本文關(guān)鍵詞：前饋神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)設(shè)計(jì)研究及其復(fù)雜化工過(guò)程建模應(yīng)用　出處：《北京化工大學(xué)》2016年博士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 過(guò)程建模 前饋神經(jīng)網(wǎng)絡(luò) 數(shù)據(jù)驅(qū)動(dòng) 結(jié)構(gòu)設(shè)計(jì) 復(fù)雜化工過(guò)程

【摘要】：流程工業(yè)的飛快發(fā)展使得該行業(yè)呈現(xiàn)出一些特點(diǎn)：生產(chǎn)過(guò)程高非線性、相關(guān)變量多維性、工藝高度復(fù)雜和流程高度綜合等,因此導(dǎo)致采用第一原理即機(jī)理分析的方法用于過(guò)程建模越來(lái)越棘手。近些年,先進(jìn)的傳感技術(shù)使得過(guò)程數(shù)據(jù)越來(lái)越容易被采集和存儲(chǔ),這些過(guò)程數(shù)據(jù)背后蘊(yùn)含著重要的過(guò)程知識(shí),所以基于數(shù)據(jù)驅(qū)動(dòng)策略的方法在解決復(fù)雜流程工業(yè)過(guò)程建模的問(wèn)題中發(fā)揮著越來(lái)越重要的作用。眾多數(shù)據(jù)驅(qū)動(dòng)的方法中,神經(jīng)網(wǎng)絡(luò)技術(shù)由于其具有學(xué)習(xí)、并行計(jì)算以及強(qiáng)非線性映射的能力已被廣泛成功地應(yīng)用到建模、控制、優(yōu)化等諸多領(lǐng)域。前饋神經(jīng)網(wǎng)絡(luò)因其結(jié)構(gòu)簡(jiǎn)單和算法易用等特點(diǎn)受到越來(lái)越多的關(guān)注。然而傳統(tǒng)前饋網(wǎng)絡(luò)模型并不能夠滿足復(fù)雜流程工業(yè)過(guò)程建模的要求,由此研究建立性能優(yōu)越的前饋網(wǎng)絡(luò)模型對(duì)豐富神經(jīng)網(wǎng)絡(luò)模型和進(jìn)一步推動(dòng)神經(jīng)網(wǎng)絡(luò)技術(shù)應(yīng)用于復(fù)雜流程工業(yè)過(guò)程建模具有重大的意義。本文主要從遞階結(jié)構(gòu)設(shè)計(jì)和雙并行結(jié)構(gòu)設(shè)計(jì)對(duì)遞階神經(jīng)網(wǎng)絡(luò)和極限學(xué)習(xí)機(jī)兩種前饋網(wǎng)絡(luò)展開(kāi)研究,最終將其應(yīng)用于復(fù)雜化工過(guò)程建模。遞階神經(jīng)網(wǎng)絡(luò)是一種善于處理高維數(shù)據(jù)的網(wǎng)絡(luò)模型,然而其子網(wǎng)結(jié)構(gòu)設(shè)計(jì)一直是個(gè)難點(diǎn)。極限學(xué)習(xí)機(jī)為近些年來(lái)機(jī)器學(xué)習(xí)領(lǐng)域的研究的熱點(diǎn)之一,該模型具有快速的學(xué)習(xí)速度和較好的泛化性能。然而面對(duì)帶有噪聲、共線性等特點(diǎn)的過(guò)程數(shù)據(jù),極限學(xué)習(xí)機(jī)模型仍存在一些問(wèn)題：1、噪聲處理性能低；2、傳統(tǒng)三層結(jié)構(gòu)限制模型性能;3、共線性數(shù)據(jù)對(duì)性能影響大。本文逐一解決上述問(wèn)題,旨在為復(fù)雜化工過(guò)程建模特定問(wèn)題下提供可靠的網(wǎng)絡(luò)模型,最終取得一些研究成果總結(jié)如下：(1)針對(duì)遞階神經(jīng)網(wǎng)絡(luò)子網(wǎng)設(shè)計(jì)困難的問(wèn)題,提出一種基于輸入屬性空間劃分的子網(wǎng)設(shè)計(jì)方法,進(jìn)而建立基于輸入屬性空間劃分的遞階神經(jīng)網(wǎng)絡(luò),為多參數(shù)輸入的復(fù)雜化工過(guò)程提供可靠的模型。該子網(wǎng)設(shè)計(jì)方法避免繁瑣的專家知識(shí),首先采用先進(jìn)的可拓聚類算法對(duì)輸入屬性的高維空間進(jìn)行聚類；然后依據(jù)輸入屬性空間的聚類結(jié)果確定遞階神經(jīng)網(wǎng)絡(luò)的子網(wǎng)個(gè)數(shù)；最后依據(jù)每個(gè)子屬性空間的輸入屬性確定子網(wǎng)的輸入。該設(shè)計(jì)方法能夠同時(shí)解決子網(wǎng)數(shù)目確定以及子網(wǎng)輸入屬性選取的兩個(gè)難題,從而提供一種簡(jiǎn)單有效設(shè)計(jì)遞階神經(jīng)網(wǎng)絡(luò)子網(wǎng)的方法。(2)針對(duì)極限學(xué)習(xí)機(jī)處理噪聲性能低的問(wèn)題,提出一種具有遞階結(jié)構(gòu)的極限學(xué)習(xí)機(jī)模型。所提出的遞階極限學(xué)習(xí)機(jī)模型中,原輸入變量沒(méi)有直接作為模型的輸入,而是先輸入到自聯(lián)想濾波子網(wǎng),一方面去除噪聲,另一方面對(duì)多維輸入空間實(shí)現(xiàn)降維；隨后將自聯(lián)想濾波子網(wǎng)的隱含層輸出數(shù)據(jù)作為極限學(xué)習(xí)機(jī)的輸入,進(jìn)而有效地避免噪聲對(duì)模型精度的影響。采用帶有噪聲的工業(yè)數(shù)據(jù)對(duì)模型進(jìn)行測(cè)試,仿真結(jié)果驗(yàn)證了該模型的有效性和可行性。(3)針對(duì)極限學(xué)習(xí)機(jī)三層網(wǎng)絡(luò)結(jié)構(gòu)的限制問(wèn)題,采用基于雙并行結(jié)構(gòu)的設(shè)計(jì)方法增強(qiáng)網(wǎng)絡(luò)性能。雙并行結(jié)構(gòu)能夠較好解決極限學(xué)習(xí)機(jī)中的結(jié)構(gòu)限制問(wèn)題,但會(huì)帶來(lái)另外兩個(gè)問(wèn)題：1、增加極限學(xué)習(xí)機(jī)模型復(fù)雜性；2、增加共線性信息。為解決第一個(gè)問(wèn)題,通過(guò)研究雙并行網(wǎng)絡(luò)結(jié)構(gòu)以及皮爾遜相關(guān)系數(shù),提出一種輸入輸出皮爾遜相關(guān)系數(shù)導(dǎo)向的雙并行極限學(xué)習(xí)機(jī)模型。該模型中利用輸入輸出屬性間的相關(guān)系數(shù)將輸入屬性分為正、負(fù)兩種屬性,隨后建立正、負(fù)屬性相互獨(dú)立的雙并行結(jié)構(gòu)。工業(yè)數(shù)據(jù)仿真結(jié)果表明與傳統(tǒng)的雙并行極限學(xué)習(xí)機(jī)以及極限學(xué)習(xí)機(jī)模型相比,所提出的改進(jìn)雙并行極限學(xué)習(xí)機(jī)模型參數(shù)數(shù)目少、響應(yīng)速度快等特點(diǎn)。(4)針對(duì)極限學(xué)習(xí)機(jī)不能較好處理雙并行結(jié)構(gòu)中共線性數(shù)據(jù)的問(wèn)題,提出一種基于偏最小二乘學(xué)習(xí)的穩(wěn)健雙并行極限學(xué)習(xí)機(jī)模型。該模型采用偏最小二乘學(xué)習(xí)算法代替原來(lái)廣義逆的學(xué)習(xí)方法來(lái)求取輸出權(quán)值。偏最小二乘算法一方面能夠有效的剔除原輸入數(shù)據(jù)間以及隱含層節(jié)點(diǎn)輸出數(shù)據(jù)間的共線性信息,另一方面通過(guò)選取隱含變量數(shù)目有效避免了隱含層節(jié)點(diǎn)數(shù)目選取的難題。實(shí)驗(yàn)仿真結(jié)果表明該模型相對(duì)其他模型具有魯棒性強(qiáng)和泛化性能穩(wěn)定的特點(diǎn),為復(fù)雜化工過(guò)程建模提供可靠模型。
[Abstract]:The rapid development of process industry, the industry is showing some characteristics: the production process of highly nonlinear and multidimensional related variables and process flow is highly complex and highly integrated, thus adopting the analysis method of the first principle is the mechanism for process modeling is more and more difficult. In recent years, advanced sensor technology makes the process more easily by data acquisition and storage process behind these data contains important process knowledge, so based on data driven method in solving complex problems in the modeling process and play an increasingly important role. In many data driven methods, neural network technology has been widely applied to many fields, such as modeling, control, optimization and so on, because of its ability of learning, parallel computing and strong nonlinear mapping. Feedforward neural network has attracted more and more attention because of its simple structure and easy to use algorithm. However, the traditional feedforward network model can not satisfy the requirements of complex industrial process modeling, this study established a feedforward network model to enrich the superior performance of the neural network model and further promote the application of neural network technology has great significance in the process of complex industrial process modeling. In this paper, two feedforward networks, hierarchical neural network and extreme learning machine, are studied from the perspective of hierarchical structure design and dual parallel structure design. Finally, it is applied to modeling complex chemical processes. Hierarchical neural network is a network model that is good at processing high dimensional data. However, the design of its subnet structure is always a difficult problem. The limit learning machine is one of the hot topics in the field of machine learning in recent years. This model has fast learning speed and good generalization performance. However, in the face of the process data with noise and collinearity, there are still some problems in the extreme learning machine model: 1, the noise processing performance is low; 2, the performance of the traditional three level structure constraint model; 3, the collinear data have great influence on the performance. We solve the above problems, in order to provide a reliable model for complex chemical process modeling on specific issues, the final results are summarized as follows: (1) aiming at the difficult problem of the design of hierarchical neural network subnet, proposed a partition of the input attribute space subnet design method based on neural network, and establish a hierarchical partition of the input based on the attribute space, provide a reliable model for complex chemical process input parameters. The subnet design method of expert knowledge to avoid tedious, first uses the advanced extension clustering algorithm on the input attributes of high dimensional space for clustering; then according to the number of sub network hierarchical neural network to determine the clustering results of input attribute space; finally the subnet input is determined based on the input attributes of each sub attribute space. The design method can solve the two difficult problems of subnet number determination and subnet input attribute selection simultaneously, so as to provide a simple and effective method for designing hierarchical neural network subnet. (2) to solve the problem of low noise performance in extreme learning machine, a model of limit learning machine with hierarchical structure is proposed. The proposed hierarchical extreme learning machine model, the original input variables not directly as input of the model, but the first input to the associative filter network, a noise removal, on the other hand, dimensionality reduction of multidimensional input space; then from the hidden layer output data as extreme learning machine input subnet filtering Lenovo then, effectively avoid the influence of noise on the accuracy of the model. The model is tested by the industrial data with noise, and the simulation results verify the validity and feasibility of the model. (3) aiming at the limitation of the three layer network structure of the limit learning machine, the design method based on double parallel structure is adopted to enhance the network performance. The dual parallel structure can solve the structural constraint problem in the extreme learning machine, but it will bring two other problems: 1, increase the complexity of the extreme learning machine model; 2, increase the collinear information. To solve the first problem, by studying the double parallel network structure and Pearson correlation coefficient, we propose a dual parallel extreme learning machine model based on input and output Pearson correlation coefficient. In this model, the input attribute is divided into two attributes, positive and negative, by using the correlation coefficient between input and output attributes, and then a double parallel structure with independent positive and negative attributes is established. The industrial data simulation results show that compared with the traditional dual parallel extreme learning machine and extreme learning machine model, the proposed dual parallel extreme learning machine model has fewer parameters and faster response speed. (4) in view of the problem that the extreme learning machine can not deal with the problem of CO linear data in dual parallel structure, a robust dual parallel extreme learning machine model based on partial least squares learning is proposed. The model uses the partial least squares learning algorithm to replace the original generalized inverse learning method to obtain the output weights. On the one hand, partial least squares algorithm can effectively eliminate the collinearity information between the original input data and the hidden layer node output data. On the other hand, by selecting the number of hidden variables, it effectively avoids the problem of selecting the number of hidden layer nodes. The experimental simulation results show that the model has the characteristics of strong robustness and stable generalization performance compared with other models, and provides a reliable model for complex chemical process modeling.
【學(xué)位授予單位】：北京化工大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2016
【分類號(hào)】：TP18

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 時(shí)振偉;紀(jì)志成;;基于多元線性回歸模型的永磁同步電機(jī)參數(shù)辨識(shí)的新方法（英文）[J];系統(tǒng)仿真學(xué)報(bào);2015年08期

2 賀彥林;徐圓;耿志強(qiáng);朱群雄;;基于自聯(lián)想遞階神經(jīng)網(wǎng)絡(luò)的多輸入?yún)?shù)化工過(guò)程軟傳感器(英文)[J];Chinese Journal of Chemical Engineering;2015年01期

3 賀彥林;王曉;朱群雄;;基于主成分分析-改進(jìn)的極限學(xué)習(xí)機(jī)方法的精對(duì)苯二甲酸醋酸含量軟測(cè)量[J];控制理論與應(yīng)用;2015年01期

4 劉菲菲;彭荻;賀彥林;朱群雄;;基于極限學(xué)習(xí)的過(guò)程神經(jīng)網(wǎng)絡(luò)研究及化工應(yīng)用[J];上海交通大學(xué)學(xué)報(bào);2014年07期

5 劉學(xué)藝;宋春躍;李平;;基于Vapnik-Chervonenkis泛化界的極限學(xué)習(xí)機(jī)模型復(fù)雜性控制[J];控制理論與應(yīng)用;2014年05期

6 劉國(guó)超;賀彥林;朱群雄;;隱含層組合型ELM研究及應(yīng)用[J];計(jì)算機(jī)與應(yīng)用化學(xué);2013年12期

7 李作成;錢斌;胡蓉;羅蓉娟;張桂蓮;;遺傳-分布估計(jì)算法求解化工生產(chǎn)中一類帶多工序的異構(gòu)并行機(jī)調(diào)度問(wèn)題[J];化工學(xué)報(bào);2014年03期

8 高慧慧;賀彥林;彭荻;朱群雄;;基于數(shù)據(jù)屬性劃分的遞階ELM研究及化工應(yīng)用[J];化工學(xué)報(bào);2013年12期

9 王改堂;李平;蘇成利;;ELM嶺回歸軟測(cè)量建模方法[J];合肥工業(yè)大學(xué)學(xué)報(bào)(自然科學(xué)版);2011年08期

10 陳建宏;劉浪;周智勇;永學(xué)艷;;基于主成分分析與神經(jīng)網(wǎng)絡(luò)的采礦方法優(yōu)選[J];中南大學(xué)學(xué)報(bào)(自然科學(xué)版);2010年05期

，

本文編號(hào)：1347685