabling it to extract from input data the main geological structures and lateral velocity variations. Experimental results indicate that the proposed method, when deployed, has the potential to increase the speed of geophysical reservoir characterization processes, saving on time and computational resources.

This paper presents a systematic procedure to fairly compare experimental performance scores for machine learning methods for fault diagnosis based on vibration signals. In the vast majority of related scientific publications, the estimated accuracy and similar performance criteria are the sole quality parameter presented. However, the experimental design giving rise to these results is mostly biased, based on unacceptably simple validation methods and on recycling identical patterns in test data sets, previously used for training. Moreover, the methods in general overfit their hyperparameters, introducing additional overoptimistic results. In order to remedy this defect, we critically analyse the usual training-validation-test division and propose an algorithmic guideline in the form of a validation framework. This allows a well defined comparison of experimental results. In order to illustrate the ideas of the paper, the Case Western Reserve University Bearing Data benchmark is used as a case study. Four distinct classifiers are experimentally compared, under gradually more difficult generalization tasks using the proposed evaluation framework: K-Nearest-Neighbor, Support Vector Machine, Random Forest and One-Dimensional Convolutional Neural Network. An extensive literature review suggests that most vibration based research papers, particularly for the Case Western Reserve University Bearing Data, use similar patterns for training and testing, making their classification an easy task.

This work presents the concept of Cascade Feature Selection to combine feature selection methods. Fast and weak methods, like ranking, are placed on the top of the cascade to reduce the dimensionality of the initial feature set. Thus, strong and computationally demanding methods, placed on the bottom of the cascade, have to deal with less features. Three cascade combinations are tested with the Extreme Learning Machine as the underlying classification architecture. The Tennessee Eastman chemical process simulation software and one high-dimensional data set are used as sources of the benchmark data. Experimental results suggest that the cascade arrangement can produce smaller final feature subsets, expending less time, with higher classification performances than a feature selection based on a Genetic Algorithm. Many works in the literature have proposed mixed methods with specific combination strategies. The main contribution of this work is a concept able to combine any existent method using a single strategy. Provided that the Cascade Feature Selection requirements are fulfilled, the combinations might reduce the time to select features or increase the classification performance of the classifiers trained with the selected features.

Distinct feature extraction methods are simultaneously used to describe bearing faults. This approach produces a large number of heterogeneous features that augment discriminative information but, at the same time, create irrelevant and redundant information. A subsequent feature selection phase filters out the most discriminative features. The feature models are based on the complex envelope spectrum, statistical time- and frequency-domain parameters, and wavelet packet analysis. Feature selection is achieved by conventional search of the feature space by greedy methods. For the final fault diagnosis, the k-nearest neighbor classifier, feedforward net, and support vector machine are used. Performance criteria are the estimated error rate and the area under the receiver operating characteristic curve (AUC-ROC). Experimental results are shown for the Case Western Reserve University Bearing Data. The main contribution of this paper is the strategy to use several different feature models in a single pool, together with feature selection to optimize the fault diagnosis system. Moreover, robust performance estimation techniques usually not encountered in the context of engineering are employed.

The main motivation to develop this work is to create a diagnosis system able to facilitate the work of human experts responsible for detecting faults before acquisition of submersible petroleum motor pump systems. A new approach for multiclass learning by reduction to multiple, binary classifiers, in a one-versus-one scheme, is presented as an alternative artificial intelligence solution to diagnose faults. Such an idea is based on the hypothesis that each pair of process conditions has different optimal feature sets to improve the classification performance. Thus, features are selected from datasets containing only two classes. Then, classifiers are trained with the selected features. The combination uses the average confidence of each classifier pair prediction to calculate the ensemble answer. Experimental results show that the proposed approach improves classification performance in a statistically significant way, when compared with correlated work. A secondary contribution is the analysis of the most difficult fault to be identified, namely rubbing.

This paper presents an extension of a comparative study of classifier architectures for automatic fault diagnosis, with a special emphasis on the Extreme Learning Machine (ELM), with and without kernel mapping. Besides the explanation of the ELM model, an attempt is made to find theoretical hints of the excellent generalization capabilities of this model, based on the findings of Cover about dichotomies and the equivalence of Mean Squared Error minimization in the high-dimensional feature spaces induced by kernels, and spaces defined by a finite sample set. The field of application is a practical problem in the context of offshore petroleum exploration where sophisticated submersible motor pumps are extensively tested before being deployed. The work juxtaposes the performance of ELM to an existing statistically sound comparison of state of the art classifier methods for a hand-crafted feature model tailored specially to the spectra of the vibrational signals of the pump. The results suggest the remarkably good generalization capability of ELM, exhibiting the highest scores for the chosen F-measure performance criterion.

In this paper, an artificial intelligence solution to diagnose faults before acquisition of submersible petroleum motor pump systems is presented. Proper fault identification is time consuming and demands highly trained human experts. The diagnosis system is intended to facilitate the work of the human component of this important process by replicating the decision of highly trained experts through a classifier. To perform the automatic diagnosis, firstly intermediate features are extracted as the vibration spectra. Subsequently, high level features are extracted and fed into a classifier that outputs the final diagnose. To validate our proposal and to select the best classifier (among K-Nearest-Neighbour, Random Forest, Support Vector Machine and Decision Tree) for this problem, we performed a comparative study using real data acquired in tests accomplished before acquisition of submersible motor pumps. Our dataset comprises thousands of entries of accelerometer sensors (vertically distributed along the particular system components) data labelled by an human expert to one of the considered scenarios (normal pump, faulty sensor, faulty pump with rubbing, misalignment or unbalance). Results have showed that the evaluated classifiers have equivalent performance for the given problem, and that the standardization procedure can improve the performance of some classifiers. The performance of the classifiers is sufficient to facilitate the work performed by humans and consequently reduce the time spent in the pump fault diagnosis process.

This work presents a framework for fault detection and diagnosis of multimode industrial processes. It is shown that the proposed method to generate features for the faults provides a generalization that can allow the fault diagnosis in modes for which no training was performed. The proposed method is applied to a sophisticated simulator of a continuous stirred tank reactor (CSTR). Experimental results suggest the usefulness of this approach for the detection of faults under operating conditions not considered during training phase.

In pattern recognition systems, ensemble techniques claim a potential performance improvement compared to single classifier approaches. Decision templates (DT) were proposed as a simple and effective method for combining continuous valued outputs of an ensemble of classifiers. In this paper, the concept of decision template single-label multi-class classifier combination is extended to the multi-label case. The different classifiers needed for a combination are obtained from the continuous re-estimation used in the Recursive Dependent Binary Relevance multi-label classifier. Each base classifier used in this work, delivers besides the class label, a continuous output for the class that can be used to assemble the DTs.

We present a generic procedure for diagnosing faults using features extracted from noninvasive machine signals, based on supervised learning techniques to build the fault classifiers. An important novelty of our research is the use of 2000 examples of vibration signals obtained from operating faulty motor pumps, acquired from 25 oil platforms off the Brazilian coast during five years. Several faults can simultaneously occur in a motor pump. Each fault is individually detected in an input pattern by using a distinct ensemble of support vector machine (SVM) classifiers. We propose a novel method for building a SVM ensemble, based on using hill-climbing feature selection to create a set of accurate, diverse feature subsets, and further using a grid-search parameter tuning technique to vary the parameters of SVMs aiming to increase their individual accuracy. Thus our ensemble composing method is based on the hybridization of two distinct, simple techniques originally designed for producing accurate single SVMs. The experiments show that this proposed method achieved a higher estimated prediction accuracy in comparison to using a single SVM classifier or using the well-established genetic ensemble feature selection (GEFS) method for building SVM ensembles.

There are many applications of plate identification in pattern recognition and machine vision. These applications range from complex security systems to commons areas and from parking admission to urban traffic control. Car license plate recognition (CLPR) has complex characteristics due to diverse effects as fog, rain, shadows, irregular illumination conditions, partial occlusion, variable distances, cars' velocity, scene's angle on frame, plate rotation and conservation, number of vehicles in the scene and other. These factors make plate recognition much more complex and difficult than the traditional pattern or optical character recognition (OCR) systems. The main objective of this work is to show a system that solves the practical problem of car identification for real scenes. All steps of the process, from image scene acquisition to optical character recognition are considered to achieve an automatic identification of plates. It can be used with all type of country rules or plates design and adapted to each situation. The system is computationally very efficient and it is suitable for others related image recognition applications.

We give the analytical definitions of the Chernoff, Bhattacharyya and Jeffreys-Matusita probabilistic distances between two Dirichlet distributions and two Beta distributions as its special case. For all other known probabilistic distances we show their inappropriateness in the analytical case. We discuss the parameter learning of the Dirichlet distribution from a finite sample set and present an application for split-and-merge image segmentation.