以云南省天星站和坡脚站10、20、40 cm 3个土层的土壤含水量观测数据为基础,通过改进时变滤波经验模态分解(TVFEMD)和快速学习网(FLN)方法构建基于多种优化算法的预测模型(TVFEMD-BSLO/AO/IVYA/EGO/PSO-FLN),提升土壤含水量时间序列预...以云南省天星站和坡脚站10、20、40 cm 3个土层的土壤含水量观测数据为基础,通过改进时变滤波经验模态分解(TVFEMD)和快速学习网(FLN)方法构建基于多种优化算法的预测模型(TVFEMD-BSLO/AO/IVYA/EGO/PSO-FLN),提升土壤含水量时间序列预测精度。通过比较各优化算法的模型性能,为土壤水分预测提供更优的建模方法。结果表明,TVFEMD分解效果主要受带宽阈值和B样条阶数2个关键参数影响。采用IVYA算法优化这2个参数可提升时间序列分解质量,进而改善模型预测性能。TVFEMD-BLSO/AO/IVYA/EGO-FLN模型在训练集上表现出卓越的预测性能,其平均绝对百分比误差(MAPE)为0.002%~0.077%,决定系数(R^(2))为0.9997~1.0000;预测集中的MAPE为0.006%~0.459%,R^(2)为0.9966~1.0000。与TVFEMD-PSO-FLN模型相比,TVFEMD-BLSO/AO/IVYA/EGO-FLN模型在拟合性能和预测精度方面均有明显提升。采用BLSO、AO、IVYA和EGO算法优化FLN超参数可有效提升模型性能,其中IVYA算法的优化效果较突出。展开更多
Several methods for evaluating the sublayer suspension beneath old pavement with falling weight deflectormeter(FWD), were summarized and the respective advantages and disadvantages were analyzed. Based on these method...Several methods for evaluating the sublayer suspension beneath old pavement with falling weight deflectormeter(FWD), were summarized and the respective advantages and disadvantages were analyzed. Based on these methods, the evaluation principles were improved and a new type of the neural network, functional-link neural network was proposed to evaluate the sublayer suspension with FWD test results. The concept of function link, learning method of functional-link neural network and the establishment process of neural network model were studied in detail. Based on the old pavement over-repairing engineering of Kaiping section, Guangdong Province in G325 National Highway, the application of functional-link neural network in evaluation of sublayer suspension beneath old pavement based on FWD test data on the spot was investigated. When learning rate is 0.1 and training cycles are 405, the functional-link network error is less than 0.000 1, while the optimum chosen 4-8-1 BP needs over 10 000 training cycles to reach the same accuracy with less precise evaluation results. Therefore, in contrast to common BP neural network,the functional-link neural network adopts single layer structure to learn and calculate, which simplifies the network, accelerates the convergence speed and improves the accuracy. Moreover the trained functional-link neural network can be (adopted) to directly evaluate the sublayer suspension based on FWD test data on the site. Engineering practice indicates that the functional-link neural model gains very excellent results and effectively guides the pavement over-repairing construction.展开更多
Molten iron temperature as well as Si, P, and S contents is the most essential molten iron quality （MIQ） indices in the blast furnace （BF） ironmaking, which requires strict monitoring during the whole ironmaking p...Molten iron temperature as well as Si, P, and S contents is the most essential molten iron quality （MIQ） indices in the blast furnace （BF） ironmaking, which requires strict monitoring during the whole ironmaking production. However, these MIQ parameters are difficult to be directly measured online, and large-time delay exists in off-line analysis through laboratory sampling. Focusing on the practical challenge, a data-driven modeling method was presented for the prediction of MIQ using the improved muhivariable incremental random vector functional-link net- works （M-I-RVFLNs）. Compared with the conventional random vector functional-link networks （RVFLNs） and the online sequential RVFLNs, the M-I-RVFLNs have solved the problem of deciding the optimal number of hidden nodes and overcome the overfitting problems. Moreover, the proposed M I RVFLNs model has exhibited the potential for multivariable prediction of the MIQ and improved the terminal condition for the multiple-input multiple-out- put （MIMO） dynamic system, which is suitable for the BF ironmaking process in practice. Ultimately, industrial experiments and contrastive researches have been conducted on the BF No. 2 in Liuzhou Iron and Steel Group Co. Ltd. of China using the proposed method, and the results demonstrate that the established model produces better estima ting accuracy than other MIQ modeling methods.展开更多
In the contemporary era, the proliferation of information technology has led to an unprecedented surge in data generation, with this data being dispersed across a multitude of mobile devices. Facing these situations a...In the contemporary era, the proliferation of information technology has led to an unprecedented surge in data generation, with this data being dispersed across a multitude of mobile devices. Facing these situations and the training of deep learning model that needs great computing power support, the distributed algorithm that can carry out multi-party joint modeling has attracted everyone’s attention. The distributed training mode relieves the huge pressure of centralized model on computer computing power and communication. However, most distributed algorithms currently work in a master-slave mode, often including a central server for coordination, which to some extent will cause communication pressure, data leakage, privacy violations and other issues. To solve these problems, a decentralized fully distributed algorithm based on deep random weight neural network is proposed. The algorithm decomposes the original objective function into several sub-problems under consistency constraints, combines the decentralized average consensus (DAC) and alternating direction method of multipliers (ADMM), and achieves the goal of joint modeling and training through local calculation and communication of each node. Finally, we compare the proposed decentralized algorithm with several centralized deep neural networks with random weights, and experimental results demonstrate the effectiveness of the proposed algorithm.展开更多
文摘以云南省天星站和坡脚站10、20、40 cm 3个土层的土壤含水量观测数据为基础,通过改进时变滤波经验模态分解(TVFEMD)和快速学习网(FLN)方法构建基于多种优化算法的预测模型(TVFEMD-BSLO/AO/IVYA/EGO/PSO-FLN),提升土壤含水量时间序列预测精度。通过比较各优化算法的模型性能,为土壤水分预测提供更优的建模方法。结果表明,TVFEMD分解效果主要受带宽阈值和B样条阶数2个关键参数影响。采用IVYA算法优化这2个参数可提升时间序列分解质量,进而改善模型预测性能。TVFEMD-BLSO/AO/IVYA/EGO-FLN模型在训练集上表现出卓越的预测性能,其平均绝对百分比误差(MAPE)为0.002%~0.077%,决定系数(R^(2))为0.9997~1.0000;预测集中的MAPE为0.006%~0.459%,R^(2)为0.9966~1.0000。与TVFEMD-PSO-FLN模型相比,TVFEMD-BLSO/AO/IVYA/EGO-FLN模型在拟合性能和预测精度方面均有明显提升。采用BLSO、AO、IVYA和EGO算法优化FLN超参数可有效提升模型性能,其中IVYA算法的优化效果较突出。
文摘Several methods for evaluating the sublayer suspension beneath old pavement with falling weight deflectormeter(FWD), were summarized and the respective advantages and disadvantages were analyzed. Based on these methods, the evaluation principles were improved and a new type of the neural network, functional-link neural network was proposed to evaluate the sublayer suspension with FWD test results. The concept of function link, learning method of functional-link neural network and the establishment process of neural network model were studied in detail. Based on the old pavement over-repairing engineering of Kaiping section, Guangdong Province in G325 National Highway, the application of functional-link neural network in evaluation of sublayer suspension beneath old pavement based on FWD test data on the spot was investigated. When learning rate is 0.1 and training cycles are 405, the functional-link network error is less than 0.000 1, while the optimum chosen 4-8-1 BP needs over 10 000 training cycles to reach the same accuracy with less precise evaluation results. Therefore, in contrast to common BP neural network,the functional-link neural network adopts single layer structure to learn and calculate, which simplifies the network, accelerates the convergence speed and improves the accuracy. Moreover the trained functional-link neural network can be (adopted) to directly evaluate the sublayer suspension based on FWD test data on the site. Engineering practice indicates that the functional-link neural model gains very excellent results and effectively guides the pavement over-repairing construction.
基金Item Sponsored by National Natural Science Foundation of China(61290323,61333007,61473064)Fundamental Research Funds for Central Universities of China(N130108001)+1 种基金National High Technology Research and Development Program of China(2015AA043802)General Project on Scientific Research for Education Department of Liaoning Province of China(L20150186)
文摘Molten iron temperature as well as Si, P, and S contents is the most essential molten iron quality （MIQ） indices in the blast furnace （BF） ironmaking, which requires strict monitoring during the whole ironmaking production. However, these MIQ parameters are difficult to be directly measured online, and large-time delay exists in off-line analysis through laboratory sampling. Focusing on the practical challenge, a data-driven modeling method was presented for the prediction of MIQ using the improved muhivariable incremental random vector functional-link net- works （M-I-RVFLNs）. Compared with the conventional random vector functional-link networks （RVFLNs） and the online sequential RVFLNs, the M-I-RVFLNs have solved the problem of deciding the optimal number of hidden nodes and overcome the overfitting problems. Moreover, the proposed M I RVFLNs model has exhibited the potential for multivariable prediction of the MIQ and improved the terminal condition for the multiple-input multiple-out- put （MIMO） dynamic system, which is suitable for the BF ironmaking process in practice. Ultimately, industrial experiments and contrastive researches have been conducted on the BF No. 2 in Liuzhou Iron and Steel Group Co. Ltd. of China using the proposed method, and the results demonstrate that the established model produces better estima ting accuracy than other MIQ modeling methods.
文摘In the contemporary era, the proliferation of information technology has led to an unprecedented surge in data generation, with this data being dispersed across a multitude of mobile devices. Facing these situations and the training of deep learning model that needs great computing power support, the distributed algorithm that can carry out multi-party joint modeling has attracted everyone’s attention. The distributed training mode relieves the huge pressure of centralized model on computer computing power and communication. However, most distributed algorithms currently work in a master-slave mode, often including a central server for coordination, which to some extent will cause communication pressure, data leakage, privacy violations and other issues. To solve these problems, a decentralized fully distributed algorithm based on deep random weight neural network is proposed. The algorithm decomposes the original objective function into several sub-problems under consistency constraints, combines the decentralized average consensus (DAC) and alternating direction method of multipliers (ADMM), and achieves the goal of joint modeling and training through local calculation and communication of each node. Finally, we compare the proposed decentralized algorithm with several centralized deep neural networks with random weights, and experimental results demonstrate the effectiveness of the proposed algorithm.
