Probabilistic Neural Network Algorithm Research Articles

Intelligent processing of electromagnetic data using detrended and identification

The application of the electromagnetic method has accelerated due to the demand for the development of mineral resource, however the strong electromagnetic interference seriously lowers the data quality, resolution and detect effect. To suppress the electromagnetic interference, this paper proposes an intelligent processing method based on detrended and identification, and applies for wide field electromagnetic method (WFEM) data. First, we combined the improved intrinsic time scale decomposition and detrended fluctuation analysis algorithm for removing the trend noise. Then, we extracted the time domain characteristics of the WFEM data after removing the trend noise. Next, the arithmetic optimization algorithm was utilized to search for the optimal smoothing factor of the probabilistic neural network (PNN) algorithm, which realized to intelligently identify the noise data and WFEM effective data. Finally, the Fourier transform was performed to extract the spectrum amplitude of the effective frequency points from the reconstructed WFEM data, and the electric field curve was obtained. In these studies and applications, the fuzzy c-mean and PNN algorithm are contrasted. The proposed method indicated that the trend noise can be adaptively extracted and eliminated, the abnormal waveform or noise interference can be intelligently identified, the reconstructed WFEM data can effectively recover the pseudo-random signal waveform, and the shape of electric field curves were more stable. Simulation experiments and measured applications has verified that the proposed method can provide technical support for deep underground exploration.

A machine learning-based approach for flames classification in industrial Heavy Oil-Fire Boilers

The burner combustion tuning is a complex problem that has been studied through flame monitoring and characterization. It has been observed that the flame electromagnetic spectrum and flickering contain specific flame information in combustion processes. This information is helpful for combustion stoichiometry tuning on burners. This paper described a method for selecting the best flame feature subset that can be computed from the scanner signal, in order to get the flame index and induce combustion stoichiometry on burners under specific combustion conditions. We propose a method for selecting a reduced subset with only the useful flame features for flame index classification. To extract the most relevant flame features we use a feature subset selection (FSS) algorithm and to determine the combustion state in burners, five flame indices were defined that represent the most common flame states in oil fuel-fired boilers. FSS includes complete, sequential, and random searches in order to eliminate redundant and noisy flame features to decrease the flame feature set dimension. A probabilistic neural network (PNN) algorithm was implemented for flame feature clustering. Signals from the actual flame scanner system and relevant variables from the boiler data acquisition system were used by the algorithms to calculate the burner flame index. A set of parametric tests was done in a heavy oil-fired boiler under well-known flame and index conditions to train and test the flame classifier. The results showed that only the four more relevant features are enough to classify flames with a good performance (92.3% accuracy), which is useful for burner combustion monitoring and optimization.

Comparison of Probabilistic Neural Network (PNN) and k-Nearest Neighbor (k-NN) Algorithms for Diabetes Classification

Purpose: This study aims to compare algorithms to determine the accuracy of the algorithm and determine the speed of the algorithm used for diabetes classification.
 Methods: There are two algorithms used in this study, namely Probabilistic Neural Network (PNN) and k-Nearest Neighbor (k-NN). The data used is the Pima Indians Diabetes Database. The data contains 768 data with 8 attributes and 1 target class, namely 0 for no diabetes and 1 for diabetes. The dataset has been divided into 80% training data and 20% testing data.
 Result: Accuracy is obtained after implementing k-fold cross validation with a value of k = 4. The accuracy results show that the k-Nearest Neighbor algorithm is superior and has better quickness compared to the Probabilistic Neural Network. The k-Nearest Neighbor algorithm obtains an accuracy of 74.6% for all features and 78.1% for four features
 Novelty: The novelty of this paper is optimizing and improving accuracy which is implemented with by focusing on data preprocessing, feature selection and k-fold cross validation in the classification algorithm

Machine learning-based rail corrugation recognition: a metro vehicle response and noise perspective.

Rail corrugation is a common problemin metro lines, and its efficient recognition is always an issue worth studying. To recognize the wavelength and amplitude of rail corrugation, a particle probabilistic neural network (PPNN) algorithm is developed. The PPNN is incorporated with the particle swarm optimization algorithm and the probabilistic neural network. On the basis ofthe above, the in-vehicle noise characteristics measured in the field are used to recognize normal rail wavelengths of 30 and 50 mm. A stepwise moving window search algorithm suitable for selecting features with a fixed order was developed to select in-vehicle noise features. Sound pressure levels at 400, 500, 630 and 800 Hz of in-vehicle noise are fed into the PPNN, and the average accuracy can reach 96.43%. The bogie acceleration characteristics calculated by the multi-body dynamics simulation model are used to recognize normal rail amplitudes of 0.1 and 0.2 mm. The bogie acceleration is decomposed by the complete ensemble empirical mode decomposition with adaptive noise, and a reconstructional signal is obtained. The energy entropy of the reconstructional signal is fed into the PPNN, and the average accuracy can reach 95.40%. This article is part of the theme issue 'Artificial intelligence in failure analysis of transportation infrastructure and materials'.

Research on Intelligent Identification of Magnetizing Inrush Current based on Empirical Modal Decomposition

The accurate identification of the converter transformer magnetizing inrush current can ensure the safety of the operation of the UHV DC transmission system. In the differential protection of the converter transformer, it is extremely hard to distinguish between the excitation inrush current generated when the transformer is closed and the Fault current due to internal fault, which leads to frequent misoperation of the transformer protection to the extent that the transformer access fails. This paper proposes a current identification model based on empirical modal decomposition (EMD) and probabilistic neural network (PNN) algorithm with particle swarm (PSO) optimization. The EMD algorithm decomposes the current waveform, and the eigenmode functions (IMFs) of the three highest correlation bars are selected. The energy, cliffs, and standard deviation of the IMFs are extracted as the original waveform eigenvectors.

Transformer fault diagnosis based on probabilistic neural networks combined with vibration and noise characteristics

When the transformer is running, the vibration which is generated in the core and winding will spread outward through the medium of metal, oil, and air. The magnetic field of the core changes with the variation of the transformer excitation source and the state of the core, so the corresponding vibration and noise change. Therefore, the vibration and noise of the transformer contain a lot of information. If the information can be associated with the fault characteristics of the transformer, it is significant to evaluate the running state of the transformer through the vibration and noise signal, which improve the intelligence, safety, and stability of the transformer operation. Based on this, modeling and simulation of transformer multi-point grounding, DC bias, and short-circuit between silicon steel sheets fault are first carried out in this paper, and vibration and noise distribution of transformer under different faults are given. Second, a fault diagnosis method based on transformer vibration and noise characteristics is proposed. In the process of implementation, vibration and noise signals under multi-point grounding, DC bias, and short-circuit between silicon steel sheets are taken as the sample data, and the probabilistic neural network algorithm is used to effectively predict the transformer fault. Finally, the effectiveness of the proposed scheme is verified by identifying the simulation faults-the proposed fault diagnosis method based on PNN can be effectively applied to transformer.

The Vehicle Intention Recognition with Vehicle-Following Scene Based on Probabilistic Neural Networks

In the vehicle-following scenario of autonomous driving, the change of driving style in the front vehicle will directly affect the decision on the rear vehicle. In this paper, a strategy based on a probabilistic neural network (PNN) for front vehicle intention recognition is proposed, which enables the rear vehicle to obtain the driving intention of the front vehicle without communication between the two vehicles. First, real vehicle data with different intents are collected and time—frequency domain variables are extracted. Secondly, Principal Component Analysis (PCA) is performed on the variables in order to obtain comprehensive features. Meanwhile, two cases are classified according to whether the front vehicle can transmit data to the rear vehicle. Finally, two recognition models are trained separately according to a PNN algorithm, and the two models obtained from the training are verified separately. When the front vehicle can communicate with the rear vehicle, the recognition accuracy of the corresponding PNN model reaches 96.39% (simulation validation) and 95.08% (real vehicle validation). If it cannot, the recognition accuracy of the corresponding PNN model reaches 78.18% (simulation validation) and 73.74% (real vehicle validation).

Использование алгоритма вероятностной нейронной сети для измерения уровня качества образования

The traditional method of determining the quality of education is too unambiguous and unreasonable, which is not suitable for a comprehensive assessment of students' abilities. The purpose of the article is to justify the use of a probabilistic neural network algorithm. Research methods. The reliability of the presented results is ensured by the analysis of scientific literature, modeling of a probabilistic neural network, comparative analysis of models and evaluation of the effectiveness of the model. Research results. In this paper, a probabilistic neural network (PNN) algorithm is used to determine the quality of education by considering the important influence between different student achievements. The PNN algorithm comes from the Bayesian decision rule and uses the nonlinear Gauss Parsen window as a probability density function. Since the PNN model has strong nonlinear and anti-interference properties, it is suitable for determining the quality of education by classifying student achievements. In addition, this article also discusses the impact of various evaluation models on the accuracy and effectiveness of classification. In addition, the influence of the spread value on the PNN model is also discussed. Scope of application. Finally, evidence is used to determine the quality of education. Conclusions. Experimental results show that the detection accuracy can reach 95%, and the detection time is only 0.0156 s based on the proposed method. That is, the method is a very practical detection algorithm with high accuracy and efficiency. In addition, it also contains information on how to further improve the quality of teaching. It is proved that the use of the PNN model makes it possible to accurately classify the achievements of students according to the quality criterion.

Correlation Filters in Machine Learning Algorithms to Select Demographic and Individual Features for Autism Spectrum Disorder Diagnosis

Autism spectrum disorder is currently considered one of the main neurodevelopmental disorders with predominant characteristics of difficulty in social communication and cognitive skills, and limited and repetitive patterns. This disorder has no cure and has different levels of severity that vary according to the appearance of symptoms in each patient. Generally, the waiting time for the diagnosis of autism spectrum disorder is slow, having as one of the reasons for this situation the lack of development of simple screening procedures to be implemented and which have efficient results. The objective of this work is to analyze a public database in order to find patterns of the autism spectrum, that is, to isolate the attributes that together with behavioral characteristics can bring greater reliability to the precursor model. The preliminary results showed that the probabilistic neural network algorithm performed well in this classification. In addition, the application of correlation filters demonstrated greater efficiency in accuracy. By applying eight data mining algorithms and aggregating the demographic, individual, and behavioral attributes, and excluding some attributes, we obtained an accuracy of 100% through the support vector machine. Finally, the results with machine learning have shown that the patient’s ethnicity, continent, and the presence of jaundice tend to reveal more likely that the patient will be diagnosed with autism spectrum disorder.

Intelligent Metal Welding Defect Detection Model on Improved FAST-PNN

In order to solve the problem of accurate and efficient detection of welding defects in the process of batch welding of metal parts, an improved Probabilistic Neural Network (PNN) algorithm was proposed to build an automatic identification model of welding defects. Combined with the characteristics of the PNN model, the structure and algorithm flow of the FAST-PNN algorithm model are proposed. Extraction of welding defect image texture features of metal welded parts by a Gray Level Co-occurrence Matrix (GLCM) screens out the characteristic indicators that can effectively characterize welding defects. Weld defect texture features are used as input to build a defect classification model with FAST-PNN, for accurate and efficient classification of welding defects. The results show that the improved FAST-PNN model can effectively identify the types of welding defects such as burn-through, pores and cracks, etc. The classification recognition accuracy and recognition efficiency have been significantly improved. The proposed defect welding identification method can accurately and effectively identify the damage types of welding defects based on a small number of defect sample images. Welding defects can be quickly identified and classified by simply collecting weld images, which helps to solve the problem of intelligent, high-precision, fast real-time online detection of welding defects in modern metal structures; it provides corresponding evidence for formulating response strategies, with a certain theoretical basis and numerical reference.

The Identification of ECG Signals Using Wavelet Transform and WOA-PNN.

Electrocardiogram (ECG) signal identification technology is rapidly replacing traditional fingerprint, face, iris and other recognition technologies, avoiding the vulnerability of traditional recognition technologies. This paper proposes an ECG signal identification method based on the wavelet transform algorithm and the probabilistic neural network by whale optimization algorithm (WOA-PNN). Firstly, Q, R and S waves are detected by wavelet transform, and the P and T waves are detected by local windowed wavelet transform. The characteristic values are constructed by the detected time points, and the ECG data dimension is smaller than that of the non-reference detection. Secondly, combined with the probabilistic neural network, the mean impact value algorithm is used to screen the characteristic values, the characteristic values with low influence are eliminated, and the input and complexity of the model are simplified. Finally, a WOA-PNN combined classification method is proposed to intelligently optimize the hyper parameters in the probabilistic neural network algorithm to improve the model accuracy. According to the simulation verification on three databases, ECG-ID, MIT-BIH Normal Sinus Rhythm and MIT-BIH Arrhythmia, the identification accuracy of a single ECG cycle is 96.97%, and the identification accuracy of three ECG cycles is 99.43%.

Research on the Effectiveness of Probabilistic Stochastic Convolution Neural Network Algorithm in Physical Education Teaching Evaluation.

In practice, PE teaching evaluation based on probabilistic convolutional neural network still faces some practical problems. At present, the existing research mainly focuses on how to improve the accuracy of PE (physical education) teaching evaluation, but ignores the balance between accuracy and speed of the model, which is the key to achieve efficient PE teaching estimation. Aiming at the problem of optimization contradiction existing in the traditional probabilistic stochastic convolution neural network regression method, a position adaptive probabilistic stochastic convolution neural network regression method was proposed. Firstly, the basic principle of probabilistic and random convolution neural network regression method is given. Secondly, the contradiction and reasons between hot trial regression and coordinated regression are analyzed. It is found that the process heat will return to the optimization with irreconcilable contradiction with the coordinates due to the lack of learning parameters when the hot trial transforms the coordinates. The optimization contradiction will make the model tonot obtain the exact coordinates of the nodes. Then, based on the above analysis, the learnable parameters are introduced into the Softmax function, and the position adaptive Softmax model is proposed. Combining the model with the probabilistic stochastic convolution neural network regression method, the position adaptive probabilistic stochastic convolution neural network integral regression method is obtained. In order to reduce the training cost of this method, a simplified training strategy is proposed. Finally, the simulation software MATLAB is used for verification, and the functions of sample maintenance, probabilistic stochastic convolution neural network training, and neural network evaluation are realized. The experimental data show that the probabilistic stochastic convolutional neural network is feasible for teaching quality evaluation, meets the accuracy requirements, and indeed provides a convenient and practical tool for PE teaching quality evaluation.

A Gray Wolf Optimization-Based Improved Probabilistic Neural Network Algorithm for Surrounding Rock Squeezing Classification in Tunnel Engineering

Surrounding rock squeezing deformation is a common and prominent hazard in tunnel engineering projects, which often induces the shield jamming disaster during the TBM tunneling process. Based on the 139 groups of historical squeezing deformation cases, this study developed a hybrid PCA-IWGO-PNN model for squeezing classification. According to the influencing factors and characteristics of squeezing deformation, the strength-stress ratio, tunnel burial depth, tunnel equivalent diameter, rock mass quality index, and support stiffness were selected to establish the prediction index system of squeezing level. Because the probabilistic neural network (PNN) requires that the input variables are independent, principal component analysis (PCA) was used to preprocess the original data to eliminate the correlation between prediction indexes and achieve dimensionality reduction. The spread coefficient was the critical hyper-parameter in the PNN, and the improved gray wolf optimization (IGWO) algorithm was used to realize its efficient automatic optimization. Then, the PNN model was applied to engineering practice. Only 1 of 20 test samples was misjudged, achieving the 95% prediction accuracy. Finally, the comparison analysis with the artificial neural network (ANN) model, support vector machine (SVM) model, and random forest (RF) model was conducted. Among them, the PNN model achieved the highest prediction accuracy, followed by the artificial neural network (85%), RF (85%), and SVM (80%). In addition, the PNN model had the fastest running speed, which only consumed 5.6350 s, while the running time of ANN, SVM, and RF was 8.8340, 6.2290, and 6.9260 s, respectively. The hybrid PCA-IWGO-PNN model developed in this research provides an effective method for surrounding rock squeezing classification, and it has superiorities in both prediction accuracy and running speed.

Study on the parameterization response of probabilistic neural networks for seismic facies classification in the Gulf of Mexico

Nowadays, there are many unsupervised and supervised machine learning techniques available for performing seismic facies classification. However, those classification methods either demand high computational costs or do not provide an accurate measure of confidence. Probabilistic neural networks (PNNs) overcome these limitations and have demonstrated their superiority among other algorithms. PNNs have been extensively applied for some prediction tasks, but they have not been well studied regarding the prediction of seismic facies volumes using seismic attributes. We have explored the capability of the PNN algorithm when classifying large- and small-scale seismic facies. In addition, we evaluate the impact of user-chosen parameters on the final classification volumes. After performing seven tests, each with a parameter variation, we assess the impact of the parameter change on the resultant classification volumes. We find that the processing task can have a significant impact on the classification volumes, but we also find how the most geologically complex areas are the most challenging for the algorithm. Moreover, we determine that even if the PNN technique is performing and producing considerably accurate results, it is possible to overcome those limitations and significantly improve the final classification volumes by including the geologic insight provided by the geoscientist. We conclude by proposing a new workflow that can guide future geoscientists interested in applying PNNs, to obtain better seismic facies classification volumes by considering some initial steps and advice.

The Use of Artificial Intelligence in the Construction of Operation Safety Early Warning Management System of Urban Rail Transit

In order to reduce the risk of safety management of urban rail transit (URRT) and improve the safety early warning of URRT, artificial intelligence (AI) is used to construct the early warning management system of URRT. Firstly, the four factors that affect the safety of URRT are discussed, the URRT system is divided into several modules, and the functions of each module are introduced. Secondly, the safety early warning analysis module is designed based on neural network technology to make it have the ability of safety early warning. Finally, the security prediction system is tested. The results show that the overall prediction error of probabilistic neural network algorithm used is less than 5%, and the classification accuracy of PNN is about 50%, which can achieve excellent prediction of data. The safety early warning system is used to predict the passenger flow of a station, and compared with the actual passenger flow, the prediction model has good prediction accuracy and credibility. Therefore, the safety early warning management system of URRT designed can meet the needs of practical application. This paper provides a reference for the safety management of URRT.

Mode recognition and coordinated magnetisation control method for variable flux memory machine

The hybrid permanent magnet (PM) variable flux memory machine employs high- and low-coercive-force hybrid PMs to simultaneously obtain a high torque density and regulable PM flux property. The operating trajectories and steady working-points of this type of machine switch back and forth in multiple regions. To identify operating modes under various PM magnetisation states and speed ranges, this paper proposes a mode recognition and coordinated control method based on probabilistic neural networks (PNNs). A PNN algorithm is trained offline and executed in real-time to obtain a certain operating mode with the maximum probability of current state. According to the mode recognition result, a coordinated control scheme for applying short-time pulses or continuous flux-weakening current in a reasonable manner is presented. The proposed method exhibits an excellent capability in real-time operating modes identification, hence achieving magnetisation current control of multiple modes easily. Its feasibility and effectiveness are validated by experimental results on a hybrid PM VFMM prototype.

Exhaustive probabilistic neural network for attribute selection and supervised seismic facies classification

Machine learning (ML) algorithms, such as principal component analysis, independent component analysis, self-organizing maps, and artificial neural networks, have been used by geoscientists to not only accelerate the interpretation of their data, but also to provide a more quantitative estimate of the likelihood that any voxel belongs to a given facies. Identifying the best combination of attributes needed to perform either supervised or unsupervised ML tasks continues to be the most-asked question by interpreters. In the past decades, stepwise regression and genetic algorithms have been used together with supervised learning algorithms to select the best number and combination of attributes. For reasons of computational efficiency, these techniques do not test all of the seismic attribute combinations, potentially leading to a suboptimal classification. In this study, we have developed an exhaustive probabilistic neural network (PNN) algorithm that exploits the PNN’s capacity in exploring nonlinear relationships to obtain the optimal attribute subset that best differentiates target seismic facies of interest. We determine the efficacy of our proposed workflow in differentiating salt from nonsalt seismic facies in a Eugene Island seismic survey, offshore Louisiana. We find that from seven input candidate attributes, the exhaustive PNN is capable of removing irrelevant attributes by selecting a smaller subset of four seismic attributes. The enhanced classification using fewer attributes also reduces the computational cost. We then use the resulting facies probability volumes to construct the 3D distribution of the salt diapir geobodies embedded in a stratigraphic matrix.

Maritime automatic target recognition for ground-based scanning radars by using sequential range profiles

Classification of marine targets using radar data products has become an important area for modern researchsociety. However, due to several reasons such as the similarity between ship structures and spatial specifications,classification of marine targets constitutes a challenging problem. In almost all of the studies, this problem has beenhandled by focusing on a single instance of range profiles or synthetic aperture radar data. However, this approachis seen to achieve only a particular success. This study introduces a novel classification approach that is shown toprovide additional classification enhancements by exploiting the extra information extracted from sequential rangeprofiles generated by ground-based marine surveillance radars. With this purpose, both synthetic and measurementalrange profiles are taken into consideration. Synthetic profile data are generated for seven marine targets by using anelectromagnetic scattering simulation tool (RASES)1. On the other hand, a total of 2387 range profile data of 171different target tracks are collected for five different marine target class types by using an X-band marine surveillanceradar. Each target tracked for a long period of time to gather sequential HRRP data subsets. HRRP data subsets areused to generate HMM based transition matrix probabilities and sequential classification results by evaluating proposedmethod. Probabilistic neural network (PNN) and convolutional neural network (CNN) classification algorithms appliedto gather classification results. The proposed method results are compared with both single value classification andmajority voting rule (MVR) method results. According to the examination results, the proposed classification approachprovides remarkable enhancements in the correct classification rates when compared to the case of single profile dataapproach.

Object Classification with Roadside LiDAR Data Using a Probabilistic Neural Network

Object classification is important information for different transportation areas. This research developed a probabilistic neural network (PNN) classifier for object classification using roadside Light Detection and Ranging (LiDAR). The objective was to classify the road user on the urban road into one of four classes: Pedestrian, bicycle, passenger car, and truck. Five features calculated from the point cloud generated from the roadside LiDAR were selected to represent the difference between different classes. A total of 2736 records (2062 records for training, and 674 records for testing) were manually marked for training and testing the PNN algorithm. The data were collected at three different sites representing different scenarios. The performance of the classification was evaluated by comparing the result of the PNN with those of the support vector machine (SVM) and the random forest (RF). The comparison results showed that the PNN can provide the results of classification with the highest accuracy among the three investigated methods. The overall accuracy of the PNN for object classification was 97.6% using the testing database. The errors in the classification results were also diagnosed. Discussions about the direction of future studies were also provided at the end of this paper.

Power quality disturbance classification based on efficient adaptive Arrhenius artificial bee colony feature selection

Automatic power quality disturbance (PQD) classification is a challenging task for industry and utility. This work presents an algorithm for classifying the PQDs based on the combination of novel feature selection with machine learning (ML) techniques. In this work, discrete wavelet transform (DWT) is utilized to decompose the PQD signals and hence minimize the size of input vectors with multi-resolution analysis. Novel optimal feature selection-based classification is performed using probabilistic neural network and adaptive Arrhenius artificial bee colony (PNN- adaptive AABC) algorithm. The selection of optimal features may discard the redundant features and retain the useful features. Also, this adaptive AABC algorithm provides a higher convergence rate, and this is used for accurate feature selection since this work is applied for 16 PQD events. In other case, features extracted with DWT are processed and trained with various ML algorithms such as Decision Tree, support vector machine (SVM), and K-nearest neighbor (KNN). Finally, the comparison accuracy of the proposed classifier is compared with the existing classifiers. The proposed technique found to give improved results in most of the cases like accuracy, prediction speed, and training time, and the outcomes acquired as 99.98%, ~1260 obs/seconds, and 13.113 seconds, respectively.