Neural Network Prediction Model Research Articles

Prediction of pipe-jacking forces using a physics-constrained neural network

Purpose A common design driver for pipe-jacking projects is the jacking force required to advance the tunnel boring machine and pipe string. Empirical methods are popular in industry but are well known to lack accuracy, while there is a strong desire to supplement such approaches with robust data-driven techniques, typically small construction datasets present significant challenges. Design/methodology/approach To address this challenge, this paper develops a physics-constrained neural network predictive model for pipe-jacking forces. Information used as input into the model includes principal design information and soil type. Findings The physics constrained model was found to predict jacking force to a higher accuracy than current industry practice and better discern meaningful patterns in data than a purely data-driven artificial neural network. The results reveal promising performance for this initial dataset such that there is motivation, as a longer-term objective, to train the present approach on a more comprehensive drive database for more reliable and cost effective solutions for new projects. Originality/value Novel contributions include (a) a bespoke framework to constrain a neural network using a pipe-jacking mechanistic model which includes stoppage-induced friction increases, (b) built-in model uncertainty for greater confidence in model outputs, (c) new historical drive data for model training and (d) one-hot encoding of soil type as a model input. The model is calibrated and validated against 14 tunnel drives across four different sites with four distinctive ground types.

Evaluation of keratometric and total corneal astigmatism measurements from optical biometers and anterior segment tomographers and mapping to reconstructed corneal astigmatism vector components.

To investigate different measures for corneal astigmatism in the context of reconstructed corneal astigmatism (recCP) as required to correct the pseudophakic eye, and to derive prediction models to map measured corneal astigmatism to recCP. Retrospective single centre study of 509 eyes of 509 cataract patients with monofocal (MX60P) IOL. Corneal power measured with the IOLMaster 700 keratometry (IOLMK), and Galilei G4 keratometry (GK), total corneal power (TCP2), and Alpin's integrated front (CorT) and total corneal power (CorTTP). Feedforward shallow neural network (NET) and linear regression (REG) prediction models were derived to map the measured C0 and C45 power vector components to the respective recCP components. Both the NET and REG models showed superior performance compared to a constant model correcting the centroid error. The mean squared prediction errors for the NET/REG models were: 0.21/0.33 dpt for IOLMK, 0.23/0.36 dpt for GK, 0.24/0.35 for TCP2, 0.23/0.39 dpt for CorT and 0.22/0.36 dpt for CorTTP respectively (training data) and 0.27/0.37 dpt for IOLMK, 0.26/0.37 dpt for GK, 0.38/0.42 dpt for TCP2, 0.35/0.36 dpt for CorT, and 0.44/0.45 dpt for CorTTP respectively on the test data. Crossvalidation with model optimisation on the training (and validation) data and performance check on the test data showed a slight overfitting especially with the NET models. Measurement modalities for corneal astigmatism do not yield consistent results. On training data the NET models performed systematically better, but on the test data REG showed similar performance to NET with the advantage of easier implementation.

CLSSATP: Contrastive learning and self-supervised learning model for aquatic toxicity prediction.

As compound concentrations in aquatic environments increase, the habitat degradation of aquatic organisms underscores the growing importance of studying the impact of chemicals on diverse aquatic populations. Understanding the potential impacts of different chemical substances on different species is a necessary requirement for protecting the environment and ensuring sustainable human development. In this regard, deep learning methods offer significant advantages over traditional experimental approaches in terms of cost, accuracy, and generalization ability. This research introduces CLSSATP, an efficient contrastive self-supervised learning deep neural network prediction model for organic toxicity. The model integrates two modules, a self-supervised learning module using molecular fingerprints for representation, and a contrastive learning module utilizing molecular graphs. Through dual-perspective learning, the model gains clear insights into the structural and property relationships of molecules. The experiment results indicate that our model outperforms comparative methods, demonstrating the effectiveness of our proposed architecture. Moreover, ablation experiments show that the self-supervised module and contrastive learning module respectively provide average performance improvements of 9.43 % and 10.98 % to CLSSATP. Furthermore, by visualizing the representations of our model, we observe that it correctly identifies the substructures that determine the molecular properties, granting itself with interpretability. In conclusion, CLSSATP offers a novel and effective perspective for future research in aquatic toxicity assessment. All of codes and datasets are freely available online at https://github.com/zhaoqi106/CLSSATP.

Optimizing Electrical Energy Management in Apartment Buildings Through Ensemble Neural Network Prediction Model

Optimizing Electrical Energy Management in Apartment Buildings Through Ensemble Neural Network Prediction Model

Research on Utility Analysis and Risk Prediction of Economic Management of Construction Enterprises in Digital Economy

This paper determines the evaluation index system of economic management utility and prediction of construction enterprises based on the theory of sustainable development of the construction industry, the theory of economics, and the principle of construction of evaluation index system. Information entropy and fuzzy evaluation are used to measure the economic management utility of construction enterprises in the digital economy. And the economic management risk aspect is analyzed by BP neural network model for prediction. After analyzing, the comprehensive evaluation results of the economic management utility of the construction enterprise [3.9848, 3.2449, 1.2547, 0.5276, 0.9895] indicate that the economic management utility of the construction enterprise is in the excellent. Based on the results of risk prediction of economic management of the construction enterprise ( R 2 = 0.9842), it was confirmed that the model has a good ability to predict the risk of economic management. The results of the study make the economic management of construction enterprises better guaranteed and stabilize the development of the regional basic economy to a certain extent.

A Typical Infrared Background Radiation Prediction Model Based on RF-VMD and Optimized Hybrid Neural Network

ABSTRACT The short-term prediction and adjustment of a target’s infrared radiation hold significant value in military camouflage applications. Existing radiation prediction models generally require real-time environmental and meteorological data support, resulting in lag in active camouflage. To meet the demand for active camouflage of background infrared (IR) radiation, a short-term background IR radiation prediction method based on historical data is proposed. First, a random forest (RF) is used to filter the collected multidimensional meteorological parameters. Variational mode decomposition (VMD) is applied for time-frequency analysis on these parameters, optimizing them with Bayesian algorithms and decomposing them into multivariate intrinsic mode functions (IMFs) with similar frequencies to reduce the impact of nonlinearity in the data. Based on the superimposed IMFs as inputs, a hybrid deep neural network prediction model is established. The model optimizes the CNN-LSTM network with residual connections and introduces a multi-head self-attention mechanism to enhance spatiotemporal feature extraction of the multidimensional meteorological parameters, focusing on key temporal feature regions. According to the experimental results, the constructed model demonstrates high prediction accuracy and adaptability across different background environments, with a low parameter count and fast prediction capability, meeting the practical application needs for various complex backgrounds.

Lower Limb Joint Angle Prediction Based on Multistream Signaling and Quantile Regression, Temporal Convolution Network–Bidirectional Long Short-Term Memory Network Neural Network

In recent years, the increasing number of patients with spinal cord injuries, strokes, and lower limb disabilities has led to the gradual development of rehabilitation-assisted exoskeleton robots. A critical aspect of these robots is their ability to accurately sense human movement intentions to achieve smooth and natural control. This paper describes research carried out on predicting the motion angles of human lower limb joints. Based on the design of a signal acquisition system for physiological muscle signals and inertial measurement unit (IMU) data, a hybrid neural network prediction model (QRTCN-BiLSTM) and a single neural network prediction model (QRBiLSTM) were constructed using quantile regression, temporal convolution network (TCN) and bidirectional long short-term memory network (BiLSTM), respectively. At the same time, 7-channel surface electromyographic signals (sEMG) and 12-channel IMU data from hip and knee joints were collected and input into the QRBiLSTM and QRTCN-BiLSTM models to unfold the training and analyze the comparison. The results show that the QRTCN-BiLSTM model can more accurately infer human movement intention and provide a more reliable and accurate prediction tool for human–robot interaction research in rehabilitation robotics.

Criteria for the spatial distribution of polymetallic ore objects as a basis for creating a predictive search model using a neural network approach (using the example of the territory of South-Eastern Transbaikalia)

The work is aimed at identifying and substantiating criteria that indirectly or actually control ore objects in order to create a predictive neural network model of the metallogenic potential of southeastern Transbaikalia. For this purpose, geological, geophysical and cartographic materials were collected and processed, including the results of the analysis of remote sensing data. Statistical analysis of the array of collected data made it possible to establish a list of the minimum necessary information to identify criteria for the localization of polymetallic ore objects within the territory of southeastern Transbaikalia. As a result, thematic schemes have been prepared reflecting the relationship between the distribution of known polymetallic mineralization zones and the identified geological and spatial features. A correlation analysis was carried out between all the criteria in order to assess the suitability of using the selected features as input data for a future neural network model.

Construction of a back propagation neural network model for predicting urosepsis after flexible ureteroscopic lithotripsy based on heparin-binding protein and C-reactive protein levels.

To analyze the association of serum heparin-binding protein (HBP) and C-reactive protein (CRP) levels with urosepsis following flexible ureteroscopic lithotripsy (FURL) and to construct a back propagation neural network prediction model. A total of 428 patients with kidney stones who underwent FURL were enrolled. Patients were divided into sepsis group (n=42) and control group (n=386) according to whether post-operative urosepsis developed. Logistic regression analysis was used to determine the risk factors and interactions of post-FURL urosepsis. A Logistic regression model and a neural network model were developed for predicting post-FURL urosepsis following FURL, and their predictive performance was evaluated using ROC curves. Univariate analysis showed that stone surgery history, gender, urine culture, stone diameter, diabetes, operation time, white blood cell (WBC), platelet, CRP, and HBP levels were significantly associated with post-FURL urosepsis (P<0.05). Multivariate analysis identified positive urine culture, CRP, and HBP levels as independent risk factors for post-FURL urosepsis (P<0.05). Interaction analysis revealed that CRP and HBP showed both additive (RERI=8.453, 95%CI: 2.645-16.282; AP=0.696, 95%CI: 0.131-1.273; S=3.369, 95%CI: 1.176-7.632) and multiplicative (OR=1.754, 95%CI: 1.218-3.650) interactions, while CRP and urine culture demonstrated multiplicative interaction (OR=2.449, 95%CI: 1.525-3.825). The neural network model showed superior predictive performance compared to the Logistic regression model. CRP and HBP levels are independent risk factors for post-FURL urosepsis. The neural network model based on CRP and HBP exhibits higher predictive accuracy than the Logistic regression model, which may provide a reliable risk assessment tool for early discrimination and intervention of post-FURL urosepsis.

A Novel Friction Mean Effective Pressure Model for a Heavy-Duty Diesel Engine by Neural Network and Its Applications to Heat Release Rate Profile Optimization

&lt;div&gt;The prediction of friction mean effective pressure (FMEP) is important when engine performance is estimated in the model-based development process. The Chen–Flynn model as a function of the maximum in-cylinder pressure (P&lt;sub&gt;max&lt;/sub&gt;) and mean piston speed (C&lt;sub&gt;m&lt;/sub&gt;) is often used to predict FMEP because of its simplicity to utilize; however, this study inferred from the results of multiple regression analysis between FMEP and factors related to combustion phase and rate of heat release profile (ROHR) that the Chen–Flynn model may be difficult to accurately estimate FMEP in a modern diesel engine with common rail fuel injection system, which allows the control of fuel injection pressure (P&lt;sub&gt;inj&lt;/sub&gt;) and combustion phase. In this study, a neural network with machine learning was applied to predict FMEP based on the expectation that the ROHR profile, which allows the reduction of FMEP may be possible to be found. 7666 points experimental results that include FMEP and combustion parameters in the heavy-duty single-cylinder diesel engine were utilized as training and validation data for machine learning. The pre-trained neural network prediction model for FMEP can predict the tendency of FMEP better than the Chen–Flynn model when start of combustion (SOC) and P&lt;sub&gt;inj&lt;/sub&gt; were varied. The error between the predicted FMEP results by the pre-trained neural network and the experimental results of FMEP were within 4% of the experimental results when SOC was advanced from top dead center to −7 deg. ATDC. Whereas, the predicted FMEP by Chen-–Flynn model had a maximum error of 15% compare to the experimental results. Furthermore, the predicted FMEP by Chen–Flynn model had a maximum error of 9% compare to the experimental results when P&lt;sub&gt;inj&lt;/sub&gt; was varied from 120 MPa to 200 MPa, whereas the error between the predicted FMEP results by the pre-trained neural network and the experimental results of FMEP were within 4% of the experimental results. The pre-trained neural network model was confirmed to be predictive better than the Chen–Flynn under the conditions that the combustion phase and P&lt;sub&gt;inj&lt;/sub&gt; are changing. In addition, the parameter study using the pre-trained neural network prediction model for FMEP and a one-dimensional engine performance simulation tool was conducted to find the ideal ROHR profile that improve brake thermal efficiency (BTE). The result of parameter study suggests that the optimum ROHR profile to improve BTE under the same conditions of P&lt;sub&gt;max&lt;/sub&gt; and the degree of constant volume combustion is to shorten combustion duration and to retard the peak of ROHR away from top dead center.&lt;/div&gt;

Research on prediction of high energy microseismic events in rock burst mines based on BP neural network

In response to the frequent occurrence of high-energy microseismic events in coal mines in China, a back propagation neural network (BPNN) prediction model based on surface subsidence data has been proposed to provide a basis for safely and efficiently predicting coal mine disasters. Theoretical research on the relationship between surface displacement, mining disturbance, and high-energy microseismic event levels has demonstrated a significant correlation among these factors. When there is a sudden increase or decrease in surface displacement or mining disturbance, the advancing working face typically exhibits dynamic characteristics. Therefore, feature parameters relevant to predicting high-energy microseismic event levels were selected as input variables for the BPNN model. Raw data from 88 sets of microseismic events in the 301 working face of the third mining area of a certain coal mine in Inner Mongolia were extracted. First, outlier preprocessing was conducted to obtain a complete dataset, which is then divided into a training set and a testing set. The BPNN model was trained with the training set and subsequently tested to evaluate its predictive performance. Finally, by comparing several model evaluation metrics, it was found that the BPNN model outperforms other common models in predicting high-energy events. The overall prediction accuracy is 86.4%, and the root mean square error (RMSE) is 0.45, indicating that the BPNN-based prediction model for high-energy microseismic events in coal mines is feasible.

Thickness monitoring of threshing mixture on the oscillating plate of corn grain harvester

Cleaning is a critical operation in the corn grain harvesting process, and the operation is seriously affected by the uneven distribution of the threshing mixture on the oscillating plate. To monitor the thickness of the threshing mixture on the oscillating plate of the corn grain harvester, in this study, we adopted the lever principle to design a corn threshing mixture thickness monitoring device based on the piezoresistive effect and theoretically analyzed the lever impact mechanics. The factors affecting the force on the sensor of the monitoring device are the vibration frequency of the oscillating plate, the thickness of the threshing mixture, and grain moisture content, which were obtained through theoretical analysis. The operational performance of the corn threshing mixture thickness monitoring device was assessed through a set of single-factor experiments, revealing the influence laws of three factors on the impact voltage: vibration frequency, thickness of the threshing mixture, and grain moisture content. A back-propagation (BP)11Back-propagation (BP) neural network prediction model was established using an orthogonal experiment with impact voltage, vibration frequency, and grain moisture content as input layers and the thickness of the corn threshing mixture as the output layer. An STM32 microcontroller was used as the primary control unit to build a thickness monitoring data acquisition and processing system that could process and operate the impact voltage collected by the pressure sensor and the vibration frequency signal collected by a Hall sensor. The data were wirelessly transmitted to a PC to predict the thickness of the threshing mixture by using the trained BP neural network. The experimental results of monitoring the variation in corn threshing mixture thickness demonstrated that the device could effectively monitor changes in thickness. The monitoring accuracy of the monitoring device was found to be over 90%, and the system can be used for monitoring the threshing mixture thickness on the oscillating plate of the corn grain harvester.

Study on the Constitutive Relationship and Mechanical Properties Fitting of SFRC Based on Optimized BP Neural Network

Abstract In order to accurately characterize the complex mechanical behavior of steel fiber reinforced concrete (SFRC) under high-speed impact and accurately describe the constitutive relationship of its mechanical behavior, the dynamic compression test of plain concrete and SFRC with two different fiber contents, 0.5% and 1.5%, was carried out based on the split Hopkinson pressure bar (SHPB). Firstly, the stress-strain curves and their failure morphology at different strain rates in the range of 47s−1-214s−1 are obtained; Next, a back propagation (BP) neural network prediction model optimized by particle swarm optimization (PSO) algorithm was established based on SFRC stress, taking strain rate, fiber content, and true strain as input parameters, and outputting the true stress; Finally, the fitted constitutive relationship of SFRC is obtained by inference based on the established model. The results show that SFRC has a strong strain rate enhancement effect in strength, and its peak stress will be continuously enhanced with the increase of strain rate. The steel fibers can effectively inhibit the generation and expansion of cracks, thereby improving the impact resistance of concrete. The PSO-BP neural network model proposed in this paper can effectively fit the constitutive relationship of SFRC, and the deviation of the model inference results is less than 10%, with the model correlation coefficient R of 0.9563 and the average relative error E of 5.92%. At the same time, the proposed model can also be used for the prediction of the performance indexes of elastic modulus and ultimate compressive strength, and the model has a good generalization ability and good fitting accuracy.

Influence and prediction of chemical composition of alfalfa on the quality and specific energy consumption of alfalfa briquettes

The chemical components and water content are related to the physical and mechanical properties of straw, affecting the processing technology that utilizes straw as energy and feed. The ratio of the viscous component to the elastic component, denoted as q, characterizes different chemical components of alfalfa. The influence of eigenvalue q on the densification of alfalfa particles was analyzed. By examining the morphology of alfalfa briquettes, the main binding mode of particles in the alfalfa briquettes was identified as the solid bridge formed by viscous components. A GA-BP-ANN artificial neural network prediction model was established to predict energy consumption and alfalfa briquette quality using q, improving the model's accuracy and stability while reducing the number of iterations. The research results contribute to determining the appropriate compression process for alfalfa, thereby promoting the efficient and high-quality utilization of alfalfa.

High-fidelity prediction of forming quality for self-piercing riveted joints in aluminum alloy based on machine learning

The effect of rivet length and sheet thickness on the cross-sectional formation and tensile-shear performance of self-piercing riveted joints in AA5754 aluminum alloy was examined through experimental investigation. The influence degree of joining parameters on the forming quality was analyzed. It was revealed that rivet length and sheet thickness are pivotal factors influencing the tensile-shear strength of the joint, culminating in the identification of four optimal riveting process schemes:LCh1Ah2A、LAh1Ah2B、LAh1Ah2B and LBh1Ah2B. A simulation model for self-piercing riveting was established, employing the GISSMO failure model and the modified Mohr-Coulomb (MMC) failure criteria to predict the damage and fracture of the aluminum alloy. A plethora of high-quality datasets depicting the cross-sections of the joints were derived from simulation analysis. Subsequently, the structure and hyperparameter determination method of traditional neural network prediction models were elucidated. By amalgamating the Aquila Optimization (AO) algorithm with the African Vultures Optimization Algorithm (AVOA), a hybrid optimization algorithm model known as MIC_AOAVOA was developed. This model effectively harnesses the strengths of various algorithms to augment search efficiency and optimization capabilities. Strategies for population initialization and adaptive weight adjustments were incorporated to enhance the algorithm's convergence velocity and the quality of solutions. The cauchy opposition-based learning (COBL) and fitness-distance balance (FDB) strategy further refined the composite algorithm, bolstering its global search capabilities and population diversity. Comparative analyses were performed with single algorithm models and traditional BP neural network models, with an in-depth examination of the MIC_AOAVOA_BP model's prediction outcomes. Comprehensive evaluations utilizing error statistics and composite evaluation indicators demonstrated that the model consistently achieved mean absolute percentage error (MAPE) values below 10%, correlation coefficients (R²) exceeding 0.98, and stable mean squared error (MSE) values around 0.0002 across the prediction of three metrics. These results underscore the model's high precision and stability. Consequently, the proposed enhanced model offers a solution that is more stable, accurate, and robust for the prediction of forming quality in self-piercing riveted joints within engineering applications.

Strip-wise controller with neural network predictive model for annealing furnace under operational constraints

Strip-wise controller with neural network predictive model for annealing furnace under operational constraints

Short-Beam Shear Strength of New-Generation Glass Fiber-Reinforced Polymer Bars Under Harsh Environment: Experimental Study and Artificial Neural Network Prediction Model.

In this study, the short-beam shear strength (SBSS) retention of two types of glass fiber-reinforced polymer (GFRP) bars-sand-coated (SG) and ribbed (RG)-was subjected to alkaline, acidic, and water conditions for up to 12 months under both high-temperature and ambient laboratory conditions. Comparative assessments were also performed on older-generation sand-coated (SG-O) and ribbed (RG-O1 and RG-O2) GFRP bars exposed to identical conditions. The results demonstrate that the new-generation GFRP bars, SG and RG, exhibited significantly better durability in harsh environments and exhibited SBSS retentions varying from 61 to 100% in SG and 90-98% in RG under the harshest conditions compared to 56-69% in SG-O, 71-80% in RG-O1, and 74-88% in RG-O2. Additionally, predictive models using both artificial neural networks (ANNs) and linear regression were developed to estimate the strength retention. The ANN model, with an R2 of 0.95, outperformed the linear regression model (R2 = 0.76), highlighting its greater accuracy and suitability for predicting the SBSS of GFRP bars.

Frost resistance prediction for rubberized concrete based on artificial neural network

Using waste rubber to partially replace fine aggregate to make rubber concrete can not only reduce black pollution to alleviate the dilemma of natural sand resource depletion, but also improve the frost resistance of concrete, which is undoubtedly a win–win solution. Aim to promote the application of rubber concrete seasonal cold regions, it is of great significance to evaluate and predict its frost-resistance. Different from ordinary concrete, the existence of rubber changes the inherent characteristics of concrete to varying degrees, which makes the durability of rubber concrete more complicated and the establishment of prediction models more challenging. In this paper, an artificial neural network (ANN) model was proposed to predict the frost-resistance of rubberized concrete. Using water-cement ratio, cement, sand, sand rate, rubber content and the number of freeze–thaw cycles as input variables and relative dynamic elastic modulus as output variables, a three-layer BP neural network (BPNN) prediction model with a hidden layer was established on the basis of a large number of experimental data of another author. The prediction results show that the proposed BPNN model has a strong ability to predict the frost resistance of rubberized concrete with satisfactory accuracy (R2 = 0.9825, MAPE = 1.5609%), which opens up a new way to improve the prediction accuracy of frost-resistance for rubberized concrete.

A Multivariable Probability Density-Based Auto-Reconstruction Bi-LSTM Soft Sensor for Predicting Effluent BOD in Wastewater Treatment Plants.

The precise detection of effluent biological oxygen demand (BOD) is crucial for the stable operation of wastewater treatment plants (WWTPs). However, existing detection methods struggle to meet the evolving drainage standards and management requirements. To address this issue, this paper proposed a multivariable probability density-based auto-reconstruction bidirectional long short-term memory (MPDAR-Bi-LSTM) soft sensor for predicting effluent BOD, enhancing the prediction accuracy and efficiency. Firstly, the selection of appropriate auxiliary variables for soft-sensor modeling is determined through the calculation of k-nearest-neighbor mutual information (KNN-MI) values between the global process variables and effluent BOD. Subsequently, considering the existence of strong interactions among different reaction tanks, a Bi-LSTM neural network prediction model is constructed with historical data. Then, a multivariate probability density-based auto-reconstruction (MPDAR) strategy is developed for adaptive updating of the prediction model, thereby enhancing its robustness. Finally, the effectiveness of the proposed soft sensor is demonstrated through experiments using the dataset from Benchmark Simulation Model No.1 (BSM1). The experimental results indicate that the proposed soft sensor not only outperforms some traditional models in terms of prediction performance but also excels in avoiding ineffective model reconstructions in scenarios involving complex dynamic wastewater treatment conditions.

Experimental analysis and low-damage machining strategy for composite ultrasonic vibration-assisted grinding of silicon carbide based on DA-MLP-NSGA-II algorithm

At present, because of the lack of ultrasonic composite vibration assisted grinding mechanism, neural network optimization algorithm (NNOA) is used to optimize the processing results. In NNOA, multi-layer perceptron (MLP) neural network model and non-dominated sorting genetic algorithm-II (NSGA-II) are very efficient and accurate methods. In this paper, based on the measurement and analysis of the specific ultrasonic vibration device, the CUVAG experiments on silicon carbide (SiC) ceramic were carried out to investigate the influence of processing parameters on the grinding forces, the ground surface roughness and morphology, and the subsurface damage. Then, the brittle-ductile removal behavior of hard-and-brittle materials could be revealed according to the above analysis. After that, MLP model and NSGA-II were utilized to predict and optimize the processing results in CUVAG. The results show that the grinding forces are basically constant, the surface quality deteriorates, and the subsurface damage increases with increased axial vibration amplitude and workpiece infeed speed, but all fluctuate with enlarged wheel speed, and turns at the inflection point of brittle-ductile transition with increased elliptic vibration amplitude. The fitting goodness R2 of the established MLP neural network prediction model is between 0.94 and 0.975, and the process parameters calculated by the NSGA-II optimization algorithm are verified. With optimized processing parameters, the grinding forces are reduced by about 13 %, the surface roughness is reduced to Ra0.037 μm (by 29 %), and the depth of subsurface damage is reduced by 68 %.