Texture features for image classification and retrieval.

Artificial intelligence (AI)-based rotor fault diagnosis (RFD) poses a variety of challenges to the prognostics and health management (PHM) of the Industry 4.0 revolution. Rotor faults have drawn more attention from the AI research community in terms of utilizing fault-specific characteristics in its feature engineering, compared to any other rotating machinery faults. While the rotor faults, specifically structural rotor faults (SRF), have proven to be the root cause of most of the rotating machinery issues, the research in this field largely revolves around bearing and gear faults. Within this scenario, this paper is the first of its kind to attempt to review and define the role of AI in RFD and provides an all-encompassing review of rotor faults for the researchers and academics. In addition, this study is unique in three ways: (i) it emphasizes the use of fault-specific characteristic features with AI, (ii) it is grounded in fault-wise analysis rather than component-wise analysis with appropriate fault categorization, and (iii) it portrays the current research and analysis in accordance with different phases of an AI-based RFD framework. Finally, the section on future research directions is aimed at bridging the gap between a laboratory-based solution and a real-world industrial solution for RFD.

Role of artificial intelligence in rotor fault diagnosis: a comprehensive review

In condition based maintenance, different signal processing techniques are used to sense the faults through the vibration and acoustic emission signals, received from the machinery. These signal processing approaches mostly utilise time, frequency, and time-frequency domain analysis. The features obtained are later integrated with the different machine learning techniques to classify the faults into different categories. In this work, different statistical features of vibration signals in time and frequency domains are studied for the detection and localisation of faults in the roller bearings. These are later classified into healthy, outer race fault, inner race fault, and ball fault classes. The statistical features including skewness, kurtosis, average and root mean square values of time domain vibration signals are considered. These features are extracted from the second derivative of the time domain vibration signals and power spectral density of vibration signals. The vibration signal is also converted to the frequency domain and the same features are extracted. All three feature sets are concatenated, creating the time, frequency and spectral power domain feature vectors. These feature vectors are finally fed into the K- nearest neighbour, support vector machine and kernel linear discriminant analysis for the detection and classification of bearing faults. With the proposed method, the reduction percentage of more than 95% percent is achieved, which not only reduces the computational burden but also the classification time. Simulation results show that the signals are classified to achieve an average accuracy of 99.13% using KLDA and 96.64% using KNN classifiers. The results are also compared with the empirical mode decomposition (EMD) features and Fourier transform features without extracting any statistical information, which are two of the most widely used approaches in the literature. To gain a certain level of confidence in the classification results, a detailed statistical analysis is also provided.

https://www.mdpi.com/1424-8220/22/5/2012/pdf?version=1646386251

A New Statistical Features Based Approach for Bearing Fault Diagnosis Using Vibration Signals

Applying computer vision to mobile robot navigation has been studied more than two decades. For the commercial off-the-shelf (COTS) automated guided vehicles (AGV) products, the cameras are still not widely used for the acquisition of guidance information from the environment. One of the most challenging problems for a vision guidance system of AGVs lies in the complex illumination conditions. Compared to the applications of computer vision where on-machine cameras are fixed in place, it is difficult to structure the illumination circumstance for an AGV that needs to travel through a large work space. In order to distinguish the original color features of path images from their illumination artifacts, an illumination-adaptive image partitioning approach is proposed based on the support vector machine (SVM) classifier with the slack constraint and the kernel function, which is utilized to divide a path image to low-, normal-, and high-illumination regions automatically. Moreover, an intelligent path recognition method is developed to carry out guide color enhancement and adaptive threshold segmentation in different regions. Experimental results show that the SVM-based classifier has the satisfactory generalization ability, and the illumination-adaptive path recognition approach has the high adaptability to the complex illumination conditions, when recognizing the path pixels in the field of view with both high-reflective and dark-shadow regions. The 98% average rate of path recognition will significantly facilitate the subsequent operation of path fitting for vision guidance of AGVs.

SVM-based image partitioning for vision recognition of AGV guide paths under complex illumination conditions

The invention discloses an electronic medical record classifying method based on convolutional neural network and active learning and a system thereof. The method comprises the steps of acquiring an entity vector which corresponds with each electronic medical record in a sample set; acquiring a marking data set and an unmarked data set; training a preset convolutional neural network model by meansof the entity vector which corresponds with the marking data set, and obtaining a trained convolutional neural network model; predicting the distribution probability of the electronic medical recorddata in the unmarked data set by means of the obtained convolutional neural network model, selecting the electronic medical record which satisfies a preset condition and adding into the marking data set; training the trained convolutional neural network model by means of an entity vector which corresponds with the electronic medical record that is added into the marking data set; after repeated training, obtaining the trained convolutional neural network model; and realizing electronic medical record classification through the trained model. According to the electronic medical record classifying method, through combining the convolutional neural network and active learning, a more accurate classification effect can be realized on the condition of relatively small number of samples.

Electronic medical record classifying method based on convolutional neural network and active learning and system thereof

Damage of road surface, e.g., Cracks, is the critical problems in road maintenance. Previous automotive road damage detection methods mainly focus on hand-crafted features and shallow classifier models. Recently, deep learning methods have also been proposed. The deep neural networks consist of dozens of parameters, which is usually optimized by the Mini-batch Stochastic Gradient Descent Algorithm (MB-SGD). However, MB-SGD is awkward for online update when new training samples from a drifting system condition, e.g., illumination, are received. In this paper, we first present an experimental study on how the illumination change affects the generalization of a pre-trained deep convolutional neural networks. Then, we propose a novel Kalman Filter based method for online updating the network parameters. Experimental results convince that the illumination change can affect the performance of a pre-trained CNN using training samples from a fixed illumination condition. By using the proposed method, the CNN can online adapt its parameters in the classifier layer to the received training samples sequentially, which leads to a better classification performance. The proposed method alleviates the need of huge amount of training samples covering all system conditions, which are hard to collect and costly.

An Online-Updating Deep CNN Method Based on Kalman Filter for Illumination-Drifting Road Damage Classification

Early and accurately detecting faults in rotating machineries is crucial for operation safety of modern manufacturing system. In this paper, we proposed a novel deep CNN method based on knowledge-transferring from shallow models for rotating machinery fault diagnosis with scarce labeled samples. It is based on the idea that shallow models trained with different hand-crafted features can reveal the latent prior knowledge or diagnostic expertise and have good generalization ability even with scarce labeled samples. First, The raw vibration signal is transformed into time-frequency domain by applying the short-time Fourier transform (STFT) to extract integral features accordingly. Then, we train the SVM model with scarce labeled samples and make predictions on unlabeled samples. The predicted labels can be regarded as the data format of expert knowledge learned by the SVM model, which are combined together with the scarce fine labeled samples. Finally, they are used to train a deep CNN model of better discriminative ability. Experimental results demonstrate the effectiveness the proposed method that it achieves better performance than SVM model and original deep CNN model trained with only scarce labeled samples. Moreover, it is computational efficient and is promising for real-time rotating machinery fault diagnosis.

Fault Diagnosis for Rotating Machinery with Scarce Labeled Samples: A Deep CNN Method Based on Knowledge-Transferring from Shallow Models

The invention discloses a small sample depth learning method based on knowledge transfer of a shallow model The invention firstly preprocesses the data, and then transforms the original signal into different transform domains according to the prior knowledge and expert experience of the related field, and calculates the artificial features According to artificial features, different shallow models are selected and trained based on a small amount of labeled sample data According to classification accuracy/prediction error and other indicators, different shallow models with different featurecombinations are screened to form candidate model pool Then, based on the candidate model pool, the model is selected to predict the unlabeled samples, and the prediction tags are obtained, and a plurality of prediction tags are fused The prediction tags are combined with a small number of existing tagged samples to construct the training set Deep neural network structure is designed for the specific task, and the training is based on the above-mentioned mixed training set The validity of the proposed method is verified by the rotating machinery fault diagnosis data set

A small sample depth learning method based on knowledge transfer of shallow model

The invention discloses a road damage identification method based on a convolutional neural network and Kalman filtering. Before the road is repaired, a road maintainer needs to research the road surface condition, and the project needs to consume huge manpower, material resources and financial resources. The method comprises the following steps: 1, preprocessing an image; And 2, image enhancementand convolutional neural network training. And 3, k = 1, 2,..., m, and executing the step 4 to the step 6 in sequence. Wherein m is the number of the detected images. And 4, amplifying the k tested image and adjusting the k tested image to 300 * 300 resolution. And 5, inputting the kth tested extended image obtained in the step 5 into the convolutional neural network obtained by training in the step 2. And 6, optimizing the weight initial value obtained in the step 5 through a Kalman filtering algorithm. According to the invention, methods such as feedforward operation, a randomgradient descent method, feedback operation, PCA dimension reduction, Kalman filtering and the like are adopted to carry out real-time parameter updating, and a convolutional neural network model with high accuracy is established.

convolutional neural network and Kalman filtering-based road damage identification method

Texture is an important visual clue in images, which is ubiquitous and hard to describe. Classical methods use different kinds of hand-crafted features and classifiers to tackle the texture detection and classification problem. With the development of deep learning, the end-to-end modeling capacity of deep neural network can achieve better performance than classical methods. In this paper, we propose a new method that utilizes Long Short-Term Memory (LSTM) units together with deep convolutional network to modelling the spatial correlations in texture image. To this end, two bilateral LSTM units are attached to the feature maps outputted by the convolutional layer of base network. Extensive experiments on two benchmark datasets DTD and FMD were conducted to compare the proposed method with state-of-the-art methods. Results convince that the proposed model has stronger representation capacity for texture images and achieves better performance. Meanwhile, it is computationally efficient with the use of GPU parallel computing. Codes will be made publicly available [29].

Mingyue Yang

Papers

An Online-Updating Deep CNN Method Based on Kalman Filter for Illumination-Drifting Road Damage Classification

Fault Diagnosis for Rotating Machinery with Scarce Labeled Samples: A Deep CNN Method Based on Knowledge-Transferring from Shallow Models

A small sample depth learning method based on knowledge transfer of shallow model

convolutional neural network and Kalman filtering-based road damage identification method

Modelling Spatial Correlations by Using Deep CNN and LSTM for Texture Image Classification