Prediction of primary quality variables in real time with adaptation capability for varying process conditions is a critical task in process industries.This article focuses on the development of non-linear adaptive so...Prediction of primary quality variables in real time with adaptation capability for varying process conditions is a critical task in process industries.This article focuses on the development of non-linear adaptive soft sensors for prediction of naphtha initial boiling point(IBP)and end boiling point(EBP)in crude distillation unit.In this work,adaptive inferential sensors with linear and non-linear local models are reported based on recursive just in time learning(JITL)approach.The different types of local models designed are locally weighted regression(LWR),multiple linear regression(MLR),partial least squares regression(PLS)and support vector regression(SVR).In addition to model development,the effect of relevant dataset size on model prediction accuracy and model computation time is also investigated.Results show that the JITL model based on support vector regression with iterative single data algorithm optimization(ISDA)local model(JITL-SVR:ISDA)yielded best prediction accuracy in reasonable computation time.展开更多
In this paper we apply the nonlinear time series analysis method to small-time scale traffic measurement data. The prediction-based method is used to determine the embedding dimension of the traffic data. Based on the...In this paper we apply the nonlinear time series analysis method to small-time scale traffic measurement data. The prediction-based method is used to determine the embedding dimension of the traffic data. Based on the reconstructed phase space, the local support vector machine prediction method is used to predict the traffic measurement data, and the BIC-based neighbouring point selection method is used to choose the number of the nearest neighbouring points for the local support vector machine regression model. The experimental results show that the local support vector machine prediction method whose neighbouring points are optimized can effectively predict the small-time scale traffic measurement data and can reproduce the statistical features of real traffic measurements.展开更多
Aiming at reducing the deficiency of the traditional fire pre-warning algorithms and the intelligent fire pre-warning algorithms such as artificial neural network,and then to improve the accuracy of fire prewarning fo...Aiming at reducing the deficiency of the traditional fire pre-warning algorithms and the intelligent fire pre-warning algorithms such as artificial neural network,and then to improve the accuracy of fire prewarning for high-rise buildings,a composite fire pre-warning controller is designed according to the characteristic（ nonlinear,less historical data,many influence factors）,also a high-rise building fire pre-warning model is set up based on the support vector regression（ SV R）. Then the wood fire standard history data is applied to make empirical analysis. The research results can provide a reliable decision support framework for high-rise building fire pre-warning.展开更多
Model updating for aircraft in a high temperature environment(HTE)is proposed based on the hierarchical method.With this method,the problem can be decomposed into temperature field updating and dynamic structural upda...Model updating for aircraft in a high temperature environment(HTE)is proposed based on the hierarchical method.With this method,the problem can be decomposed into temperature field updating and dynamic structural updating.In order to improve the estimation accuracy,the model updating problem is turned into a multi-objective optimization problem by constructing the objective function which combined with residues of modal frequency and effective modal mass.Then the metamodeling,support vector regression(SVR)is introduced to improve the optimization efficiency,and the solution can be determined by adaptive weighted-sum method(AWS).Finally,the proposed method is tested on a finite element(FE)model of a reentry vehicle model.The results show that the multi-objective model updating method in HTE can identify the input parameters of the temperature field and structure with good accuracy.展开更多
目的采用4种机器学习算法分别构建结直肠癌患者术前营养不良的临床风险预测模型,探讨其预测价值。方法回顾性收集2023年1月—2024年5月在新疆医科大学附属肿瘤医院胃肠外科就诊的412例结直肠癌患者的术前资料;按7∶3的比例随机分为训练...目的采用4种机器学习算法分别构建结直肠癌患者术前营养不良的临床风险预测模型,探讨其预测价值。方法回顾性收集2023年1月—2024年5月在新疆医科大学附属肿瘤医院胃肠外科就诊的412例结直肠癌患者的术前资料;按7∶3的比例随机分为训练集(n=288)和验证集(n=124),采用单因素分析及二元logistic回归分析筛选出术前营养不良的预测因子;基于逻辑回归(LR)、支持向量机(SVM)、轻量级梯度提升(LightGBM)、多层感知机(MLP)4种机器学习算法分别构建结直肠癌患者术前营养不良风险预测模型,绘制ROC曲线评价4种算法模型预测效能,通过Delong检验比较4种模型的AUC差异。选择最优算法模型,采用校准曲线和临床决策曲线(DCA曲线)进行验证。结果(1)结直肠癌患者术前营养不良发生率为33.7%,年龄、Braden评分是其独立危险因素;(2)训练集中LightGBM算法模型预测结直肠癌患者术前发生营养不良的AUC高于LR、SVM、MLP算法模型(0.941 VS 0.874、0.830、0.831);(3)ROC曲线结果提示,LightGBM算法模型验证集中预测结直肠癌患者术前发生营养不良的AUC为0.926(95%CI:0.882~0.969);校准曲线显示,LightGBM算法模型预测结直肠癌患者术前发生营养不良的曲线与实际发生营养不良一致性良好;DCA曲线结果显示,LightGBM算法模型在阈值概率区间为0.16~0.79可以提供显著临床净收益。结论基于LightGBM算法构建的临床预测模型在预测结直肠癌患者术前发生营养不良中有较高价值,可以为临床人员实施营养管理提供参考。展开更多
Aiming at the problem that the traditional Unscented Kalman Filtering(UKF) algorithm can't solve the problem that the measurement covariance matrix is unknown and the measured value contains outliers,this paper pr...Aiming at the problem that the traditional Unscented Kalman Filtering(UKF) algorithm can't solve the problem that the measurement covariance matrix is unknown and the measured value contains outliers,this paper proposes a robust adaptive UKF algorithm based on Support Vector Regression(SVR).The algorithm combines the advantages of support vector regression with small samples,nonlinear learning ability and online estimation capability of adaptive algorithm based on innovation.Firstly,the SVR model is trained by using the innovation in the sliding window,and the new innovation is monitored.If the deviation between the estimated innovation and the measured innovation exceeds a given threshold,then measured innovation will be replaced by the predicted innovation,and then the processed innovation is used to calculate the measurement noise covariance matrix using the adaptive estimation algorithm.Simulation experiments and measured data experiments show that SVRUKF is significantly better than the traditional UKF,robust UKF and adaptive UKF algorithms for the case where the covariance matrix is unknown and the measured values have outliers.展开更多
文摘Prediction of primary quality variables in real time with adaptation capability for varying process conditions is a critical task in process industries.This article focuses on the development of non-linear adaptive soft sensors for prediction of naphtha initial boiling point(IBP)and end boiling point(EBP)in crude distillation unit.In this work,adaptive inferential sensors with linear and non-linear local models are reported based on recursive just in time learning(JITL)approach.The different types of local models designed are locally weighted regression(LWR),multiple linear regression(MLR),partial least squares regression(PLS)and support vector regression(SVR).In addition to model development,the effect of relevant dataset size on model prediction accuracy and model computation time is also investigated.Results show that the JITL model based on support vector regression with iterative single data algorithm optimization(ISDA)local model(JITL-SVR:ISDA)yielded best prediction accuracy in reasonable computation time.
基金Project supported by the National Natural Science Foundation of China (Grant No 60573065)the Natural Science Foundation of Shandong Province,China (Grant No Y2007G33)the Key Subject Research Foundation of Shandong Province,China(Grant No XTD0708)
文摘In this paper we apply the nonlinear time series analysis method to small-time scale traffic measurement data. The prediction-based method is used to determine the embedding dimension of the traffic data. Based on the reconstructed phase space, the local support vector machine prediction method is used to predict the traffic measurement data, and the BIC-based neighbouring point selection method is used to choose the number of the nearest neighbouring points for the local support vector machine regression model. The experimental results show that the local support vector machine prediction method whose neighbouring points are optimized can effectively predict the small-time scale traffic measurement data and can reproduce the statistical features of real traffic measurements.
基金Supported by the National Natural Science Foundation of China(11072035)
文摘Aiming at reducing the deficiency of the traditional fire pre-warning algorithms and the intelligent fire pre-warning algorithms such as artificial neural network,and then to improve the accuracy of fire prewarning for high-rise buildings,a composite fire pre-warning controller is designed according to the characteristic（ nonlinear,less historical data,many influence factors）,also a high-rise building fire pre-warning model is set up based on the support vector regression（ SV R）. Then the wood fire standard history data is applied to make empirical analysis. The research results can provide a reliable decision support framework for high-rise building fire pre-warning.
基金supported by the National Natural Science Foundation of China(No.11472132)the Fundamental Research Funds for Central University (No. NJ20160050)the Fundamental Research Funds for Central University(No.NJ2016098)
文摘Model updating for aircraft in a high temperature environment(HTE)is proposed based on the hierarchical method.With this method,the problem can be decomposed into temperature field updating and dynamic structural updating.In order to improve the estimation accuracy,the model updating problem is turned into a multi-objective optimization problem by constructing the objective function which combined with residues of modal frequency and effective modal mass.Then the metamodeling,support vector regression(SVR)is introduced to improve the optimization efficiency,and the solution can be determined by adaptive weighted-sum method(AWS).Finally,the proposed method is tested on a finite element(FE)model of a reentry vehicle model.The results show that the multi-objective model updating method in HTE can identify the input parameters of the temperature field and structure with good accuracy.
文摘目的采用4种机器学习算法分别构建结直肠癌患者术前营养不良的临床风险预测模型,探讨其预测价值。方法回顾性收集2023年1月—2024年5月在新疆医科大学附属肿瘤医院胃肠外科就诊的412例结直肠癌患者的术前资料;按7∶3的比例随机分为训练集(n=288)和验证集(n=124),采用单因素分析及二元logistic回归分析筛选出术前营养不良的预测因子;基于逻辑回归(LR)、支持向量机(SVM)、轻量级梯度提升(LightGBM)、多层感知机(MLP)4种机器学习算法分别构建结直肠癌患者术前营养不良风险预测模型,绘制ROC曲线评价4种算法模型预测效能,通过Delong检验比较4种模型的AUC差异。选择最优算法模型,采用校准曲线和临床决策曲线(DCA曲线)进行验证。结果(1)结直肠癌患者术前营养不良发生率为33.7%,年龄、Braden评分是其独立危险因素;(2)训练集中LightGBM算法模型预测结直肠癌患者术前发生营养不良的AUC高于LR、SVM、MLP算法模型(0.941 VS 0.874、0.830、0.831);(3)ROC曲线结果提示,LightGBM算法模型验证集中预测结直肠癌患者术前发生营养不良的AUC为0.926(95%CI:0.882~0.969);校准曲线显示,LightGBM算法模型预测结直肠癌患者术前发生营养不良的曲线与实际发生营养不良一致性良好;DCA曲线结果显示,LightGBM算法模型在阈值概率区间为0.16~0.79可以提供显著临床净收益。结论基于LightGBM算法构建的临床预测模型在预测结直肠癌患者术前发生营养不良中有较高价值,可以为临床人员实施营养管理提供参考。
文摘Aiming at the problem that the traditional Unscented Kalman Filtering(UKF) algorithm can't solve the problem that the measurement covariance matrix is unknown and the measured value contains outliers,this paper proposes a robust adaptive UKF algorithm based on Support Vector Regression(SVR).The algorithm combines the advantages of support vector regression with small samples,nonlinear learning ability and online estimation capability of adaptive algorithm based on innovation.Firstly,the SVR model is trained by using the innovation in the sliding window,and the new innovation is monitored.If the deviation between the estimated innovation and the measured innovation exceeds a given threshold,then measured innovation will be replaced by the predicted innovation,and then the processed innovation is used to calculate the measurement noise covariance matrix using the adaptive estimation algorithm.Simulation experiments and measured data experiments show that SVRUKF is significantly better than the traditional UKF,robust UKF and adaptive UKF algorithms for the case where the covariance matrix is unknown and the measured values have outliers.
