Prediction Error Of Model Research Articles

Investigating the relationship between CSAT scores and prefrontal fNIRS signals during cognitive tasks using a quantum annealing algorithm

Academic achievement is a critical measure of intellectual ability, prompting extensive research into cognitive tasks as potential predictors. Neuroimaging technologies, such as functional near-infrared spectroscopy (fNIRS), offer insights into brain hemodynamics, allowing understanding of the link between cognitive performance and academic achievement. Herein, we explored the association between cognitive tasks and academic achievement by analyzing prefrontal fNIRS signals. A novel quantum annealer (QA) feature selection algorithm was applied to fNIRS data to identify cognitive tasks correlated with CSAT scores. Twelve features (signal mean, median, variance, peak, number of peaks, sum of peaks, range, minimum, kurtosis, skewness, standard deviation, and root mean square) were extracted from fNIRS signals at two time windows (10- and 60-s) to compare results from various feature variable conditions. The feature selection results from the QA-based and XGBoost regressor algorithms were compared to validate the former’s performance. In a two-step validation process using multiple linear regression models, model fitness (adjusted R2) and model prediction error (RMSE) values were calculated. The quantum annealer demonstrated comparable performance to classical machine learning models, and specific cognitive tasks, including verbal fluency, recognition, and the Corsi block tapping task, were correlated with academic achievement. Group analyses revealed stronger associations between Tower of London and N-back tasks with higher CSAT scores. Quantum annealing algorithms have significant potential in feature selection using fNIRS data, and represents a novel research approach. Future studies should explore predictors of academic achievement and cognitive ability.

Data-Driven Generative Model Aimed to Create Synthetic Data for the Long-Term Forecast of Gas Turbine Operation

Abstract The prediction of gas turbine (GT) future health state plays a strategic role in the current energy sector. However, training an accurate prognostic model is challenging in case of limited historical data (e.g., new installation). Thus, this paper develops a Generative Adversarial Network (GAN) model aimed to generate synthetic data that can be used for data augmentation. The GAN model includes two neural networks, i.e., a generator and a discriminator. The generator aims to generate synthetic data that mimic the real data. The discriminator is a binary classification network. During the training process, the generator is optimized to fool the discriminator in distinguishing between real and synthetic data. The real data employed in this paper were taken from the literature, gathered from three GTs, and refer to two quantities, i.e., corrected power output and compressor efficiency, which are tracked during several years. Three different analyses are presented to validate the reliability of the synthetic dataset. First, a visual comparison of real and synthetic data is performed. Then, two metrics are employed to quantitively evaluate the similarity between real and synthetic data distributions. Finally, a prognostic model is trained by only using synthetic data and then employed to predict real data. The results prove the high reliability of the synthetic data, which can be thus exploited to train a prognostic model. In fact, the prediction error of the prognostic model on the real data is lower than 2.5% even in the case of long-term prediction.

The study of solid flame model of diffusion jet flame for the ammonia and propane mixed gas under the inclined ceiling

In order to solve the environmental pollution and energy dilemma, the ammonia has been considered as a new fuel and a carbon-free hydrogen carrier, which can be blended into the traditional hydrocarbon fuels. In this study, the evolution of flame emissivity, view factor and radiative heat flux have been investigated under different mass flow rates of ammonia and propane, and inclined angle of ceiling. A radiation prediction model of ammonia-propane mixed gas jet flame under inclined ceiling with different inclined angles has been developed. This prediction model can accurately describe the flame morphology and simplify the calculations, and consider the influence of mixing ratio on combustion reaction. The experimental results indicated that the flame emissivity, view factor, and radiative heat flux decrease with the increase of ammonia concentration. The prediction error of the traditional model will be enlarged with the increase in the inclined angle of ceiling. The modified model added the undersurface of ceiling jet flame, which reduces the prediction error from 25 % to 14 %. These prediction models are helpful in predicting the thermal radiation of jet flame under inclined ceiling.

Modeling and identification of hysteresis of marine damper considering shock environment based on evolutionary sparrow search algorithm

The mechanical properties of marine dampers are usually highly nonlinear in order to adapt to dynamic and shock environments. The dampers are subjected to dynamic and shock tests and the hysteresis curves are found to be rate-dependent and asymmetric. To be able to describe the dynamic hysteresis and shock hysteresis, a rate-dependent generalized Prandtl-Ishlinskii (RDGPI) model is proposed based on the generalized Prandtl-Ishlinskii (GPI) model, which is a hybrid of the GPI model and the radial basis function (RBF) neural network. Due to the large number of parameters in the model, thus aiming to improve the effect of parameter identification, an evolutionary sparrow search algorithm mixed with sparrow search algorithm and differential evolutionary algorithm is proposed. Compared with other algorithms, sparrow search algorithm is verified to have ideal optimization effect and convergence speed, and it is not easy to fall into the local optimum. Finally, the RDGPI model is experimentally verified, and the results show that the prediction error of the RDGPI model for dynamic hysteresis is within 0.06 kN, and the prediction error for shock hysteresis is within 0.8 kN, which is capable of describing the rate-dependent characteristics of the damper hysteresis. And it has a very good generalization ability, which makes it very suitable to be used in the modeling of the mechanical characteristics of the marine damper.

ATHENA-FL: Avoiding Statistical Heterogeneity with One-versus-All in Federated Learning

Federated learning (FL) is a distributed approach to train machine learning models without disclosing private data from participating clients to a central server. Nevertheless, FL training struggles to converge when clients have distinct data distributions, which leads to an increased training time and model prediction error. We propose ATHENA-FL, a federated learning system that considers clients with heterogeneous data distributions to generate accurate models in fewer training epochs than state-of-the-art approaches. ATHENA-FL reduces communication costs, providing an additional positive aspect for resource-constrained scenarios. ATHENA-FL mitigates data heterogeneity by introducing a preliminary step before training that clusters clients with similar data distribution. To handle that, we use the weights of a locally trained neural network used as a probe. The proposed system also uses the one-versus-all model to train one binary detector for each class in the cluster. Thus, clients can compose complex models combining multiple detectors. These detectors are shared with all participants through the system's database. We evaluate the clustering procedure using different layers from the neural network and verify that the last layer is sufficient to cluster the clients efficiently. The experiments show that using the last layer as input for the clustering algorithm transmits 99.68% fewer bytes to generate clusters compared to using all the neural network weights. Finally, our results show that ATHENA-FL correctly identifies samples, achieving up to 10.9% higher accuracy than traditional training. Furthermore, ATHENA-FL achieves lower training communication costs compared with MobileNet architecture, reducing the number of transmitted bytes between 25% and 97% across evaluated scenarios.

Research on the Development Height Prediction Model of Water-Conduction Fracture Zones under Conditions of Extremely Thin Coal Seam Mining

Addressing the difficult problem of predicting the height of water-conducting fracture zones in shallow and thin coal seams, a prediction model of water-conduction fracture zones based on a backpropagation (BP) neural network was developed by integrating theoretical analysis, field measurements, and algorithmic advancements. Firstly, through overburden migration analysis and correlation tests, the height index system of the water-conducting fracture zone was determined. This system includes mining height, buried depth, dip angle, working face width, and overburden rock lithology, with five groups of characteristic parameters. Then, 35 pairs of minefield-measured data were collected to establish the measured height data set of the water-conducting fracture zone. Secondly, a BP neural network prediction model and a traditional support vector regression (SVR) prediction model were constructed based on a Pytorch framework, and the models were trained and tested by selecting data sets. Thirdly, the optimal prediction model was determined by comparing the model with the empirical model and multiple regression model of mining regulations for coal pillar maintenance and pressure in buildings, water bodies, railways, and main shafts. Finally, a typical mine was selected for application to verify the suitability of the optimal model. The results show that: (1) the predicted value of the neural network model is consistent with the change trend of the measured value, which accords with the theoretical law; (2) compared with traditional forecasting methods, the error of the BP neural network prediction model is stable and the prediction effect is the best; (3) dropout can effectively mitigate mitigation training overfitting, achieve regularization, and improve prediction accuracy; (4) the field application further verified that the BP neural network model is the best for predicting the height of water-conducting fracture zones of extremely thin coal seams, and the research results can provide technical guidance for similar fragile coal seams.

High-speed machining simulation of Ti6Al4V using a thermo-mechanical coupling model and velocity-dependent friction model

PurposeThis study aims to establish a thermally coupled two-dimensional orthogonal cutting model to further improve the modeling process for systematic evaluation of material damage, stiffness degradation, equivalent plastic strain and other material properties, along with cutting temperature distribution and cutting forces. This enhances modeling efficiency and accuracy.Design/methodology/approachA two-dimensional orthogonal cutting thermo-mechanical coupled finite element model is established in this study. The tanh material constitutive model is used to simulate the mechanical properties of the material. Velocity-dependent friction model between the workpiece and the tool is considered. Material characteristics such as material damage, stiffness degradation, equivalent plastic strain and temperature field during cutting are evaluated through computation. Contact pressure and shear stress on the tool surface are extracted for friction analysis.FindingsSpeed-dependent friction models predict cutting force errors as low as 8.6%. The prediction errors of various friction models increase with increasing cutting forces and depths of cut, and simulation results tend to be higher than experimental data.Social implicationsThe current research results provide insights into understanding and controlling tool-chip friction in metal cutting, offering practical recommendations for friction modeling and machining simulation work.Originality/valueThe originality of this research is guaranteed, as it has not been previously published in any journal or publication.Peer reviewThe peer review history for this article is available at: https://publons.com/publon/10.1108/ILT-05-2024-0162/

An integrated optimization method for measurement points layout and error modeling for digital twin of CNC machine tools

Thermal error significantly influences the accuracy of precision computer numerical control (CNC) machine tools. The key to compensating thermal error lies in selecting appropriate temperature measurement points and establishing an accurate error prediction model. Traditional methods separate measurement points selection and prediction modeling, that is, selecting temperature measurement points first and then establishing the prediction model using these points. These methods are difficult to achieve optimal matching between measurement points and the prediction model, resulting in shortcomings in modeling accuracy. To address these challenges, an integrated optimization method for measurement points layout and error modeling for digital twin of CNC machine tools is proposed. A dual-stage attention-based long short-term memory combined with convolutional neural network (DA-CLSTM) error modeling method is proposed to accurately predict the thermal error for different numbers of temperature measurement points. Then, an integrated method of virtual-real temperature measurement points layout and error modeling is proposed, which ensures the temperature measurement points and the error prediction model are closely matched, offering rational sensors layout and high-accuracy prediction. Finally, experiments are conducted on the spindle system test bench to validate the method proposed in this research. Through integrated optimization of measurement points layout and the prediction model, the proposed method achieves high-accuracy thermal error prediction across various sensor counts and maintains reliable prediction ability when the number of sensors is reduced. This method is applied to a digital twin system of MKL7150 grinding machine, and effectively improves the accuracy of actual machining.

An offline data-driven dual-surrogate framework considering prediction error for history matching

High computer power has long been a critical ingredient that affects the effectiveness and efficiency of history matching. Data-driven surrogate modeling as an efficient strategy can accelerate the history-matching process by constructing machine learning-based models with high computing speed but reduced accuracy. However, the applicability of surrogate models for different history-matching problems is uncertain due to the influence of data quality and quantity, model architectures, and hyperparameters. To overcome this issue, an offline data-driven dual-surrogate framework (ODDF) that considers the prediction error of surrogate models for history matching is proposed, where one surrogate model predicts the production data of reservoirs and the other one learns the prediction error of the former surrogate. The first surrogate model considers the time-series characteristics of production data using a recurrent neural network, while the second surrogate model regards the two-dimensional spatial correlation characteristics of multivariate prediction error using a fully convolutional neural network. Furthermore, an enhanced error model is applied to incorporate the prediction error into the objective function to reduce the influence of the prediction error on inversion results. Based on this hybrid framework, one can improve the prediction accuracy of surrogate models in history matching when the architectures or hyperparameters of surrogate models are not optimal. Additionally, one can obtain satisfactory results for history matching and uncertainty quantification based on surrogate modeling. The proposed framework is validated on the history matching of two- and three-dimensional reservoir models. The results show that the proposed method is robust in constructing the surrogate models and predicting the production data of reservoirs, which improves the efficiency and reliability of history matching.

Predicting error for machining thin-walled blades considering initial error

A multistage machining process is employed to machine the blades with low stiffness. Nonetheless, machining errors can be transferred and accumulate throughout the multistage machining process, complicating the precise prediction of the final accuracy of thin-walled blades. Consequently, this paper introduces a machining accuracy prediction model for thin-walled blades that takes into account initial error. The machining error prediction model of thin-walled blades is developed using Gaussian process regression optimized by the sparrow search algorithm (SSA-GPR) with the initial contour error, depth of cut, feed per tooth, and spindle speed as inputs, and the machining error as the output. And the results show that the prediction accuracy of the SSA-GPR is 6.73 % higher than that of the Gaussian process regression (GPR), 13.73 % higher than that of the back propagation neural network (BPNN), and 32.32 % higher than that of the support vector machine regression (SVR). The influence of the initial error and milling parameters on the machining error is analyzed through the length-scales of the Gaussian kernel function. The findings indicate that the depth of cut, feed per tooth and initial error significantly affect the machining error, whereas the spindle speed has a minor impact on the machining error. Furthermore, the 3D graph based on the SSA-GPR shows that the increase of the initial error will increase the machining error of thin-walled blades. This research provides a theoretical foundation for the process optimization of thin-walled blades.

An effective hybrid attention model for crop yield prediction using IoT-based three-phase prediction with an improved sailfish optimizer

Internet of Things (IoT)-based smart agriculture approaches have been adopted to achieve promising results in terms of food productivity. Moreover, high cost and technological complexity are major problems in existing agricultural management systems. Thus, a new IoT-based smart agriculture farming model is implemented in three phases namely crop detection, crop disease detection, and crop yield prediction. In the first phase, the plant leaf images are gathered from the Kaggle source PlantifyDr. The collected images are passed to the deep feature retrieval phase, where the Visual Geometry Group (VGG16) is utilized to calculate feature (f1). Then, a Hybrid Attention-based Crop Type Detection Network (HA-CTDecNet) is used for detecting the crop types, in which the parameters are optimized by Attack Power Improved Sailfish Optimizer (API-SFO). In the second step of crop disease detection phase, the input images are fed to Otsu Thresholding for Abnormality segmentation. The deep attributes from the segmented images are haul out using VGG16 and attain the feature (f2). The extracted features are forwarded to the Hybrid Attention-based Crop Disease Detection Network (HA-CDDecNet), where the constraints are optimized using API-SFO. In the third crop yield prediction phase, the data related to crop location are collected from the crop yield forecast dataset. Then, the collected data and the extracted features (f1) and (f2) are concatenated and fed to the Principle Component Analysis (PCA). Then, these features are given to the Hybrid Attention-based Crop Yield Prediction Network (HA-CYPreNet) with the parameter optimization of API-SFO. The investigation is accepted to verify the effectiveness of the suggested IoT-based smart agriculture model with the traditional approaches by considering distinct measures. The Mean Absolute Error (MAE) of the crop yield prediction model is improved with 35% than CNN, 27.5% than LSTM, 17.5% than GRU, and 7.5% than LSTM_GRU while considering the learning percentage value as 55.

The research model of forecasting agricultural carbon emissions based on ARIMA-LSTM

To accurately forecast the volume of agricultural carbon emissions in China, this paper proposes a forecasting model that combines the Autoregressive Integrated Moving Average Model (ARIMA) with Long Short-Term Memory networks (LSTM) to better handle both linear and non-linear parts in time series. Using the data on China's agricultural carbon emissions collected from 2000 to 2023 as the research subject, the data from 2000 to 2016 were first used as the training set, and the data from 2017 to 2023 were used as the test set. The Spearman's rank correlation coefficient method was utilized to study the correlation between various types of carbon emissions. Then, the established ARIMA-LSTM forecasting model was applied to predict the emissions in the test samples and make comparisons. The Absolute Relative Error (ARE) was adopted as the evaluation metric for different models, comparing the prediction errors of the ARIMA-LSTM model with those of the ARIMA and LSTM single models through analysis. The results show that the positive correlation between diesel carbon emissions and film carbon emissions was the highest; the absolute relative error of the ARIMA-LSTM model did not exceed 7%, which is significantly lower than the other two comparison models. This research model achieves high-accuracy predictions of carbon emissions, which is beneficial for the control and management of the greenhouse effect.

A feature-specific prediction error model explains dopaminergic heterogeneity.

The hypothesis that midbrain dopamine (DA) neurons broadcast a reward prediction error (RPE) is among the great successes of computational neuroscience. However, recent results contradict a core aspect of this theory: specifically that the neurons convey a scalar, homogeneous signal. While the predominant family of extensions to the RPE model replicates the classic model in multiple parallel circuits, we argue that these models are ill suited to explain reports of heterogeneity in task variable encoding across DA neurons. Instead, we introduce a complementary 'feature-specific RPE' model, positing that individual ventral tegmental area DA neurons report RPEs for different aspects of an animal's moment-to-moment situation. Further, we show how our framework can be extended to explain patterns of heterogeneity in action responses reported among substantia nigra pars compacta DA neurons. This theory reconciles new observations of DA heterogeneity with classic ideas about RPE coding while also providing a new perspective of how the brain performs reinforcement learning in high-dimensional environments.

Reduction of the model prediction error in vibration-based SHM applications

The accuracy of an SHM scheme is partly controlled by the accuracy of the predictive models involved within the design process of the detector that associates sensor data with a corresponding health state. On the other hand, the accuracy of the predictive model, which is quantified by the error between predictions and observations, is partly controlled by its parameter magnitudes. This work provides a unified framework that allows for reducing the model prediction error by updating the model parameters in the light of experimental observations. The problem is cast in a probabilistic setting and tackled through inferential statistics. Forward and inverse uncertainty quantification approaches are employed in a sequential manner. The framework is able to yield point estimates in accordance to a Maximum Likelihood Estimator (MLE) and is also able to construct credible intervals through computational Bayesian updating. The former is treated by genetic algorithms whereas the later through Markov Chain Monte Carlo computational statistics. The framework is showcased on a vibration problem of the simplified benchmark case named as “Jim Beam”. The first ten modal parameters are considered as the discriminant features of interest . A surrogate of the finite element model of the considered specimen serves as the predictive model and the eigenfrequencies are extracted from a linear modal analysis. The parameters of inferential interest are the Young’s modulus and an artificial stiffness related parameter that is introduced so as to model the flexural rigidity of the specimen’s bolted connections. Experimental Modal Analysis and Operational Modal Analysis are the two approaches that were employed for experimentally identifying the modes and their corresponding frequencies. Results are provided from both the MLE and the Bayesian updating process and the effectiveness of each approach is discussed.

Research on load prediction of low-calorific fuel fired gas turbine based on data and knowledge hybrid model

The high-precision load prediction technology plays a vital role in load control and health management for gas turbines. Low-calorific fuel fired gas turbines pose an especially significant challenge for load prediction due to the frequent and wide fluctuations of gas parameters. First, an improved hybrid-dimension physical-based model is proposed, which is corrected based on segmented operation data and integrates zero-dimensional thermodynamic knowledge of gas turbines with three-dimensional fluid dynamics simulation knowledge. Secondly, an optimal data-driven model is proposed through conducting a systematic comparative study on four popular machine learning models with two types of input variables feature extraction. Then, a novel hybrid model is proposed based on the improved physical-based model, optimal data-driven model, and the principle of minimizing hybrid model errors. The average relative prediction error of the hybrid model over the complete operation range is 0.68%. Compared with the physical-based model and the data-driven model, the proposed hybrid model can improve the prediction accuracy by 63% and 29%, respectively. As the high-calorific fuel fired gas turbine system is relatively simple and has a small range of fuel fluctuations, the method established in this paper can be extended to high-calorific fuel fired gas turbines, which is of great significance for promoting the dynamic balance of loads in a new type of electric power system based on new energy.

Environmental information-assisted intelligent fusion localization for vehicles in urban area

In this paper, we propose a localization methodology aimed at improving accuracy through two primary aspects: environment information assisting fusion algorithm and the elaborate localization errors prediction. First, the environment-assisted particle filter (EA-PF) is proposed by enhancing the particle filter algorithm with perceived environmental feature information, which is reflected by the rasterized map constructed based on LiDAR sensors. Then, an elaborate localization error prediction model is designed to accurately forecast Inertial Navigation System (INS) errors. The model comprises the noise characteristic feature extraction and encoding network and multi-feature error prediction network. The localization errors in the EA-PF output can be further compensated. The proposed localization methodology can be divided into two operational modes based on the availability of Global Navigation Satellite System (GNSS) signal: GNSS-available and GNSS-unavailable modes. In the GNSS-available mode, the EA-PF algorithm yields precise localization results, while in the GNSS-unavailable modes, the elaborate localization error prediction model can effectively predict INS errors and provide the compensation for the EA-PF. To validate the effectiveness of the proposed methodology, relevant test experiments were performed using the KITTI dataset. The experimental validation results demonstrate the feasibility and effectiveness of the proposed methodology. The experimental results demonstrate that the EA-PF outperforms traditional PF methods by improving the root mean square error (RMS.) by 32.70 % on average during 90 s GNSS outages. Moreover, the proposed localization methodology achieves 2.30 m RMS error on average during 90 s GNSS outage.

Estimation of biomass in various components of Pinus koraiensis based on Bayesian methods

IntroductionPinus koraiensis is a dominant tree species in northeastern China. Estimating its biomass is required for forest carbon stock monitoring and accounting.MethodsThis study investigates biomass estimation methods for P. koraiensis components. A Bayesian approach was used to synthesize the parameter distributions of 298 biomass models as prior information to estimate the trunk, branch, leaf, and root biomass of P. koraiensis. The results were compared with non-informative prior and the minimum least squares (MLS).ResultsThe results indicated that the Bayesian approach outperformed the other methods regarding model fit and prediction error. In addition, the responses of different components to tree height varied. The models of trunk and root biomass exhibited a smaller response to tree height, whereas those of branches and leaves showed a larger response to tree height. The model parameters yield precise estimations.DiscussionIn sum, this study highlights the potential of the Bayesian methods in estimating P. koraiensis biomass and proposes further enhancements to improve estimation accuracy.

The bi-long short-term memory based on multiscale and mesoscale feature extraction for electric load forecasting

Accurate power load prediction is beneficial to the efficient use of electric energy and the orderly development of power systems. Given the strong volatility and complexity of power load series, a hybrid load forecasting method based on multiscale and mesoscale information fusion, signal decomposition, model optimization, and bi-long-short-term memory (BiLSTM) is proposed. Firstly, the load sequence is analyzed on different time scales, and the extracted multi-scale information and mesoscale information are fused to improve the perception ability. Secondly, the empirical wavelet transform (EWT) with adaptive decomposition ability is used to decompose the sequence and extract the rich feature information. Thirdly, the complexity, volatility, and uncertainty of each mode component were analyzed, the data features were fully mined, and the feature fusion was carried out by the TOPISIS evaluation method. The BiLSTM model and the GWO-BiLSTM model are used to predict the low-frequency component and the high-frequency component, respectively. The optimization of Grey Wolf optimization (GWO) algorithm can improve the BiLSTM model's ability to learn long-term time series. Finally, the analysis of application examples shows that compared with various prediction models, the prediction error of mixed model EWT-SGEO-BiLSTM is the smallest, MAPE is as low as 1.07 %, and goodness of fit R2 is 0.99 which verifies the accuracy and applicability of the intelligent model.

Toward a Comprehensive Evaluation of Student Knowledge Assessment for Art Education: A Hybrid Approach by Data Mining and Machine Learning

By analyzing students’ understanding of a certain subject’s knowledge and learning process, and evaluating their learning level, we can formulate students’ learning plans and teachers’ curricula. However, the large amount of data processing consumes a lot of manpower and time resources, which increases the burden on educators. Therefore, this study aims to use a machine learning model to build a model to evaluate students’ learning levels for art education. To improve the prediction accuracy of the model, SVM was adopted as the basic model in this study, and was combined with SSA, ISSA, and KPCA-ISSA algorithms in turn to form a composite model. Through the experimental analysis of prediction accuracy, we found that the prediction accuracy of the KPCA-ISSA-SVMM model reached the highest, at 96.7213%, while that of the SVM model was only 91.8033%. Moreover, by putting the prediction results of the four models into the confusion matrix, it can be found that with an increase in the complexity of the composite model, the probability of classification errors in model prediction gradually decreases. It can be seen from the importance experiment that the students’ achievements in target subjects (PEG) have the greatest influence on the model prediction effect, and the importance score is 9.5958. Therefore, we should pay more attention to this characteristic value when evaluating students’ learning levels.

A Novel Approach to Surface Roughness Virtual Sample Generation to Address the Small Sample Size Problem in Ultra-Precision Machining.

Surface roughness is one of the main bases for measuring the surface quality of machined parts. A large amount of training data can effectively improve model prediction accuracy. However, obtaining a large and complete surface roughness sample dataset during the ultra-precision machining process is a challenging task. In this article, a novel virtual sample generation scheme (PSOVSGBLS) for surface roughness is designed to address the small sample problem in ultra-precision machining, which utilizes a particle swarm optimization algorithm combined with a broad learning system to generate virtual samples, enriching the diversity of samples by filling the information gaps between the original small samples. Finally, a set of ultra-precision micro-groove cutting experiments was carried out to verify the feasibility of the proposed virtual sample generation scheme, and the results show that the prediction error of the surface roughness prediction model was significantly reduced after adding virtual samples.