Validation and data splitting in predictive regression modeling of honing surface roughness data

Model validation is critical in predicting the performance of manufacturing processes. In predictive regression, proper selection of variables helps minimize the model mismatch error, proper selection of models helps reduce the model estimation error, and proper validation of models helps minimize the model prediction error. In this paper, the literature is briefly reviewed and a rigorous procedure is proposed for evaluating the validation and data splitting methods in predictive regression modeling. Experimental data from a honing surface roughness study will be used to illustrate the methodology. In particular, the individual versus average data splitting methods as well as the fivefold versus threefold cross-validation methods are compared. This paper shows that statistical tests and prediction errors evaluation are important in subset selection and cross-validation of predictive regression models. No statistical differences were found between the fivefold and the threefold cross-validation methods, and between use of the individual and average data splitting methods in predictive regression modeling.     

Predicting glass furnace output using statistical and neural computing methods

This paper describes the development of predictive models for glass production at a regional manufacturing company. The objectives of the models are to predict the actual batch tonnage produced per week from the glass furnace based on the planned production schedule. Four modelling methods were explored: (i) linear regression; (ii) nonlinear regression; (iii) artificial neural network using back-propagation; and (iv) radial basis function neural network. Using 175 cases of production schedule data and subsequent furnace output, the two neural network-based prediction models resulted in lower average absolute error and lower maximum absolute error than the linear or nonlinear regression models. Accurate neural network-based prediction models of furnace output will subsequently be used in the overall production planning system by utilizing estimates of furnace output to determine the necessary energy, raw material, repair and personnel requirements of the glass manufacturing facility.     

Cost-tolerance analysis model based on a neural networks method

The purpose of tolerance design in product components is to produce a product with the least manufacturing cost possible, while meeting all functional requirements of the product. The product designer and process planner must fully understand the process accuracy and manufacturing cost of all kinds of manufacturing process to perform a good process plan job. Usually, the cost-tolerance model is constructed by a linear or non-linear regression analysis based on the data of the cost-tolerance experiment and to derive the correlation curve between the two. Though these correlation curves can show the relationship between manufacturing cost and tolerance, a fitting error is inevitable. In particular, there is considerable discrepancy in terms of the non-experimental data. A cost-tolerance analysis model based on a neural networks method is proposed. The cost-tolerance experimental data are used to set the training sets to establish a cost-tolerance network. Three representation modes of the cost-tolerance relationship are presented. First, the cost-tolerance relationship is derived from the grid points setting by the required tolerance accuracy. Second, a reasonable manufacturing cost of an unknown cost-tolerance experimental pair can be derived by the simulation of a cost-tolerance network. Third, an inference model based on a network's output is proposed to express the scope of the cost variation of various tolerances by means of a cost band. Comparison is also made with the high-order polynomial power function and exponential function cost-tolerance curves adopted by Yeo et al . Analytical results prove that the application of the cost-tolerance analysis model based on neural networks yields better performance in controlling the average fitting error than all conventional fitting models. The representation model using a cost band can identify precisely the possible cost variation range and reduce the chances of error in the tolerance design and cost estimation. It can thus provide important references for tolerance designers and process planners.     

Surface roughness predictive modeling: neural networks versus regression

Surface roughness plays an important role in product quality and manufacturing process planning. This research focuses on developing an empirical model for surface roughness prediction in finish turning. The model considers the following working parameters: work-piece hardness (material), feed, cutter nose radius, spindle speed and depth of cut. Two competing data mining techniques, nonlinear regression analysis and computational neural networks, are applied in developing the empirical models. The values of surface roughness predicted by these models are then compared with those from some of the representative models in the literature. Metal cutting experiments and tests of hypothesis demonstrate that the models developed in this research have a satisfactory goodness of fit. It has also presented a rigorous procedure for model validation and model comparison. In addition, some future research directions are outlined.     

表面滚压强化工艺对FV520B钢表面完整性的影响及预测模型建立

目的 基于多元回归法和BP神经网络建立预测模型,实现对滚压后试件表面完整性指标的精准控制,从而指导实际加工生产。方法 以FV520B钢为研究对象,以滚压工艺参数(压强、进给量、滚压速度)为影响因素,以材料表面完整性指标(表面粗糙度、表面硬度、塑性变形层深度)为评价指标,设计了正交试验。通过对正交试验数据进行方差分析和信噪比分析,探究了滚压工艺参数对FV520B钢表面完整性的影响。基于正交试验数据构建了多元回归预测模型和BP神经网络预测模型,并对2种模型的有效性和精准度进行了分析和比较。结果 进给量对表面粗糙度有显著影响,随着进给量的增大,表面粗糙度也显著增大。压强和进给量对塑性变形层深度均有显著影响,且塑性变形层深度随着压强的增大而增大,随着进给量的增大而减小。多元回归法建立的预测模型的拟合度较差,而BP神经网络预测模型的实验值和预测值的相对误差均在10%以下,预测效果较好。结论 相比于多元回归预测模型,BP神经网络预测模型具有误差小、泛化性能好等优点,能够实现对滚压后试件表面完整性指标的精准控制,为实际的加工生产提供一定的指导。     

Surface roughness modelling in hard turning operation of AISI 4140 using CBN cutting tool

In this study, the influence of hardness (H) and spindle speed (N) on surface roughness (Ra) in hard turning operation of AISI 4140 using CBN cutting tool has been studied. A multiple regression analysis using analysis of variance is conducted to determine the performance of experimental values and to show the effect of hardness and spindle speed on the surface roughness. Artificial neural network (ANN) and regression methods have been used for modelling of surface roughness in hard turning operation of AISI 4140 using CBN cutting tool. The input parameters are selected to be as hardness and spindle speed and the output is the surface roughness. Regression and artificial neural network optimum models have been presented for predicting surface roughness. The predicted surface roughness by the employed models has been compared with the experimental data which shows the preference of ANN in prediction of surface roughness during hard turning operation. Finally, a reverse ANN model is constructed to estimate the hardness and spindle speed from surface roughness values. The results indicate that the reverse ANN model can predict hardness for the train data and spindle speed for the test data with a good accuracy but the predicted spindle speed for the train data and the predicted hardness for the test data don’t have acceptable accuracy.     

基于回归方法和神经网络的电弧增材制造单道成形形貌预测

目的 针对电弧增材制造技术实际应用中工艺参数选取困难和成形结果难预测的问题,确定高效、准确的电弧增材制造单道成形形貌预测的数学方法,以快速、方便地选取丝材电弧增材制造工艺参数并指导成形质量控制。方法 在单道单层丝材电弧增材制造实验的基础上,采用多种回归方法和神经网络方法分别建立焊接电流、电压和焊枪移动速度等多个工艺参数与增材层宽度、增材层高度及熔池深度等成形形貌参数之间的数学关系模型。结果 电弧增材制造单道成形形貌与焊接电流、电压和焊枪移动速度显著相关,且各参数间存在非线性交互作用;采用多元线性回归法可较准确地预测单道增材层宽度,但对于增材层高度和熔深的预测效果较差;神经网络可良好地处理各工艺参数间复杂的非线性关系,其对增材层宽度、增材层高度和熔深的预测平均误差率分别为4.17%、6.60%和7.01%,显著优于多元线性回归法。结论 采用神经网络法可以准确预测电弧增材制造单道成形的形貌参数,进而指导增材制造工艺参数的选取和成形质量的控制。     

Design of manufacturing systems by a hybrid approach with neural network metamodelling and stochastic local search

Metamodels are models of simulation models. Metamodels are able to estimate the simulation responses corresponding to a given combination of input variables. A simulation metamodel is easier to manage and provides more insights than simulation alone. Traditionally, the multiple regression analysis is utilized to develop the metamodel from a set of simulation experiments. Simulation can consequentially benefit from the metamodelling in post-simulation analysis. A backpropagation (BP) neural network is a proven tool in providing excellent response predictions in many application areas and it outperforms regression analysis for a wide array of applications. In this paper, a BP neural network is used to generate metamodels for simulated manufacturing systems. For the purpose of optimal manufacturing systems design, mathematical models can be formulated by using the mapping functions generated from the neural network metamodels. The optimization model is then solved by a stochastic local search approach, simulated annealing (SA), to obtain an optimal configuration with respect to the objective of the systems design. Instead of triggering the detailed simulation programs, the SA-based optimization procedure evaluates the simulation outputs by the neural network metamodels. By using the SA-based optimization algorithm, the solution space of the studied problem is extensively exploited to escape the entrapment of local optima while the number of time consuming simulation runs is reduced. The proposed methodology is illustrated to be both effective and efficient in solving a manufacturing systems design problem through an example.     

Utilizing support vector machine in real-time crash risk evaluation

Real-time crash risk evaluation models will likely play a key role in Active Traffic Management (ATM). Models have been developed to predict crash occurrence in order to proactively improve traffic safety. Previous real-time crash risk evaluation studies mainly employed logistic regression and neural network models which have a linear functional form and over-fitting drawbacks, respectively. Moreover, these studies mostly focused on estimating the models but barely investigated the models’ predictive abilities. In this study, support vector machine (SVM), a recently proposed statistical learning model was introduced to evaluate real-time crash risk. The data has been split into a training dataset (used for developing the models) and scoring datasets (meant for assessing the models’ predictive power). Classification and regression tree (CART) model has been developed to select the most important explanatory variables and based on the results, three candidates Bayesian logistic regression models have been estimated with accounting for different levels unobserved heterogeneity. Then SVM models with different kernel functions have been developed and compared to the Bayesian logistic regression model. Model comparisons based on areas under the ROC curve (AUC) demonstrated that the SVM model with Radial-basis kernel function outperformed the others. Moreover, several extension analyses have been conducted to evaluate the effect of sample size on SVM models’ predictive capability; the importance of variable selection before developing SVM models; and the effect of the explanatory variables in the SVM models. Results indicate that (1) smaller sample size would enhance the SVM model's classification accuracy, (2) variable selection procedure is needed prior to the SVM model estimation, and (3) explanatory variables have identical effects on crash occurrence for the SVM models and logistic regression models.     

Modeling and optimization of spring-back in bending process using multiple regression analysis and neural computation

The current work involves both modeling and optimization approaches to achieve minimum spring-back in V-die bending process of heat treated CK67 sheets. Number of 36 experimental tests have been conducted with various levels of sheet orientation, punch tip radius and sheet thickness. Firstly, various predictive models based on statistical analysis, back-propagation neural network (BPNN), counter propagation neural network (CPNN) and radial basis function network (RBFNN) have been developed using experimental observations. Then the accuracy of the developed models has been compared based on values of mean absolute error (MAE), and root mean square error (RMSE). Secondly, the model with lowest values of MAE, and RMSE has been applied as objective function for optimization of process using imperialist competitive algorithm (ICA). After selection of optimal bending parameters, a confirmation test has been conducted to prove the optimal solutions. Results indicated that the radial basis network fulfills precise prediction of process rather than the other developed models. Also, confirmation tests proved that both RBFNN and ICA could predict and optimize the process vigorously.     

Data mining techniques applied to predictive modeling of the knurling process

Knurls are designed into a product to provide the correct frictional force for easy assembly and maintenance and sometimes for decorative purposes. The literature to date has merely studied how to realize a good and consistent knurl, but no predictive models of the knurling process have been presented. This paper applies two competing data mining techniques, regression analysis and artificial neural networks, to develop a predictive model of the knurling process. Fractional factorial design of experiments is used to plan the experiments. Four criteria, namely the PRESS statistic, the adjusted R², the C_p statistic, and the residual mean square s², are employed to select the best regression model. Hypothesis testing is conducted to test the effectiveness of each model, and to compare the two data mining schemes. This study demonstrates that for a reasonably large set of data from structurally designed experiments, the two methods produce comparable results in both model construction (or training) and model validation. Due to the explicit nature of a regression model, it is preferred to a neural network model to investigate the process.     

Modelling and optimization of turning processes for slender parts

In this paper, an abductive network is adopted in order to construct a prediction model for surface roughness and error-of-roundness in the turning operation of slender parts. The abductive network is composed of a number of functional nodes. These functional nodes are self-organized to form an optimal network architecture by using a predicted square error (PSE) criterion. Once the process parameters (workpiece length L, spindle speed n, feed rate f and depth of cut t) are given, the surface roughness and error-of-roundness can be predicted by this developed network. To verify the feasibility of the abductive network, regression analysis has been adopted to develop a second prediction model for surface roughness and error-of-roundness. Comparison of the two models indicates that the prediction model developed by the abductive network is more accurate than regression analysis. It can be found that the use of the abductive network for surface roughness and error-of-roundness is feasible. A simulated annealing optimization algorithm with a performance index is then applied to the developed network for searching the optimal process parameters. The optimal cutting condition can be obtained with the object of maximizing the metal removal rate and minimizing the surface roughness and error-of-roundness to the lowest/smallest extent permissible.     

喷墨打印纸印刷适性预测模型

目的研究广义回归神经网络对喷墨打印质量进行预测的可能性。方法测试不同喷墨打印纸的定量、平滑度、白度、光泽度、粗糙度等印刷适性,在相同条件下打印后测量印刷品色度,利用广义回归神经网络结合印刷品色度指标与喷墨打印纸印刷适性指标,并建立预测模型。结果基于广义回归神经网络的预测模型预测得到印刷品最小色差达到4.7215,最大色差达到4.8638。结论该模型可以定量描述喷墨打印纸印刷适性对印刷品色差的影响,为选纸提供试验及理论依据。     

Analysis of model building techniques for the development of machinability database systems

Computerized machinability database systems are essential for the selection of optimum cutting conditions during process planning, and these form an important component in the implementation of computer integrated manufacturing (CIM) systems. This paper presents a comparative analysis of the different model building techniques available in commercial statistical packages, to select the most suitable technique for use in machinability database systems. The techniques analysed are; backward elimination, forward selection, stepwise regression, and all possible subset regression. Experimental machining response data for the surface roughness when turning grey cast iron (154 BHN) and steel (140 BHN) have been analysed to evaluate the four model building techniques. Second order polynomial model structures with logarithmic transformation of the variables has been used for building adequate models. The adequacy of the fitted regression equation is checked by analysing the residuals. Based on the results of this analysis, the advantages and limitations of the four different techniques for building machinability models are discussed. It has been found that the backward elimination and all possible subset regression are the best suited techniques for model building in machining database systems.     

A yield forecast model for pilot products using support vector regression and manufacturing experience–the case of large-size polariser

To build up a manufacturing management model for a newly developed product is fundamentally a difficult problem, because the collected data in the early manufacturing stages is usually insufficient when data size is small. There are several researches on this topic, and most of them focus on the original data analysis such as building up virtual samples to increase the data number. As to other approaches, the usage of old or similar manufacturing experience may be an alternative approach to help in modelling a small data set, by taking advantage of the fact that the new product's manufacturing process could be based on the experience of the old one. This research proposes a combination of support vector regression (SVR) and the manufacturing experience to build up the manufacturing knowledge model for a new product. A real-problem of a new product yield forecast model in a polariser manufacturing company is demonstrated, where two approaches are proposed, and the results show that the presented approach is superior to the performance of a linear regression and back-propagation neural network. The case study shows that the input of the old or similar manufacturing experience into the forecast model can reduce the error rate and enhance the model forecasting ability.     

Convolutional and generative adversarial neural networks in manufacturing

Manufacturing is undergoing transformation driven by the developments in process technology, information technology, and data science. A future manufacturing enterprise will be highly digital. This will create opportunities for machine learning algorithms to generate predictive models across the enterprise in the spirit of the digital twin concept. Convolutional and generative adversarial neural networks have received some attention of the manufacturing research community. Representative research and applications of the two machine learning concepts in manufacturing are presented. Advantages and limitations of each neural network are discussed. The paper might be helpful in identifying research gaps, inspire machine learning research in new manufacturing domains, contribute to the development of successful neural network architectures, and getting deeper insights into the manufacturing data.     

Measuring the predictive accuracy of various models of formability of Corus Tubular Blanks

This paper assesses the predictive accuracy of various analytical models and one numerical model (a CART-ANFIS network) of springback that are available with the existing literature using the mean square error and its decomposition into systematic and random components as a comparative measure of predictive accuracy. The numerical model was found to have no systematic bias in the springback predictions made, whilst for the analytical models the systematic bias accounted for about 11% of the mean square error. The CART-ANFIS network also had the smallest MSE and the prediction errors made were all random in nature. The paper ends by giving some illustrations of the CART-ANFIS numerical model in finding the proper die contour to correct for springback so as to achieve right first-time manufacturing for a wide range of sheet steels.     

基于机器学习的电喷印精度预测方法研究

目的 节省电流体喷射打印精度预测的时间和解决电流体工艺参数的选择问题,达到提高电流体打印的质量和效率的目的。方法 为了对电流体喷射打印精度进行预测,提出有限元模型与机器学习相结合的方法。基于线性回归、支持向量回归和神经网络等机器学习算法建立4种参数与射流直径的关系模型。结果 算法结果表明:支持向量回归和神经网络预测模型的决定系数R²能达到0.9以上,表示模型可信度高;支持向量回归和神经网络预测模型指标都比线性回归预测模型的小。结论 机器学习算法可对电喷印打印精度进行有效预测,预测效率提高了十几倍,节省了精度预测的时间。     

基于神经网络的电弧增材制造铝合金力学性能预测

目的 预测不同工艺参数下电弧增材制造铝合金的力学性能。方法 通过实验建立了电弧增材制造6061铝合金及TiC增强6061铝合金力学性能的数据集,并建立了一种以焊接电流、焊接速度、脉冲频率、TiC颗粒含量为输入,以屈服强度和抗拉强度为输出的神经网预测模型,对比了反向传播神经网络(BP)、粒子群算法优化BP神经网络(PSO-BP)、遗传算法优化BP神经网络(GA-BP)3种预测模型的精度。结果 与BP模型和PSO-BP模型相比,GA-BP预测模型具有更好的预测精度。其中,GA-BP模型预测6061铝合金屈服强度最佳结果的相关系数(R)为0.965,决定系数(R²)为0.93,平均绝对误差(Mean Absolute Error,MAE)为2.35,均方根误差(Root Mean Square Error,RMSE)为2.67;预测TiC增强的6061铝合金抗拉强度最佳结果的R=1,R²高达0.99,MAE为0.46,RMSE为0.49,GA-BP具有良好的预测精度。结论 BP、PSO-BP、GA-BP 3种神经网络模型可以用来预测电弧增材制造铝合金的力学性能,GA-BP模型比其他2种模型的预测精度更优。与传统的实验方法相比,用神经网络模型预测电弧增材制造铝合金力学性能的速度更快,成本更低。     

基于BP神经网络的个性化多功能茶几设计研究

目的 通过分析消费者感性需求和多功能茶几产品形态设计要素,建立二者之间回归的联系模型,完成多功能茶几产品的个性化设计,解决茶几产品无法按照用户消费需求设计制造的难题。方法 首先运用语义差异法获取消费者对茶几产品的感性意象评价值,并利用因子分析法对评价值进行归纳整理,其次按照茶几产品设计要素对其进行模块解构,并对各部分模块进行数值化编码,再次根据整理的感性意象评价值和模块数值训练茶几产品BP神经网络,建立二者间映射关系,最后实施二次语义差异法问卷实验,验证BP神经网络的准确性。结果 根据训练的茶几产品BP神经网络可预测出感性评价值最大的茶几产品造型,实验结果验证了茶几产品BP神经网络模型的准确性,为茶几产品的个性化设计提供了有利的支撑。结论 此方法提高了茶几产品的设计效率,提升了茶几产品设计的合理性,解决了家具设计者无法精准按照用户主观需求完成客观产品设计的难题,为以消费者需求市场为导向的产品设计制造提供了有益的参考和指导。