Accuracy Of Artificial Neural Network Model Research Articles

Artificial Neural Network-Based Prediction of Physical and Mechanical Properties of Concrete Containing Glass Aggregates

This comprehensive study analyzes the use of crushed glass as both fine and coarse aggregate in concrete, as well as the prediction accuracy of Artificial Neural Networks (ANN). The primary objectives are to understand the interactions between concrete’s constituents and to assess the accuracy of ANN models in predicting concrete’s mechanical and physical properties. This is achieved using a two-decade experimental results dataset of concrete’s compressive and tensile strengths, slump, density, and the corresponding mix design proportions, including waste glass aggregate. A series of 70 concrete samples were carefully built and tested, with compressive strengths varying from 12 to 71 MPa and glass aggregate percentages ranging from 0-100%. These samples served as the basis for the creation of an input dataset and ANN targets. The ANN model underwent intensive training, validation, testing, and statistical regression analysis. The ANN models are exceptionally accurate, with a continuously low error margin of roughly 2%, highlighting their usefulness in matching experimental and predicted results. Validation techniques highlight the models' dependability, with consistently high coefficients of determination (R-values), including 0.99484, demonstrating their robustness in replicating complicated concrete properties. The data analysis shows a unique pattern, with optimum glass aggregate percentages in the range of 10–20%. Beyond this range, there is a noticeable decline in concrete properties. Finally, the study confirms the efficacy of ANN in predictive modeling while also validating the potential of crushed glass to replace natural aggregates in concrete. Doi: 10.28991/CEJ-2024-010-05-018 Full Text: PDF

Predicting hydrogen segregation energy distributions in strained regions of tungsten using artificial neural network

Employing the smooth overlap of atomic position (SOAP) descriptors, we established an artificial neural network (ANN) model with the ability to effectively and accurately predict the segregation energy Eseg distributions of hydrogen (H) atoms in various strained regions of tungsten (W). The model is verified to have comparable accuracy with the molecular statics (MS) relaxations, as quantitively evidenced by the root mean square error (RMSE) values of 0.018 eV for the homogeneously strained regions, averagely 0.025 eV for regions near different dislocations, and 0.032 eV for regions surrounding an interstitial H cluster. Besides, the predictions yield no systematic bias at both high and low Eseg extremes, which is a common issue in the elastic theory (ET) calculations. Moreover, Eseg values are precisely predicted with the RMSE value of 0.041 eV surrounding an interstitial H cluster that is not encountered during the learning phase of the model, demonstrating the model’s robust generalization capability. Considering its markedly enhanced computation speed, this accurate ANN model holds promising prospects in circumstances that require massive Eseg calculations.

Compact Modeling of Advanced Gate-All-Around Nanosheet FETs Using Artificial Neural Network.

As the architecture of logic devices is evolving towards gate-all-around (GAA) structure, research efforts on advanced transistors are increasingly desired. In order to rapidly perform accurate compact modeling for these ultra-scaled transistors with the capability to cover dimensional variations, neural networks are considered. In this paper, a compact model generation methodology based on artificial neural network (ANN) is developed for GAA nanosheet FETs (NSFETs) at advanced technology nodes. The DC and AC characteristics of GAA NSFETs with various physical gate lengths (Lg), nanosheet widths (Wsh) and thicknesses (Tsh), as well as different gate voltages (Vgs) and drain voltages (Vds) are obtained through TCAD simulations. Subsequently, a high-precision ANN model architecture is evaluated. A systematical study on the impacts of ANN size, activation function, learning rate, and epoch (the times of complete pass through the entire training dataset) on the accuracy of ANN models is conducted, and a shallow neural network configuration for generating optimal ANN models is proposed. The results clearly show that the optimized ANN model can reproduce the DC and AC characteristics of NSFETs very accurately with a fitting error (MSE) of 0.01.

Artificial Neural Networks versus Multiple Linear Regressions to Predict the Christiansen Uniformity Coefficient in Sprinkler Irrigation

The Christiansen Uniformity Coefficient (CUC) describes the distribution of water in a sprinkler system. In this study, two types of models were developed to predict the Christiansen Uniformity Coefficient (CUC) of sprinkler irrigation systems: Artificial Neural Network (ANN), specifically the feed-forward neural networks, and multiple linear regression (MLR) models. The models were trained on a dataset of published research on the CUC of sprinkler irrigation systems, which included data on a variety of design, operating, and meteorological condition variables. In order to build the predictive model of CUC, 10 input parameters were used including sprinkler height (H), working pressure (P), nozzle diameter (D and da), sprinkler line spacing (SL), sprinkler spacing (SS), wind speed (WS), wind direction (WD), temperature (T), and relative humidity (RH). Fifty percent (50%) of the data was used to train ANN models and the remaining data for cross-validation (25%) and for testing (25%). Multiple linear regression models were built using the training data. Four statistical criteria were used to evaluate the model’s predictive quality: the correlation coefficient (R), the index of agreement (d), the root mean square error (RMSE), and the mean absolute error (MAE). Statistical analysis demonstrated that the best predictive ability was obtained when the models (ANN and MLR) utilized all the input variables. The results demonstrated that the accuracy of ANN models, predicting the CUC of sprinkler irrigation systems, is higher than that of the MLR ones. During the training stage, the ANN models were more accurate in predicting CUC than MLR, with higher R (0.999) and d (0.999) values and lower MAE (0.167) and RMSE (0.456) values. The R values of the MLR model fluctuated between 0.226 and 0.960, the d values oscillated from 0.174 to 0.979, the MAE values were in the range of 2.458% and 10.792%, and the RMSE values fluctuated from 2.923% to 13.393%. Furthermore, the study revealed that WS and WD are the most influential climatic parameters. The ANN model can be used to develop more accurate tools for predicting the CUC of sprinkler irrigation systems. This can help farmers to design and operate their irrigation systems more efficiently, which can save them time and money.

Deriving groundwater major ions from electrical conductivity using artificial neural networks supported by analytical hydrochemical solutions

Tracing groundwater chemistry requires broad and continuous sampling campaigns to keep track of groundwater deterioration. The need is pressing in coastal aquifers due to high water demand to meet rapid population growth. This paper uses artificial intelligence (AI) to develop an algorithm capable of predicting seven major chemical ions in water (Ca2+, Mg2+, Na+, K+, HCO3−, Cl− and SO42−) from electrical conductivity (EC), a single input that can be easily acquired manually or via automated monitoring programs. The algorithm mainly relied on artificial neural network (ANN) to tackle complex and nonlinear relationships. It included a set of procedures among others, Levenberg-Marquardt back propagation multilayer perceptron (MLP), conjugate gradient back propagation with Fletcher-Reeves updates MLP, and support vector regression. The low accuracy of ANN models in predicting Ca2+ and HCO3− concentrations justified their calculation via advanced hydrochemical analytical solutions that were embedded in one AI scheme. The full scheme was trained and tested over an EC20 range of 400–3000 μS/cm, and then validated for new measurements elsewhere. It showed <15% discrepancy for the four ions that constitute the majority of groundwater chemical composition (Cl−, Na+, HCO3− and Ca2+). Higher uncertainties were associated with concentrations derived for EC values outside the training limits. The developed algorithm can save time and money in return of labor and laboratory work. It is an open source code publicly available for further improvements by training it on additional data covering a wider EC range.

A data-driven model of drop size prediction based on artificial neural networks using small-scale data sets

An artificial neural network (ANN) method is introduced to predict drop size in two kinds of pulsed columns with small-scale data sets. After training, the deviation between calculate and experimental results are 3.8% and 9.3%, respectively. Through ANN model, the influence of interfacial tension and pulsation intensity on the droplet diameter has been developed. Droplet size gradually increases with the increase of interfacial tension, and decreases with the increase of pulse intensity. It can be seen that the accuracy of ANN model in predicting droplet size outside the training set range is reach the same level as the accuracy of correlation obtained based on experiments within this range. For two kinds of columns, the drop size prediction deviations of ANN model are 9.6% and 18.5% and the deviations in correlations are 11% and 15%.

A novel optimization method for transcritical-subcritical parallel organic Rankine cycle under variable engine loads and cooling water temperature

Organic Rankine cycle (ORC) is an effective technology for engine waste heat recovery. However, when engine loads and cooling water temperature are variable, the optimization of ORC system will be difficult and time-consuming, especially for complex system like transcritical two-stage ORC. In this study, a novel optimization method is proposed for optimizing the transcritical-subcritical parallel ORC (TSPORC) system, and the system operates under variable engine loads and cooling water temperature. The key idea of the optimization method is a combination of artificial neural network (ANN) and non-dominated sorting genetic algorithm (NSGA-II), the ANN is used to establish a fast and accurate prediction model of the system performance to reduce computing time significantly, and the NSGA-II is furtherly used to optimize system performance fast and accurately based on the ANN prediction model. To improve the accuracy of ANN model, three different artificial neural network models are optimized and compared. To satisfy different objective optimization for application, both single-objective and multi-objective optimization are conducted. The results demonstrate that the optimization method can achieve real time optimization for TSPORC system operating under variable engine loads and cooling water temperature fast and accurately.

Prediction of control temperature and emergency temperature of monadic/binary aromatic nitro compounds by quantitative structure-property relationship: correlation study of self-accelerating decomposition temperature in thermal hazard assessment.

The thermal hazard of reactions caused by the structural instability of aromatic nitro compounds is a major concern in the field of chemical process safety and one of the main causes of major thermal runaway (TR) accidents such as fire and explosion. Among them, the self-accelerating decomposition temperature (SADT), as an important parameter, has been widely used to evaluate the thermal hazards of aromatic nitro compounds in actual storage and transportation processes. However, the control temperature (CT) and emergency temperature (ET), which depend on and are associated with SADT, have been rarely reported in previous studies. In this work, multiple linear regression (MLR) and artificial neural network (ANN) models for CT and ET were constructed based on the molecular descriptors corresponding to the stable structures of 27 monadic/binary aromatic nitro compounds, combined with advanced adiabatic accelerating calorimetric experiments and quantitative structure-property relationship (QSPR). The optimal subset of descriptors with significant contributions was screened out while the fit, predictive ability, and robustness of the four types of models were evaluated with internal and external validation parameters, and finally, two types of parameters (R2 and ARE) were selected as the main indicators for a comprehensive comparative analysis. The results show that the four models fit the experimental data well. During this period, the accuracy of ANN models is slightly higher than that of MLR models, and the QSPR models under the two modes (linear and nonlinear) are more inclined toward ET in prediction ability. Based on simplifying the calculation process and realizing rapid parameter prediction, this study is expected to provide technical support for engineering applications such as safe operation, safe storage and transportation of substances, and emergency response in the chemical industry. In this work, we tested and calculated the thermal safety parameters of 27 monadic/binary aromatic nitro compounds by ARC and AKTS and further used the PubChem database and Gaussian 09 software program to obtain and optimize their corresponding molecular structures. The geometric optimization process adopts DFT on the B3LYP level and the 6-31 + G(d, p) basis set, while the same functional and basis set was used for vibration analysis. The OpenBabel toolbox and ChemDES platform were used for transformation coding and descriptor calculation. Finally, IBM SPSS Statistics 24 and MATLAB software were used to construct MLR models and ANN models, respectively.

Instantaneous Frequency and Chirp Rate Estimation for Noisy Quadratic FM Signals by CNN

Deep learning and machine learning are widely employed in various domains. In this paper, Artificial Neural Network (ANN) and Convolution Neural Network (CNN) are used to estimate the Instantons Frequency (IF), Linear Chirp Rate (LCR), and Quadratic Chirp Rate (QCR) for Quadratic Frequency Modulated (QFM) signals under Additive White Gaussian (AWG) noise and Additive Symmetric alpha Stable (ASαS) noise. SαS distributions are impulsive noise disturbances except for a few circumstances, lack a closed-form Probability Density Function (PDF), and an infinite second-order statistic. Geometric SNR (GSNR) is used to determine the impulsiveness of mixture noise for Gaussian and SαS noise. ANN is a machine learning classifier with few layers that reduce FE, LCRE, and QCRE complexity and achieve high accuracy. CNN is a deep learning classifier that is built with multiple layers of FE, LCRE, and QCRE. CNN is more accurate than ANN when dealing with large amounts of data and determining optimal features. The results reveal that SαS noise is substantially more damaging to FE, LCRE, and QCRE than Gaussian noise, even when the magnitude is modest, and it is less damaging when alpha is greater than one. After training DCNN for FE, LCRE, and QCRE estimation of QFM signals. The 2D-CNN model accuracy achieved 98.7603 and 1D-CNN is 75.8678 for ten epochs. ANN model accuracy achieved 37.5 for 1000 epochs. The accuracy of TFD (spectrogram &amp; pspectrum) for frequency estimation of QFM signals was 38.4254 by spectrogram and 38.6746 by pspectrum.

The role of neural network for estimating real estate prices value in post COVID-19: a case of the middle east market

The main goal of this paper was to explore the use of an artificial neural network (ANN) model in predicting real estate prices in the Middle East market. Although conventional modeling approaches such as regression can be used in prediction, they have a weakness of a predetermined relationship between input and output. In this regard, using the ANN model was expected to reduce the bias and ensure non-linear relationships are also covered in the prediction process for more accurate results. The ANN model was created using Python v.3.10 program. The model exhibited a high correlation between predicted and actual house price data (R=0.658). In this respect, it was realized that the model could be effectively used in appraising real estate by investors. However, a major limitation of the model was realized to be a limited dataset for large and luxurious houses, which were not accurately predicted as data distribution between actual and predicted values became sparse for high house prices. A key recommendation made is that future research should include more variables related to luxurious houses and macroeconomic factors to increase the ANN model accuracy.

Forecasting the Final Contract Cost on the Basis of the Owner’s Cost Estimation Using an Artificial Neural Network

Raising the final contract cost (FCC) is a significant risk for project owners. This study hypothesizes that the factors that cause owner’s cost estimation (OCE) accuracy and FCC changes share the same causes, and a case study confirmed that the two variables (OCE and FCC) could be correlated. Accordingly, this study aims to develop a forecast model to predict FCC on the basis of the initial OCE, which has not been studied previously. This study utilized data from 34 Saudi Arabian projects. Two linear regression models developed the data, and the square root function transformed the data. Moreover, the artificial neural network (ANN) model was developed after data standardization using Zavadskas and Turskis’ logarithmic method. The results showed that the ANN model had a MAPE smaller than the two linear regression models. Using Zavadskas and Turskis’ logarithmic standardization method and elimination of data that had an absolute percentage error (APE) of more than 35% led to an increase in ANN model accuracy and provided a MAPE value of less than 8.5%.

Application of wavelet theory to enhance the performance of machine learning techniques in estimating water quality parameters (case study: Gao-Ping River).

There are several methods for modeling water quality parameters, with data-based methods being the focus of research in recent decades. The current study aims to simulate water quality parameters using modern artificial intelligence techniques, to enhance the performance of machine learning techniques using wavelet theory, and to compare these techniques to other widely used machine learning techniques. EC, Cl, Mg, and TDS water quality parameters were modeled using artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS). The study area in the present research is Gao-ping River in Taiwan. In the training state, using hybrid models with wavelet transform improved the accuracy of ANN models from 8.1 to 22.5% and from 25.7 to 55.3% in the testing state. In addition, wavelet transforms increased the ANFIS model's accuracy in the training state from 6.7 to 18.4% and in the testing state from 9.9 to 50%. Using wavelet transform improves the accuracy of machine learning model results. Also, the WANFIS (Wavelet-ANFIS) model was superior to the WANN (Wavelet-ANN) model, resulting in more precise modeling for all four water quality parameters.

Using Artificial Neural Networks to Predict Hardness and Impact Toughness of Aluminum Alloy 6061-T6

Predicting the material's mechanical properties is essential for reducing testing time, cost, and effort. In this study, the effect of temperature and holding time on the hardness and impact toughness of Al 6061 was investigated using the design of experiments (DOE) methodology. Analysis of variance (ANOVA) was used to analyze the results of DOE-factorial experiments. Two factors with five replicates were studied in the experiments: temperature with four levels (393.15, 423.15, 453.15, and 483.15 oK) and holding time with four levels (60, 120, 180, and 240 min). An artificial neural network (ANN) model was constructed to predict the hardness and impact toughness of precipitation-hardened 6061 aluminium alloy. The results revealed that the temperature, holding time, and interaction between them were significant factors on the hardness and impact toughness of Al 6061. ANN models' accuracy to predict the hardness and impact toughness of precipitation-hardened 6061 aluminium alloy was 99.1% and 97.6%, respectively. In this work, the ANN model accuracy was larger than ANOVA accuracy.

The Development of Advanced Mathematical Models to Predict Compaction Parameters of FineGrained Soils in Terms of Atterberg Limits

Compaction is an important engineering process that ensures the stability of soils by compressing them to a predefined strength. However, in most construction projects, particularly large-scale projects, achieving the appropriate compaction properties, such as optimum moisture content (OMC) and maximum dry density (MDD), it requires time and high cost. Predicting the compaction characteristics from the Atterberg limit, which involves simpler and faster testing techniques, becomes an important task in this scenario. The purpose of this study is to study the comparison of the multiple linear regression (MLR) method with the response surface method (RSM) and artificial neural network (ANN) to determine an accurate, efficient and simple technique to predict soil compaction parameters. For this research, 29 samples were subjected to a variety of laboratory testing. All of the parameters’ statistical relationships were analyzed. In this research, techniques are used, and the findings of these studies are discussed and analysed. To see the performance and accuracy of the model, the criteria for validation of the model used are based on the value of coefficient of determination (R2), absolute mean error (MAE), mean square error (MSE) and mean square root error (RMSE). A comparison with the test data revealed that the coefficient of determination (R2) of ANN model predictions was greater than those of other models. In addition, the findings indicate that the accuracy of ANN models are superior to the statistical models MLR and RSM.

GEU-AMP50: Enhanced antimicrobial peptide prediction using a machine learning approach

The biggest challenge to the systematic elimination and treatment of disease due to parasites, fungi, bacteria, and viruses is antimicrobial resistance (AMR). They also eliminate infections that are resistant to a variety of antibiotics. Deep learning algorithms can be used to predict AMPs with great success, which offer a viable route for novel antibiotic development. We put forward a precise computational method for the prediction of AMP using the ANN algorithm in this paper called GEU-AMP50. Presently models can predict AMP up to 30 amino acids long but GEU-AMP50 is a model that can predict AMPs with lengths ranging from 05 to 50 amino acids. This model uses the distribution patterns of amino acid which is characteristics sequence for AMPs and this pattern serve as the foundation for the prediction model. To develop models and evaluate the antimicrobial activity of peptides, Artificial Neural Network (ANN) classification techniques were employed in conjunction with validation. The model shows the training accuracy 92.7% and validation accuracy 72.2%. We compared the ANN classification technique to other models as well and Random Forest produced the best performance metrics for the dataset, with accuracy of 87.7 %, recall of 85 %, specificity of 90.7 %, precision of 84.9 %, AUC of 87.8 %, and F1 of 87.8 %. According to the results, the training accuracy of ANN model attains high efficiency for recognizing AMPs. This method of prediction could be beneficial for assessing a proposed antimicrobial peptide sequence with antimicrobial potential before moving on to experimental testing. To the best of our knowledge, the ANN Algorithm developed is among the first that has high training accuracy among all known models and can reliable predict AMP up to 50 amino acid long. This model is further being tested with wet lab experimental data.

Statistical and Artificial Neural Network Analysis for Corrosion of Mild Steel in Hydrochloric Acid in Presence of Eco-Friendly Inhibitor

In present work, the inhibition performance of the Portulaca Grandiflora Leaf Extract (PGL), as environmental-friendly corrosion inhibitor, for low-carbon steel in 0.5 M hydrochloric acid solution at variable inhibitor concentrations and temperatures is evaluated by mass loss technique. The dependent variable was corrosion rate, while the independent variables were inhibitor concentration and temperatures. Several mathematical and artificial neural network (ANN) models have been suggested. A computer aided program is used during regression and estimation processes. Several models were used. Polynomial – individual effect, polynomial – interaction effect, linear effect, exponential growth, and piecewise regression models were estimated. Results show that the Piecewise regression model was the best one with high correlation coefficient (R2) equal to 0.9994. For ANN studies, Linear Model (LM), Radial Basis Function (RBF), and Multi-Layer Perceptron (MLP) were evaluated. The data were divided into training and testing. MLP of two inputs, multi-hidden layers, and one output (2:2-8-1:1) was the highly accurate artificial neural network model ANN with a high correlation coefficient (R2=0.9826). The effect of temperature was lower than the effect of PGL concentration as shown by mathematical and ANN analysis.

Strength prediction and progressive damage analysis of carbon fiber reinforced polymer-laminate with circular holes by an efficient Artificial Neural Network

The composite laminates with circular holes find numerous applications in aerospace, automobile manufacturing and other fields due to the design and assembly of structural components. The failure analysis of composite laminates with notches or holes is of great importance in structural applications. In this work, a finite element method (FEM) based artificial neural network (ANN) model is presented to predict the strength and progressive damage behavior of carbon fiber reinforced polymer (CFRP) laminates with holes subjected to the external loads. The activation functions in the model design are reasonably chosen. The ANN prediction results are found to be in good agreement with the simulation results, thereby confirming the accuracy of ANN model. The developed ANN model is suitable for the rapid prediction of progressive damage failure behavior of the open-hole composite laminates. The elastic deformation and the progressive damage behavior of the CFRP laminates with circular holes are predicted by the proposed ANN model, which provides a good machine learning platform with high efficiency and finds potential applications in other fields.

Fast design and optimization method for an ultra-wideband perfect absorber based on artificial neural network acceleration

In recent years, ultra-wideband perfect absorbers (UWPAs) based on metamaterial nanostructures have been widely studied due to their excellent performance. However, designing and optimizing the absorber quickly and accurately remains challenging in the broad-band range. In this work, by adopting the genetic algorithm combined with artificial neural network (ANN) acceleration, an effective method with the global search ability and extremely fast optimization speed is developed, which only requires an accurate forward ANN model and offers tremendous controls to the designer. In the context of the optimization of the nano-multilayers absorber, the most of mean absolute errors (MAEs) of ANN between the actual spectral absorptance and the predicted spectral absorptance are less than 1E-2, demonstrating the outstanding consistency. What's more, the optimization speed is four orders of magnitude faster than conventional electromagnetic simulation software, 1.2 × 105 search calculations take only 573 s. Meanwhile, we also show that the hybridization between multilayer cavity plasmonic resonant modes is the main cause of the ultra-wideband absorptance. This method significantly increases the viability of more complex ultra-wideband perfect absorber designs and provides opportunities to manipulate light-matter interactions in the future.

Machine learning modeling for solubility prediction of recombinant antibody fragment in four different E. coli strains

The solubility of proteins is usually a necessity for their functioning. Recently an emergence of machine learning approaches as trained alternatives to statistical models has been evidenced for empirical modeling and optimization. Here, soluble production of anti-EpCAM extracellular domain (EpEx) single chain variable fragment (scFv) antibody was modeled and optimized as a function of four literature based numerical factors (post-induction temperature, post-induction time, cell density of induction time, and inducer concentration) and one categorical variable using artificial neural network (ANN) and response surface methodology (RSM). Models were established by the CCD experimental data derived from 232 separate experiments. The concentration of soluble scFv reached 112.4 mg/L at the optimum condition and strain (induction at cell density 0.6 with 0.4 mM IPTG for 24 h at 23 °C in Origami). The predicted value obtained by ANN for the response (106.1 mg/L) was closer to the experimental result than that obtained by RSM (97.9 mg/L), which again confirmed a higher accuracy of ANN model. To the author’s knowledge this is the first report on comparison of ANN and RSM in statistical optimization of fermentation conditions of E.coli for the soluble production of recombinant scFv.

Prediction and evaluation of fuel properties of hydrochar from waste solid biomass: Machine learning algorithm based on proposed PSO–NN model

Hydrothermal carbonization is an effective and environmentally friendly biomass pretreatment technology, which converts high moisture biomass into homogeneous, carbon–rich, and high calorific value solid hydrochar. This study aimed to predict the fuel properties of the hydrochar based on hydrothermal conditions and biomass characteristics by machine learning (ML) models. Artificial neural network (ANN) combined with particle swarm optimization (PSO) algorithm was proposed and developed based on 296 data points collected from abundant previous studies, and the prediction capability is analyzed with ordinary ANN model. The results showed that particle swarm optimization–neural network (PSO–NN) model with optimal hyper–parameters can reduce iteration time, and improve the stability and accuracy of ANN model. Fuels properties of hydrochar were predicted by PSO–NN model with R2 greater than 0.85 and the convergence speed is increased by 26.8%. Feature importance and correlation were explored by the integration of PSO–NN model and model explainer based on SHAP methodology. The result showed that the carbon content in raw biomass was the significant feature impacting mass yield, and the mass yield of hydrochar mainly depended on elemental composition of feedstock. The HTC temperature of water is the most important factor affecting HHV of the hydrochar, so raising hydrothermal temperature is the best way to improve the HHV. N content was considered as the most important parameter for the N/C molar ratio among all the evaluated features. The O content of the raw biomass had obvious influence on the ASH content of hydrochar, and the influence of operating conditions for ash content changes only accounted for 14.3%, which indicated that the removing efficiency of ash from biomass was low only by changing the operating conditions. Both DHD and DCD of hydrochars were most affected by temperature, and the ash content played a significant role in the prediction of the DHD. Furthermore, we found that most of ash remained in feedstock negatively affected DHD of the hydrochar but had a positive effect on DCD. The PSO–NN model can be used for pre–experiment condition design, which is convenient for researchers to obtain ideal hydrothermal products.