Probabilistic Neural Network Research Articles

Development of a recommendation engine for university study programme selection: A regression-based approach

This paper is the third in a series focused on bridging the gap between secondary and higher education. Our primary objective is to develop a robust theoretical framework for an innovative e-business model called the Undergraduate Study Programme Search System (USPSS). This system considers multiple criteria to reduce the likelihood of exam failure or the need for multiple retakes, while maximizing the chances of successful program completion. Testing of the proposed algorithm demonstrated that the Stochastic Gradient Boosted Regression Trees method outperforms the current method used in Lithuania for admitting applicants to 47 educational programs. Specifically, it is more accurate than the Probabilistic Neural Network for 25 programs, the Ensemble of Regression Trees for 24 programs, the Single Regression Tree for 18 programs, the Random Forest Regression for 16 programs, the Bayesian Additive Regression Trees for 13 programs, and the Regression by Discretization for 10 programs.

Model-Assisted Probabilistic Neural Networks for Effective Turbofan Fault Diagnosis

A diagnostic method for gas-path faults of turbofan engines, relying on a Probabilistic Neural Network (PNN) coupled with a thermodynamic model of the engine, is presented. The novel aspect of the method is that its training information is generated dynamically by an accompanying Engine Performance Model. In the proposed approach, the PNN efficiently addresses the first step of a diagnostic process (i.e., detection of the faulty component at the current operating point), while with the aid of an adaptive engine model, the fault is then further isolated and identified. A description of the proposed method and training aspects of the PNN are presented. The method is applied to the case of a mixed-flow turbofan engine to diagnose common gas-path faults in compressors and turbines (i.e., fouling, FOD, erosion, and tip clearance). Its performance is evaluated using realistic fault data that may be acquired at various operating conditions within a flight envelope.

Estimating The Health of Helicopter Turbine Engines by Means of Regression and Classification Using a Probabilistic Neural Network

This paper presents team Mad SoftMax’s approach to competing in the Conference of the Prognostics and Health Management Society 2024 Data Challenge. The competition tasked participants with estimating the health of helicopter turbine engines by means of regression and classification. Classification was used to categorize each observation within the dataset as belonging to a healthy or faulty engine while probabilistic regression was employed to estimate the torque margin at each measurement. Additionally, teams were challenged to provide a confidence metric to each of their estimations. These metrics reflected the trustworthiness of solutions and added a risk versus reward element to the competition. While the complete dataset for the challenge contained seven engines, four were provided in a training set to encourage robustness of solutions to handle the three withheld engines’ data during the testing and validation phase. The data was scrambled and identifiers to specific engines were removed to not give away patterns specific to a given engine. Open-source libraries such as TensorFlow were utilized to develop classification and regression models and the following paper opines on the process of understanding the data, data cleaning and model evaluation.

A Novel Neuro-Probabilistic Framework for Energy Demand Forecasting in Electric Vehicle Integration

This paper presents a novel grid-to-vehicle modeling framework that leverages probabilistic methods and neural networks to accurately forecast electric vehicle (EV) charging demand and overall energy consumption. The proposed methodology, tailored to the specific context of Medellin, Colombia, provides valuable insights for optimizing charging infrastructure and grid operations. Based on collected local data, mathematical models are developed and coded to accurately reflect the characteristics of EV charging. Through a rigorous analysis of criteria, indices, and mathematical relationships, the most suitable model for the city is selected. By combining probabilistic modeling with neural networks, this study offers a comprehensive approach to predicting future energy demand as EV penetration increases. The EV charging model effectively captures the charging behavior of various EV types, while the neural network accurately forecasts energy demand. The findings can inform decision-making regarding charging infrastructure planning, investment strategies, and policy development to support the sustainable integration of electric vehicles into the power grid.

Information-guided adaptive learning approach for active surveillance of infectious diseases

The infectious disease surveillance system is a key support tool for public health decision making. Current research concentrates on optimizing static sentinel deployment to address the problem of incomplete data due to the lack of sufficient surveillance resources. In this study, we introduce an information-guided adaptive learning strategy for the dynamic surveillance of infectious diseases. The goal is to improve monitoring effectiveness in situations where it is possible to adjust the focus of surveillance, such as serial surveys and allocation of testing tools. Specifically, we develop a probabilistic neural network model to learn spatio-temporal correlations among the numbers of infections. Based on a probabilistic model, we evaluate the information gain of monitoring a spatio-temporal target and design a greedy selection algorithm for monitoring targets selection. Moreover, we integrate two major surveillance objectives, i.e., informativeness and coverage, in the monitoring target selection. The experimental results on the synthetic dataset and two real-world datasets demonstrate the effectiveness of our approach, showcasing the promise of further exploration and application of dynamic adaptive active surveillance.

Phishing URL detection with neural networks: an empirical study

Cybercriminals create phishing websites that mimic legitimate websites to get sensitive information from companies, individuals, or governments. Therefore, using state-of-the-art artificial intelligence and machine learning technologies to correctly classify phishing and legitimate URLs is imperative. We report the results of applying deterministic and probabilistic neural network models to URL classification. Key achievements of this work are: (1) The development of a unique approach based on probabilistic neural networks that improves classification accuracy. (2) We show for the first time in URL phishing research that a machine learning model trained on a combination of open source and private datasets is successful in production. The dataset is constructed from open sources like Alexa, PhishTank, or OpenPhish and, most importantly, real-world production data from EasyDMARC. The daily validation of the model using daily reported URL data and corresponding labels, both from open-source platforms and private production, reach on average a 97% accuracy on the validation dataset, labeled by PhishTank, OpenPhish and EasdDMARC where possible mislabeled data can not be excluded and was not possible to check due to large number of URLs. Feature engineering was done without third-party dependencies. Lastly, the evaluation of both deterministic and probabilistic models shows high accuracy on short and long URLs, where short URLs are defined as having less than 50 characters.

Feasibility of Nondestructive Soluble Sugar Monitoring in Tomato: Quantified and Sorted through ATR-FTIR Coupled with Chemometrics

As a fast detection method, Fourier transform infrared attenuated total reflection (ATR-FTIR) spectroscopy is seldom used for monitoring soluble sugars in crops. This study aimed to demonstrate the feasibility of leveraging ATR-FTIR coupled with chemometrics to quantify and sort the contents of soluble sugar in tomatoes. Firstly, 192 tomato samples were scanned using ATR-FTIR; subsequently, a quantitative model was developed using PLSR with selected wavelength variables as inputs. Finally, a classification model was estimated through probabilistic neural network (PNN) to determine the samples. The results indicated that ATR-FTIR had successfully captured the spectra from the cellular layers of tomatoes, resulting in a robust PLSR model created by 468 selected variables with a R² value of 0.86, a RMSEP of 0.71%, a ratio of performance to relative percent deviation (RPD) of 1.87, and a ratio of prediction to interquartile range (RPIQ) of 2.1. Meanwhile, the PNN model demonstrated a high rate correct (RC) of 92.17% in identifying whether the samples with a higher soluble sugar content than the limit of detection (LOD at 2.1%). Overall, ATR-FTIR coupled with chemometrics has proven effective for non-destructive determination of soluble sugars in tomatoes, offering new insights into internal monitoring techniques for crop quality assurance.

Unlocking thin sand potential: a data-driven approach to reservoir characterization and pore pressure mapping

The gradual decline of production of the major fields of the Indus Basin has influenced the field pressure. Therefore, the prospects need to be assessed by the petro-elastic relationship. This study highlights the value-addition for identifying gas pockets within the E-Sand of the Lower Goru Formation through estimating and populating petro-elastic properties. A probabilistic neural network (PNN) was employed to develop the relationship of pore pressure to the inverted properties. The outcomes will help identify the high-pressure areas in accordance with the petro-elastic relationship for optimizing production strategy. In order to evaluate the petro-elastic relationship, stochastic seismic inversion was employed to holistically capture the reservoir’s variability. Pore pressure volumetric was estimated from inverted attributes using a PNN, a non-linear algorithm that was more suited to heterogeneous reservoirs. The main objective of the research is achieved via enhanced resolution of elastic properties attained through the integration of stochastic inversion and PNN processes. The high-resolution elastic properties including P-impedance, S-impedance, and Vp/Vs ratio can delineate the heterogeneities more profoundly. The petrophysical volumes obtained via enhanced inverted properties combined with pore-pressure through PNN illuminated the potential facies and the correlations in predicting the properties. The procedure provides a linkage of elastic, petrophysical, and geomechanical properties that is helpful for professionals covering a broad field of the petroleum sector. The developed workflow can be followed globally where problems occur regarding heterogeneities and early depletion.

A probabilistic neural network assessment method for insulator pollution level based on discharge noise

The insulators of a transmission line are exposed to air for a long time, a mount of dust particles will be deposited on the surface. When there rain, fog or high humidity, the surface of the insulator will discharge or even flashover to breakdown. For this, in this paper, the characteristic parameters of insulator discharge noise under different pollution levels, ambient temperature and relative humidity are extracted, and the influence of insulator discharge noise are analyzed respectively. Probabilistic neural network (PNN) model training samples are used to propose an insulator pollution level assessment method based on discharge noise and verify it. The results show that the pollution level has obvious influence on the characteristic parameters of insulator discharge noise. The higher the pollution level, the more obvious the discharge phenomenon and the greater the discharge intensity. When the relative humidity ranges from 70 % to 90 %, the greater the relative humidity, the greater the pollution discharge intensity of the insulator. The energy of the noise signal gradually concentrated in the 5.5 kHz to 8.3 kHz frequency band, and the pollution discharge in the 5.5 kHz to 16.5 kHz frequency band reflected in the low-frequency part of the information increased. Ambient temperature has no obvious effect on insulator discharge noise. The precision rate of the assessment method for pollution levels Ⅰ, Ⅱ, Ⅲ and Ⅳ are 91.5 %, 88.2 %, 86.5 % and 94.0 %, respectively.

A parallel and multi-scale probabilistic temporal convolutional neural networks for forecasting the key monitoring parameters of gas turbine

As a crucial equipment in the power industry, gas turbines need effective condition monitoring techniques to maintain safe and reliable operations. Fast yet accurate long-term forecasting of the key monitoring parameters of gas turbine is vital for achieving the overall effective condition monitoring. This however poses challenges in terms of both efficacy and efficiency for conventional time series models like recurrent neural networks (RNN), since they run in a sequential manner. To this end, this paper introduces a novel parallel probabilistic time series prediction model. Particularly, the proposed model leverages the parallelism of temporal convolutional neural network (TCN) and exploits the power of latent space in the Bayesian framework. Besides, a multi-scale feature extraction strategy based on dense connections and a lagged observation strategy are presented to improve the model performance. In the comparative study against state-of-the art competitors on four classical system identification benchmarks, the proposed model significantly reduces the training time, while consistently achieving highly accurate predictions and providing appropriate uncertainty quantification. Thereafter, the proposed model is adopted to build a systematic workflow for the fast yet accurate long-term forecasting of the temperature of blade channel of gas turbine. The comprehensive results again highlight the superiority of the proposed model in terms of both the prediction quality and the training efficiency.

Coupling physics in artificial neural network to predict the fatigue behavior of corroded steel wire

To accurately predict the fatigue life of corroded steel wire, the authors proposed an artificial neural network (ANN) model and a probabilistic physics-guided neural network (PPgNN) model. Factors including stress range, mean stress, and corrosion rate were considered as input features of these two neural networks. The ANN model exhibited the best prediction accuracy with a determination coefficient of 0.94; however, it cannot capture the dispersion of fatigue life. Through introduction of physics constraints, PPgNN model was able to predict the standard deviation of fatigue life. The results showed that 94.79% of the dataset was within the 95% confidence interval.

A Probabilistic Neural Network Assessment Method for Insulator Pollution Levels Based on Infrared Images

A Probabilistic Neural Network Assessment Method for Insulator Pollution Levels Based on Infrared Images

An Improved Convolutional Neural Network for Pipe Leakage Identification Based on Acoustic Emission

Oil and gas pipelines are the lifelines of the energy market, but due to long-term use and environmental factors, these pipelines are prone to corrosion and leaks. Offshore oil and gas pipeline leaks, in particular, can lead to severe consequences such as platform fires and explosions. Therefore, it is crucial to accurately and swiftly identify oil and gas leaks on offshore platforms. This is of significant importance for improving early warning systems, enhancing maintenance efficiency, and reducing economic losses. Currently, the efficiency of identifying leaks in offshore platform pipelines still needs improvement. To address this, the present study first established an experimental platform to simulate pipeline leaks in a marine environment. Laboratory leakage signal data were collected, and on-site noise data were gathered from the “Liwan 3-1” offshore oil and gas platform. By integrating leakage signals with on-site noise data, this study aimed to closely mimic real-world application scenarios. Subsequently, several neural network-based leakage identification methods were applied to the integrated dataset, including a probabilistic neural network (PNN) combined with time-domain feature extraction, a Backpropagation Neural Network (BPNN) optimized with simulated annealing and particle swarm optimization, and a Long Short-Term Memory Network (LSTM) combined with Mel-Frequency Cepstral Coefficients (MFCC). Corresponding models were constructed, and the effectiveness of leak detection was validated using test sets. Additionally, this paper proposes an improved convolutional neural network (CNN) leakage detection technology named SART-1DCNN. This technology optimizes the network architecture by introducing attention mechanisms, transformer modules, residual blocks, and combining them with Dropout and optimization algorithms, which significantly enhances data recognition accuracy. It achieves a high accuracy rate of 99.44% on the dataset. This work is capable of detecting pipeline leaks with high accuracy.

Prospects of Computational Intelligence in Society: Human-Centric Solutions, Challenges, and Research Areas

The future of society could be shaped by artificial intelligence (AI), which is a part of soft computing that deals with probabilistic reasoning, neural networks, fuzzy logic, and evolutionary algorithms. This paper focuses on issues that are faced by soft computing as it moves towards human-centric solutions, outlines the challenges for soft computing, and provides possible research directions for development. Currently, there is an increasing emphasis on human-centered solutions in various fields, such as the healthcare industry, transportation sector, financial institutions, and education systems in the world today. Using soft computing techniques in this way comes with its own advantages, especially when managing unclearness or ambiguity or unsureness existing within humans' decision-making processes. Soft computing based on using neural networks may help to develop computer systems capable of faking human behavior and preferences as they evolve over time by using fuzzy logic and evolutionary algorithms. This human-centric approach helps improve the user experience while creating more symbiotic ties between technology and society, leading to examples such as personalized medicine treatment plans, among others. Soft computing’s future in society will be, to a large extent, determined by the provision of human-friendly AI while addressing the ethical, legal, and social challenges of its applications. Rights on AI represent humanity and shall be accountable. Collaborating with interdisciplinary scientists and carrying out frontier studies on developing human-friendly AI systems from ethical, legal, and social perspectives and welfare has the potential to have an influence on community and policy-maker decisions, facilitate a suitable climate for soft computing and AI in society, leverage it for empowering individuals, communities, and societies, and to improve the quality of human life while respecting human values and dignity. It also paves the way for society to be more inclusive and equitable.

Harnessing probabilistic neural network with triple tree seed algorithm-based smart enterprise quantitative risk management framework

Enterprise risk management (ERM) frameworks convey vital principles that help create a consistent risk management culture, irrespective of employee turnover or industry standards. Enterprise Management System (EMS) are becoming a popular research area for assuring a company’s long-term success. Statistical pattern recognition, federated learning, database administration, visualization technology, and social networking are all used in this field, which includes artificial intelligence (AI), data science, and statistics. Risk assessment in EMS is critical for enterprise decision-making to be effective. Recent advancements in AI, machine learning (ML), and deep learning (DL) concepts have enabled the development of effective risk assessment models for EMS. This special issue seeks groundbreaking research articles that showcase the application of applied probability and statistics to interdisciplinary studies. This study offers Improved Metaheuristics with a Deep Learning Enabled Risk Assessment Model (IMDLRA-SES) for Smart Enterprise Systems. Using feature selection (FS) and DL models, the provided IMDLRA-SES technique estimates business risks. Preprocessing is used in the IMDLRA-SES technique to change the original financial data into a usable format. In addition, an FS technique based on oppositional lion swarm optimization (OLSO) is utilized to find the best subset of features. In addition, the presence or absence of financial hazards in firms is classified using the triple tree seed algorithm (TTSA) with a probabilistic neural network (PNN) model. The TTSA is used as a hyperparameter optimizer to improve the efficiency of the PNN-based categorization. An extensive set of experimental evaluations is performed on German and Australian credit datasets to illustrate the IMDLRA-SES model’s improved performance. The performance validation of the IMDLRA-SES model portrayed a superior accuracy value of 95.70% and 96.09% over existing techniques.

Permanent magnet synchronous motor demagnetization fault diagnosis based on PCA-ISSA-PNN

Aiming at the demagnetization fault problem of the permanent magnet synchronous motor (PMSM), a demagnetization fault diagnosis method based on the combination of the principal component analysis (PCA) algorithm, the improved sparrow search algorithm (ISSA), and the probabilistic neural network (PNN) algorithm is proposed. First, the principal components of phase currents are extracted using PCA. Second, ISSA is used to optimize the smoothing coefficients of the PNN algorithm, and the optimized PNN algorithm is combined with PCA to obtain the PCA-ISSA-PNN fault diagnosis model. Finally, the established fault diagnosis model was tested using the current data collected from the experiments and compared with the fault diagnosis indexes and optimization performance of the conventional PNN, PCA-PNN, PCA-GA (genetic algorithm)-PNN, PCA-DA (dragonfly algorithm)-PNN, PCA-GTO (artificial gorilla troop optimizer)-PNN, PCA-AHA-PNN, and PCA-SSA-PNN. The test results show that the fault diagnosis accuracy of PCA-ISSA-PNN reaches 95.83%, and the fault diagnosis indexes are significantly higher than those of PNN, PCA-PNN, PCA-GA-PNN, and PCA-DA-PNN; its optimization performance is also significantly better than that of PCA-GTO-PNN, PCA-AHA-PNN, and PCA-SSA-PNN, which verifies the accuracy and efficiency of the proposed method.

Estimation of reservoir properties using pre-stack seismic inversion and neural network in mature oil field, Upper Assam basin, India

The mature oil fields require comprehensive characterization for enhanced hydrocarbon production, and subsequently demands estimation of reservoir properties. The key properties viz. volume of clay, effective-porosity, hydrocarbon-saturation has been evaluated for an aging Oligocene reservoir of Upper Assam basin, located in northeastern India from seismic and well log data. Elastic properties (acoustic and shear impedance) and density are derived from pre-stack inversion of 3D seismic data. These elastic properties are analyzed for their sensitivity for discrimination of lithology and fluid-content, and many derived attributes are computed from elastic properties. These attributes are assessed for their predictability to predict the target reservoir properties using multi-attribute analysis. For each of the target property neural network is trained with the most predictable attributes, and multi-dimensional, non-linear neural network models are created using multilayered feed forward neural network (MLFN), followed by Probabilistic neural network (PNN). The specific neural network models for each target property are employed for quantitative estimate of volume of clay, effective-porosity, hydrocarbon-saturation in inter-well regions. The estimated properties leverage the identification of untapped oil reserves and provide promising opportunity for enhanced production through drilling of infill wells.

Uncertainty quantification for damage detection in 3D-printed auxetic structures using ultrasonic guided waves and a probabilistic neural network

Auxetic structures hold significant potential for applications due to their outstanding properties. Ultrasonic waves and neural networks are the popular technologies used for structural health monitoring (SHM). To increase the reliability of the neural network output for SHM, comprehensive uncertainty quantification is needed for damage detection in unknown areas of auxetic structures. This paper presents the first comprehensive framework for health diagnosis and uncertainty quantification based on ultrasonic guided waves in two 3D-printed star hourglass honeycomb auxetic structures. The proposed framework integrates in-plane compression with simultaneous ultrasonic testing to receive ultrasonic signals across various deformation states. Additionally, fully elastic and elasto-plastic finite element simulations are conducted to analyze wave energy variations and mechanical responses in auxetic structure. Critical damage deformation is identified based on observed deformation patterns and variations in signal energy. The Hilbert transform is used to extract two damage-sensitive features, namely envelope and phase. These features serve as input data for the Flipout probabilistic convolutional neural network (FPCNN) model, which integrates pseudo-independent weight perturbations and a Gaussian probabilistic layer within the visual geometry group 13 architecture to predict structural deformations and associated uncertainties. The UQ framework effectively separates and quantifies the predictive variance of the FPCNN model into aleatoric and epistemic uncertainty. The framework’s effectiveness is demonstrated through the comprehensive approach, combining compression and ultrasonic tests, finite element simulation, and the FPCNN technique.

Constrained H∞ control with gain switching for a nonlinear active suspension system matching non-pneumatic wheel

In this study, a constrained H ∞ controller with gain switching is put forward for the active suspension system to improve the overall performance of vehicles equipped with non-pneumatic wheels, considering the nonlinearity of wheel stiffness, actuator saturation, and output constraints. Firstly, a quarter-vehicle model incorporating active suspension and non-pneumatic wheel (NPW) is established experimentally. Secondly, the system model is linearized using Taylor series expansion and linear fractional transformation (LFT). A constrained H ∞ control strategy and a gain switching method based on road classification are proposed, taking into account the parameter uncertainty in linearization process and the variation of performance demands under different road conditions. Then, an L ∞ state observer is designed for the required system state, and the road roughness classifier based on grey wolf optimization (GWO) and probabilistic neural network (PNN) is developed to obtain the necessary road information. A bench test is finally performed using the reconstructed actual road as input. The test results validate the effectiveness and superiority of the proposed control strategy.

A dual-stage coverage path planning method for bathymetric survey using an AUV in graph-based SLAM framework considering positioning uncertainty

Autonomous underwater vehicles (AUVs) are becoming increasingly popular for seabed coverage survey tasks but suffer from a serious problem of positioning error accumulation, which leads to low path tracking accuracy and a low coverage rate of the target area. A dual-stage AUV path planning (DSPP) approach was proposed within a simultaneous localization and mapping (SLAM) framework for coverage bathymetric surveys. In the first stage, a global coverage path is designed in a lawnmower style. Crossed paths are added to generate periodic loop closures in SLAM, which can improve the efficiency of online replanning. In the second stage, with a probabilistic neural network (PNN) used for feature-rich bathymetric sub-map identification, an entropy-based online path replanning approach was implemented using 20 templates to adjust the global path to produce accurate loop closures for positioning error correction. Two bathymetric datasets were used to test the proposed DSPP method using a self-developed experimental simulation system. Experiments verified that DSPP can significantly improve the positioning accuracy and coverage rate with acceptably longer paths. In a 6000 m× 5000 marea,the DSPP approach reduced the positioning error by 76% compared with the pure dead reckoning method, and the coverage ratio reached 98.15%.