A New Levenberg Marquardt based Back Propagation Algorithm Trained with Cuckoo Search

Unfolding of mono-energy neutron spectra using artificial neural network based on LMBP training algorithm

In this work, neutron spectra are unfolded using artificial neural networks (ANNs). The neutron response of the NE213 scintillator detector, characterized by the pulse height distribution, is calculated to obtain the necessary data for unfolding the energy spectrum. This is achieved using both analytical response functions and response functions generated by the MCNPX-PHOTRACK code. In this query, the Levenberg-Marquardt method (LMM), which has a high computational speed in the learning method, is used to train the network. The performance of the ANN for unfolding the neutron energy spectrum of the NE213 scintillation detector was evaluated by comparing its results to the established Gravel method. The ANN method consistently produced spectra with a single peak closely matching the incident energy, while the Gravel method showed additional peaks and distortions. Quantitative analysis revealed a lower relative energy peak difference (indicating higher accuracy) for the ANN method compared to Gravel, particularly when noise was introduced into the data. These results suggest that ANNs offer a more robust and accurate approach for neutron spectrum unfolding.

Multiple faults diagnosis for ocean-going marine diesel engines based on different neural network algorithms

Fault diagnosis technologies for ocean-going marine diesel engines play an important role in the safety and reliability of ship navigation. Although many fault diagnosis technologies have achieved acceptable results for single fault of diesel engines, the diagnosis of multiple faults is rarely involved. Due to the strong correlation, non-linearity and randomness of multiple faults, it is extremely difficult to make an accurate diagnosis. In this study, diagnosis methods based on thermal parametric analysis combined with different neural network algorithms were established and used for the diagnosis of multiple faults in the ocean-going marine diesel engine. The results show that the Levenberg Marquardt back propagation neural network has the highest diagnostic accuracy rate of 88.89% and 100% for multiple faults and single faults, respectively, and its diagnostic time is also relatively short, 0.78 s. The Bayesian regularization back propagation neural network can give a diagnostic accuracy rate of 100% for single faults, but for multiple faults, the diagnostic accuracy rate is only 55.56%, and the diagnosis time for the entire sample is the longest. As for the probabilistic neural network, although it has the fastest diagnosis speed, it has the lowest diagnostic accuracy rate for both single faults and multiple faults. The results may provide references for the online diagnosis of single faults and multiple faults in ocean-going marine diesel engines.

Optimization-based artificial neural networks to fit the isotherm models parameters of aqueous-phase adsorption systems

An artificial neural network (ANN) hybrid structure was proposed that, unlike the standard ANN structure optimization, allows the fit of several adsorption curves simultaneously by indirectly minimizing the real output error. To model a case study of 3-aminophenol adsorption phenomena onto avocado seed activated carbon, a hybrid ANN was applied to fit the parameters of the Langmuir and Sips isotherm models. Network weights and biases were optimized with two different methods: particle swarm optimization (PSO) and genetic algorithm (GA), due to their good convergence in large-scale problems. In addition, the data were also fitted with the Levenberg-Marquardt feedforward optimization method to compare the performance between a standard ANN model and the hybrid model proposed. Results showed that the ANN-isotherm hybrid models with both PSO and GA were able to accurately fit the experimental equilibrium adsorption capacity data using the Sips isotherm model, obtaining Pearson's correlation coefficient (R) of the order of 0.9999 and mean squared error (MSE) around 0.5, very similar to the performance of standard ANN using Levenberg-Marquardt optimization. On the other hand, the results with Langmuir isotherm models were quite inferior in the ANN-isotherm hybrid models with both PSO and GA, with R and MSE of around 0.944 and 4.04 × 10², respectively. The proposed ANN-isotherm hybrid structure was successfully applied to estimate the parameters of adsorption isotherms, reducing the computational demand and the exhausting task of estimating the parameters of each adsorption curve individually.

An Evolutionary Field Theorem: Evolutionary Field Optimization in Training of Power-Weighted Multiplicative Neurons for Nitrogen Oxides-Sensitive Electronic Nose Applications

Neuroevolutionary machine learning is an emerging topic in the evolutionary computation field and enables practical modeling solutions for data-driven engineering applications. Contributions of this study to the neuroevolutionary machine learning area are twofold: firstly, this study presents an evolutionary field theorem of search agents and suggests an algorithm for Evolutionary Field Optimization with Geometric Strategies (EFO-GS) on the basis of the evolutionary field theorem. The proposed EFO-GS algorithm benefits from a field-adapted differential crossover mechanism, a field-aware metamutation process to improve the evolutionary search quality. Secondly, the multiplicative neuron model is modified to develop Power-Weighted Multiplicative (PWM) neural models. The modified PWM neuron model involves the power-weighted multiplicative units similar to dendritic branches of biological neurons, and this neuron model can better represent polynomial nonlinearity and they can operate in the real-valued neuron mode, complex-valued neuron mode, and the mixed-mode. In this study, the EFO-GS algorithm is used for the training of the PWM neuron models to perform an efficient neuroevolutionary computation. Authors implement the proposed PWM neural processing with the EFO-GS in an electronic nose application to accurately estimate Nitrogen Oxides (NO_x) pollutant concentrations from low-cost multi-sensor array measurements and demonstrate improvements in estimation performance.

Membership Function-Weighted Non-Linear Fitting Method for Optical-Sensing Modeling and Reconstruction

Imprecise measurements present universally due to variability in the measurement error. We devised a very simple membership function to evaluate fuzzily the quality of optical sensing with a small dataset, where a normal distribution cannot be assumed. The proposed membership function was further used as a weighting function for non-linear curve fitting under expected mathematical model constraints, namely the membership function-weighted Levenberg⁻Marquardt (MFW-LM) algorithm. The robustness and effectiveness of the MFW-LM algorithm were demonstrated by an optical-sensing simulation and two practical applications. (1) In laser-absorption spectroscopy, molecular spectral line modeling was greatly improved by the method. The measurement uncertainty of temperature and pressure were reduced dramatically, by 53.3% and 43.5%, respectively, compared with the original method. (2) In imaging, a laser beam-profile reconstruction from heavy distorted observations was improved by the method. As the dynamic range of the infrared camera increased from 256 to 415, the detailed resolution of the laser-beam profiles increased by an amazing 360%, achieving high dynamic-range imaging to capture optical signal details. Therefore, the MFW-LM algorithm provides a robust and effective tool for fitting a proper physical model and precision parameters from low-quality data.