Generalized Regression Neural Network Model Research Articles

Machine learning in additive manufacturing——NiTi alloy’s transformation behavior

The laser powder bed fused NiTi alloys (LPBF-NiTi) demonstrate shape memory functionality and superelasticity as a result of their distinctive phase transition characteristics. Nevertheless, achieving precise control and regulation of the phase transition temperature poses a challenge, influenced by factors like powder composition and process parameter. In this study, a feature screening strategy and machine learning (ML) method were employed to predict and regulate the phase transition temperature of LPBF-NiTi alloy, offering a more efficient and cost-effective alternative to traditional experimental methods of regulation and control. Specifically, accuracy analysis was performed on LPBF-NiTi phase transition datasets with varying compositions and process conditions utilizing generalized regression neural network (GRNN), and other methods. The findings indicate that the interpretable features extracted through the selection strategy outlined in this study when combined with the GRNN model, demonstrate a high level of prediction accuracy (R2 = 0.97). To investigate the accuracy of the model, this study utilized various process parameters to fabricate NiTi alloys with different compositions from Ni50.8Ti49.2 alloy powder. Using this model, the study identified a novel, larger window of optimal LPBF processing that allows for controllable complex phase transitions.

Predicting chlorophyll-a dynamics in the Ashtamudi estuary using ANN and ANFIS techniques

Excessive nutrients and associated high chlorophyll-a concentration are of growing concern to the public health due to the occurrence of eutrophication in aquatic bodies. Continuous monitoring of chlorophyll-a has limitations, thereby necessitating the development of prediction models. Even though chlorophyll-a prediction models are numerous, they rarely fit into estuarine conditions, which is encountered with dynamic mixing of freshwater and sea water. Thus, the present study identifies the most appropriate driving factors that are relevant for estuarine conditions through chlorophyll-a prediction models developed using Artificial Neural Network and Adaptive Neuro Fuzzy Inference System (ANFIS) approaches in the Ashtamudi estuary, India. Salinity, which is an indirect measure of the estuarine mixing intensity, has been used to replicate the dynamic mixing in the estuary. The results indicate that the model incorporating salinity along with total nitrogen, total phosphorous and turbidity well-predicted chlorophyll-a concentration with a coefficient of determination ranging from 0.73 to 0.99. General Regression Neural Network (GRNN) and ANFIS models outperformed Back Propagation Neural Network (BPNN) and Recurrent Neural Network (RNN) models. Of the various ANFIS models, the ones that used Gaussian and Generalized Bell membership functions were most appropriate for the prediction of chlorophyll-a than the models with triangular and trapezoidal membership functions. The study identified that the models that considered salinity were less sensitive and can be used for modelling chlorophyll-a. The chlorophyll-a was observed to have a direct influence of salinity at salinity values greater than 5‰. The study highlighted that estuarine mixing further enhances the primary productivity through increased penetration of light leading to higher chlorophyll-a concentration. Further the study emplasized that the predictive models are important tools fitting well within estuarine environments due to its replicability.

Sovereign CDS Spread and Term Structure Forecasting Based on Neural Network

The article aims to forecast credit risk for BRICS countries using daily credit default swaps (CDS) spreads obtained from Datastream data base from 2018 to 2023. Our approach consists, first, of forecasting the CDS spread in order to estimate the forecasted CDS term structure. The general regression neural network (GRNN) is used to predict the CDS spread. By checking the accuracy of the prediction, the results show that the GRNN model can be recommended as an effective forecasting tool for CDS spread. Second, the predicted spreads are used to estimate the forecasted CDS term structure using the Nelson–Siegel model. The results show that for Russia, overall, the CDS spreads in the long term are less than those in the short term, which implies that the future outlook is more optimistic, given the events that occurred during the study period, but it still retains the highest level of credit risk compared to other countries. Unlike Brazil, India and South Africa, the future outlook is more pessimistic. For China, the term structure is unstable; in the short term, there is a tendency to reduce the risk, but for longer horizons, the risk will increase. Thus, BRICS countries have different risk profiles depending on investment horizons. The study’s findings help policymakers in developing tailored risk management strategies for BRICS countries and guide investors in making informed credit investment decisions. The use of advanced forecasting tools like GRNN and Nelson–Siegel models emphasizes the importance of sophisticated techniques in enhancing financial market resilience.

A Comparative Study of Baby Boomers and Gen Z on Virtual Reality Adoption for Travel Intentions: A PLS-MGA and GRNN Model

This research delves into the adoption of Virtual Reality (VR) technology for travel intentions, with a specific emphasis on understanding generational disparities between Baby Boomers and Gen Z. This study focuses on two pivotal elements necessitating further examination: the role of virtual experiences and content quality in shaping travel intentions within the realm of virtual reality. Our investigation delves into the influence of perceived ease of use and perceived usefulness on virtual experiences and the impact of information access and novelty on content quality with travel intention for VR. Utilizing the Partial Least Squares Multigroup Analysis (PLS-MGA) and Generalized Regression Neural Network (GRNN) models, we scrutinize these factors. This study aims to illuminate the technological and generational dimensions of VR adoption within the context of travel intention. The study underscores generational distinctions, yielding an intriguing outcome: virtual experiences distinctly shape the travel intentions of Gen Z but do not find favor among Baby Boomers. Conversely, the influence of content quality is evident, decisively affirming the travel intentions of Baby Boomers while lacking substantiation in the case of Gen Z. This finding provides valuable insights into how different generations engage with and adopt VR technology.

Machine learning models for daily net radiation prediction across different climatic zones of China

Net radiation (Rn), a critical component in land surface energy cycling, is calculated as the difference between net shortwave radiation and longwave radiation at the Earth’s surface and holds significant importance in crop models for precision agriculture management. In this study, we examined the performance of four machine learning models, including extreme learning machine (ELM), hybrid artificial neural networks with genetic algorithm models (GANN), generalized regression neural networks (GRNN), and random forests (RF), in estimating daily Rn at four representative sites across different climatic zones of China. The input variables included common meteorological factors such as minimum and maximum temperature, relative humidity, sunshine duration, and shortwave solar radiation. Model performance was assessed and compared using statistical parameters such as the correlation coefficient (R2), root mean square errors (RMSE), mean absolute errors (MAE), and Nash–Sutcliffe coefficient (NS). The results indicated that all models slightly underestimated actual Rn, with linear regression slopes ranging from 0.810 to 0.870 across different zones. The estimated Rn was found to be comparable to observed values in terms of data distribution characteristics. Among the models, the ELM and GANN demonstrated higher consistency with observed values, exhibiting R2 values ranging from 0.838 to 0.963 and 0.836 to 0.963, respectively, across varying climatic zones. These values surpassed those of the RF (0.809–0.959) and GRNN (0.812–0.949) models. Additionally, the ELM and GANN models showed smaller simulation errors in terms of RMSE, MAE, and NS across the four climatic zones compared to the RF and GRNN models. Overall, the ELM and GANN models outperformed the RF and GRNN models. Notably, the ELM model's faster computational speed makes it a strong recommendation for Rn estimates across different climatic zones of China.

Regression Quantitative Structure-toxicity Relationship of Pesticides on Fishes

Pesticide usage reaches several million metric tons annually worldwide, and the effects of pesticides on non-target species, such as various fishes in aquatic environments, have resulted in serious concerns. Predicting pesticide aquatic toxicity to fish is of great significance. In this paper, 20 molecular descriptors were successfully used to develop a regression quantitative structure-activity/toxicity relationship (QSAR/QSTR) model for the toxicity logLC50 of a large data set consisting of 1106 pesticides on fishes by using a general regression neural network (GRNN) algorithm. The optimal GRNN model produced correlation coefficients R of 0.8901 (rms = 0.6910) for the training set, 0.8531 (rms = 0.7486) for the validation set, and 0.8802 (rms = 0.6903) for the test set, which are satisfactory compared with other models in the literature, although a large data set of toxicity logLC50 was used in this work.

Comparative Analysis of Time Series Forecasting using ARIMA, and GRNNs Models: A Case Study of Death Rate of Diabetic Mellitus in Canada

This research aims to compare ARIMA and GRNN models alone. For this comparison the death rate for diabetes mellitus time series data of Canada is used. Autoregressive Integrated Moving Average (ARIMA), and Generalized Regression Neural Networks (GRNN) models were applied for time series prediction of the death rate for diabetes mellitus—trained data for two models from 2000 to 2015. Test data was used to compare the precision through data from 2016 to 2021. The ARIMA model was applied using the auto-command through R package which provided the least BIC and AIC values. The mean absolute deviation (MAD), root mean squared error (RMSE), and mean absolute percentage error (MAPE) were employed to measure the forecasting efficiency of the models. The ARIMA model had the highest prediction accuracy as compared to the GRNN model. ARIMA predicts the death rate for diabetes mellitus more accurately and robustly compared to the GRNNs model.

The prediction of sound absorption coefficient of film multi-cavity materials based on generalized regression neural network (GRNN)

In recent years, noise problems have attracted more and more widespread attention with the rapid development of economy, social life and transportation construction. Sound can cause the membrane to vibrate, thereby absorbing sound waves of specific frequencies. In this study, the film was transformed into a bubble structure to form a film multi-cavity structure material. The theoretical calculation model of the sound absorption performance of sound-absorbing materials has always been the focus of research. Generalized regression neural network (GRNN) is a radial basis neural network (RBF) with a flexible network structure, high fault tolerance and strong nonlinear mapping capabilities. This paper uses the generalized regression neural network (GRNN) method to predict the sound absorption coefficient of film multi-cavity materials. Twenty-four sets of film multi-cavity material samples are prepared, and two groups of GRNN model is established. Each group of GRNN model uses 12 sets of data, of which 10 sets are used as training samples for GRNN and 2 sets are used as test samples. The thickness, bubble diameter and porosity of the material are used as inputs of the GRNN, and the sound absorption coefficients at 1/3 octave center frequencies are used as the output of the network. The optimal spread factors obtained by training the two groups of GRNN models are 0.5 and 0.35 respectively. The results show that the minimum error between the model prediction value and the experimental measurement value is 0.001. The established GRNN-1 and GRNN-2 models have high accuracy and high similarity in sound absorption curves. This method is simple, effective and accurate. When there are fewer data samples, GRNN also has outstanding performance in terms of approximation ability and learning speed.

Assessment of Peanut Protein Powder Quality by Near-Infrared Spectroscopy and Generalized Regression Neural Network-Based Approach.

The global demand for protein is on an upward trajectory, and peanut protein powder has emerged as a significant player, owing to its affordability and high quality, with great future market potential. However, the industry currently lacks efficient methods for rapid quality testing. This research paper addressed this gap by introducing a portable device with employed near-infrared spectroscopy (NIR) to quickly assess the quality of peanut protein powder. The principal component analysis (PCA), partial least squares (PLS), and generalized regression neural network (GRNN) methods were used to construct the model to further enhance the accuracy and efficiency of the device. The results demonstrated that the newly established NIR method with PLS and GRNN analysis simultaneously predicted the fat, protein, and moisture of peanut protein powder. The GRNN model showed better predictive performance than the PLS model, the correlation coefficient in calibration (Rcal) of the fat, the protein, and the moisture of peanut protein powder were 0.995, 0.990, and 0.990, respectively, and the residual prediction deviation (RPD) were 10.82, 10.03, and 8.41, respectively. The findings unveiled that the portable NIR spectroscopic equipment combined with the GRNN method achieved rapid quantitative analysis of peanut protein powder. This advancement holds a significant application of this device for the industry, potentially revolutionizing quality testing procedures and ensuring the consistent delivery of high-quality products to fulfil consumer desires.

Forecasting regional water demand using multi-fidelity data and harris hawks optimization of generalized regression neural network models – A case study of Heilongjiang Province, China

Accurate forecasting of water demand plays a crucial role in decision makers to effectively allocate regional water resources and promotes the rational development and utilization of water resources. To address the challenges posed by limited raw data availability and the difficulty in selecting parameters for water demand forecasting models, this study proposes a coupled model called MF-HHO-GRNN (Multi-Fidelity Data, Hybrid Harris Hawks Optimization, and Generalized Regression Neural Network). Initially, multi-fidelity data was constructed employing water demand factor data for Heilongjiang Province, China (1999–2020) as high-fidelity data, alongside data from Liaoning and Jilin Provinces of China as low-fidelity data. Adaptive boosting algorithms, data concatenation, and data dimensionality reduction techniques were employed to construct multi-fidelity data, thereby improving the availability of the original data. Subsequently, Harris hawk optimization algorithms were implemented to optimize the smoothing factor of the GRNN to resolve the challenge of selecting suitable parameters for water demand forecasting models. Finally, the MF-HHO-GRNN model was compared with other prediction models to analyze the prediction accuracy. The optimized MF-HHO-GRNN model meeting accuracy requirements was utilized to forecast agricultural irrigation, industrial, domestic, ecological water demand, and total water demand in Heilongjiang Province for the next 5 years. Results demonstrated that the proposed MF-HHO-GRNN model attained predictive accuracy of 98 %, 98 %, and 97 % for total water demand under three distinct time horizons (1 year, 3 years, and 5 years), proving excellent predictive capabilities. Furthermore, a comprehensive comparison was made between the predicted total water demand and user-specific water demands for the next 5 years in Heilongjiang Province and the regional development plans. The results showed that the predictions of the coupled model aligned with the future development trends of Heilongjiang Province, validating the practicality of the MF-HHO-GRNN model. This study provides decision-makers with an effective method for forecasting water demand, contributing to the future realization of more intelligent and accurate water resource management and decision-making.

Precision positioning based on temperature dependence self-sensing magnetostrictive actuation mechanism

The self-sensing actuator that integrates both actuation and sensing functions is an effective way to simplify the structure of electromechanical systems. This paper proposes a temperature-dependent self-sensing actuation strategy based on a self-sensing giant magnetostrictive actuator (SSGMA) by sensing online stiffness. Utilizing the developed model considering temperature-dependent hysteresis nonlinearity, the self-sensing output characteristics of SSGMA are first evaluated. The self-sensing based control system is then designed in detail. The relationship between the self-sensing signal and the output displacement of SSGMA at different temperatures is constructed by General Regression Neural Network (GRNN) model for fast recognition by the controller. On this basis, a composite control scheme that takes into account rate-dependent asymmetric hysteresis nonlinearities and combines an adaptive feedforward controller with PID feedback controller is proposed for precision actuation of SSGMA. Finally, a temperature-controlled experimental platform is assembled for verification. Experimental results demonstrate that, based on the developed model, the SSGMA is capable of accurate self-sensing output at temperatures of 22 °C, 40 °C, and 70 °C, respectively. The SSGMA with the proposed control method enables the synchronous and precise self-sensing positioning of step and multiple-frequency sinusoidal expectations at different temperatures.

Modeling Xanthan Gum Foam's Material Properties Using Machine Learning Methods.

Xanthan gum is commonly used in the pharmaceutical, cosmetic, and food industries. However, there have been no studies on utilizing this natural biopolymer as a foam material in the insulation and packaging sectors, which are large markets, or modeling it using an artificial neural network. In this study, foam material production was carried out in an oven using different ratios of cellulose fiber and xanthan gum in a 5% citric acid medium. As a result of the physical and mechanical experiments conducted, it was determined that xanthan gum had a greater impact on the properties of the foam material than cellulose. The densities of the produced foam materials ranged from 49.42 kg/m3 to 172.2 kg/m3. In addition, the compressive and flexural moduli were found to vary between 235.25 KPa and 1257.52 KPa and between 1939.76 KPa and 12,736.39 KPa, respectively. Five machine-learning-based methods (multiple linear regression, support vector machines, artificial neural networks, least squares methods, and generalized regression neural networks) were utilized to analyze the effects of the components used in the foam formulation. These models yielded accurate results without time, material, or cost losses, making the process more efficient. The models predicted the best results for density, compression modulus, and flexural modulus achieved in the experimental tests. The generalized regression neural network model yielded impressive results, with R2 values above 0.97, enabling the acquisition of more quantitative data with fewer experimental results.

Building models of technological processes based on neuro-fuzzy technology

The work considers the issues of formalization of the extraction process in the form of a generalized regression neural network model, which are the basis for solving the problem of analysis and synthesis of the extraction process control system for obtaining petroleum products. An adaptive learning algorithm for a neural network model has been developed that is characterized by high speed and accuracy. A comparative analysis of the developed model with existing ones was made, which showed the effectiveness of the proposed algorithm for building the architecture of neural network models and learning the weight coefficients of the model.

Predicting and optimizing reactive oxygen species metabolism in Punica granatum L. through machine learning: role of exogenous GABA on antioxidant enzyme activity under drought and salinity stress

BackgroundDrought and salinity stress have been proposed as the main environmental factors threatening food security, as they adversely affect crops' agricultural productivity. As a potential solution, the application of plant growth regulators to enhance drought and salinity tolerance has gained considerable attention. γ-aminobutyric acid (GABA) is a four-carbon non-protein amino acid that accumulates in plants as a response to stressful conditions. This study focused on a comparative assessment of several machine learning (ML) regression models, including radial basis function, generalized regression neural network (GRNN), random forest (RF), and support vector regression (SVR) to develop predictive models for assessing the effect of different concentrations of GABA (0, 10, 20, and 40 mM) on various physio-biochemical traits during periods of drought, salinity, and combined stress conditions. The physio-biochemical traits included antioxidant enzyme activities (superoxide dismutase, SOD; peroxidase, POD; catalase, CAT; and ascorbate peroxidase, APX), protein content, malondialdehyde (MDA) levels, and hydrogen peroxide (H2O2) levels. The non‑dominated sorting genetic algorithm‑II (NSGA‑II) was employed for optimizing the superior prediction model.ResultsThe GRNN model outperformed the other ML algorithms and was therefore selected for optimization by NSGA-II. The GRNN-NSGA-II model revealed that treatment with GABA at concentrations of 20.90 mM and 20.54 mM, under combined drought and salinity stress conditions at 20.86 and 20.72 days post-treatment, respectively, could result in the maximum values for protein content (by 0.80 and 0.69), APX activity (by 50.63 and 51.51), SOD activity (by 0.54 and 0.53), POD activity (by 1.53 and 1.72), CAT activity (by 4.42 and 5.66), as well as lower MDA levels (by 0.12 and 0.15) and H2O2 levels (by 0.44 and 0.55), respectively, in the ‘Atabaki’ and ‘Rabab’ cultivars.ConclusionsThis study demonstrates that the GRNN-NSGA-II model, as an advanced ML algorithm with a strong predictive ability for outcomes in combined stressful environmental conditions, provides valuable insights into the significant factors influencing such multifactorial processes.

Time-series prediction of settlement deformation in shallow buried tunnels based on EMD-SSA-GRNN model

Tunnel settlement deformation monitoring is a complex task and can result in nonlinear dynamic changes. To overcome the disturbances caused by historical data and the difficulty in selecting input parameters during deformation prediction, a decomposition, reconstruction and optimization method for tunnel settlement deformation prediction is proposed. First, empirical mode decomposition (EMD) is used to decompose the in-situ monitoring data and reduce the interactions among information at different scales in sequences. Then, the monitoring data are decomposed into intrinsic mode functions (IMFs). Secondly, the smoothing factor of the generalized regression neural network (GRNN) is optimized by using the sparse search algorithm (SSA). An EMD-SSA-GRNN deformation prediction model is developed using the optimized GRNN algorithm and is used to predict the changes in the decomposed IMFs. Finally, using the measured deformation data from a shallowly buried tunnel along the Kaizhou-Yunyang Highway in Chongqing, China, the reliability and accuracy of different models are analysed. The results show that tunnel settlement deformation exhibited a trend and a slow change in the early stage, a rapid change in the middle stage and a slow change in the late stage, and the rate of change was significantly influenced by the excavation time and the upper and lower geological layers. The prediction accuracy of the EMD-SSA-GRNN model after EMD improved from 19.2 to 59.4% relative to that of the SSA-GRNN and single GRNN models. Moreover, we find that the three error evaluation indicators of the EMD-SSA-GRNN model are lower than those of the other models and that the results of the proposed model and are more strongly correlated with measured data.

Research on the Innovation Path and Enterprise Adaptation of Green Finance in the Development of Low-Carbon Economy-Strategy Analysis Based on Neural Network Modeling

Abstract Green finance is an economic form formed by the organic combination of finance and environmental awareness, and the long-term development of a low-carbon economy needs to be strongly supported by green finance. Based on the generalized regression neural network model, this paper measures the level of green finance development in a low-carbon economy by constructing a green finance evaluation system. The evaluation system for enterprise adaptability level under green finance is constructed from three dimensions: enterprise green development ability, enterprise green innovation ability, and enterprise green management ability. The entropy weight method is used to empower the indicators in the rating system, and the set-pair theory is used to portray the data of the empowered indicators, while the Markov chain combined with the set-pair linkage is introduced to portray the dynamic trend of the development of enterprise adaptability level. The generalized regression neural network model’s prediction accuracy for green finance development in Guangdong Province is above 85% after the study. Year-on-year, the level of green finance development in Guangdong Province increased by around 57.78%, creating a significant growth trend in the context of a low-carbon economy between 2010 and 2017. The level of business adaptability also leaps from “poor” to “excellent”, with the linkage value changing from −0.8574 in 2018 to 0.8031 in 2022. This paper’s research provides the perfect data support for choosing a green financial innovation path only in a low carbon economy.

A Generalized Regression Neural Network Model for Accuracy Improvement of Global Precipitation Products: A Climate Zone-Based Local Optimization

A Generalized Regression Neural Network Model for Accuracy Improvement of Global Precipitation Products: A Climate Zone-Based Local Optimization

Predição do conteúdo relativo de clorofila na copa da planta de chá usando algoritmo GRNN otimizado e imagens multiespectrais RPA

ABSTRACT To quickly and accurately assess tea plant growth, this study aims to find a new way to predict the chlorophyll content in tea plant canopies using machine learning. Using remotely piloted aircraft equipped with multispectral cameras, images of tea plantation areas are captured and reflectance from four spectral bands is extracted, leading to the calculation of vegetation indices. Simultaneously, chlorophyll relative content in the tea plant canopies was collected on the ground using a detector. Four models, namely Random Forest (RF), Backpropagation neural network (BPNN), Radial basis function network (RBFN), and General Regression Neural Network (GRNN), were constructed to predict the chlorophyll relative content in tea plant canopies. Subsequently, important remote sensing variables were identified through RF filtering, followed by a comparison of the predictive performance of machine learning models under different input conditions. Lastly, by integrating the Sparrow Search Algorithm (SSA) to optimize the smoothing factor in the GRNN, the study explores the impact of optimization algorithms on the predictive performance of the GRNN model. Experiments indicate that within the established machine learning models, the GRNN demonstrates the highest predictive accuracy. By ranking the importance of remote sensing variables through RF, 18 significant remote sensing variables were selected, which enhanced the predictive accuracy of the machine learning models. The optimization of the GRNN smoothing factor through the SSA algorithm can significantly enhance the predictive accuracy of the GRNN model. Based on a series of experiments, the established RFSSA-GRNN prediction model demonstrates good predictive performance, with an reaching 0.84.

Evaluating the efficacy of recyclable nanostructured adsorbents for rapid removal of methylparaben from aqueous solutions

In this study, we evaluate the efficiency of two novel nanostructured adsorbents – chitosan-graphitic carbon nitride@magnetite (CS-g-CN@Fe3O4) and graphitic carbon nitride@copper/zinc nanocomposite (g-CN@Cu/Zn NC) – for the rapid removal of methylparaben (MPB) from water. Our characterization methods, aimed at understanding the adsorbents’ structures and surface areas, informed our systematic examination of influential parameters including sonication time, adsorbent dosage, initial MPB concentration, and temperature.We applied advanced modeling techniques, such as response surface methodology (RSM), generalized regression neural network (GRNN), and radial basis function neural network (RBFNN), to evaluate the adsorption process. The adsorbents proved highly effective, achieving maximum adsorption capacities of 255 mg g−1 for CS-g-CN@Fe3O4 and 218 mg g−1 for g-CN@Cu/Zn NC. Through genetic algorithm (GA) optimization, we identified the optimal conditions for the highest MPB removal efficiency: a sonication period of 12.00 min and an adsorbent dose of 0.010 g for CS-g-CN@Fe3O4 NC, with an MPB concentration of 17.20 mg L−1 at 42.85 °C; and a sonication time of 10.25 min and a 0.011 g dose for g-CN@Cu/Zn NC, with an MPB concentration of 13.45 mg L−1 at 36.50 °C.The predictive accuracy of the RBFNN and GRNN models was confirmed to be satisfactory. Our findings demonstrate the significant capabilities of these synthesized adsorbents in effectively removing MPB from water, paving the way for optimized applications in water purification.

Prediction of NOx emission from two-stage combustion of NH3–H2 mixtures under various conditions using artificial neural networks

This study aims to predict NOx emission from the combustion of ammonia (NH3) and hydrogen (H2) mixtures using Artificial Neural Network (ANN) models. Chemical Reactor Network (CRN) models were utilized to calculate the NOx emission during the Rich-Quench-Lean (RQL) two-stage combustion of NH3–H2 mixtures under various conditions, such as fuel pressures, inlet temperatures, primary air humidification fractions, total equivalent ratios and primary equivalent ratios. The simulation results served as a data set for training and validation of three different ANN models: Backpropagation (BP), Radial Basis Function (RBF), and Generalized Regression Neural Network (GRNN). The results show that all three ANN models can effectively replace the CRN model in accurately predicting NOx concentrations. In particular, the BP model, optimized with 37 hidden neurons, showcased superior performance, achieving a correlation coefficient (R) value of 0.99826 with the simulated results. It had the lowest Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) values of 61.3 and 96.1, respectively. In addition, it achieved the highest coefficient of determination (R2) and Model Efficiency (ME) values of 0.9950 and 0.9908 respectively, surpassing the RBF and GRNN models. These results suggest that the BP model has significant potential as an alternative to the CRN model for predicting NOx emissions from two-stage combustion of NH3–H2 mixtures, demonstrating both accuracy and efficiency.