【摘要】：針對(duì)現(xiàn)有改進(jìn)互信息爬山(MIHC)算法精度低、耗時(shí)長(zhǎng)及簡(jiǎn)化爬山(SHC)算法產(chǎn)生大量冗余邊的問題,提出一種新的結(jié)構(gòu)學(xué)習(xí)算法,即改進(jìn)爬山(IHC)算法。通過計(jì)算互信息鏈得到貝葉斯初始結(jié)構(gòu),利用條件獨(dú)立性測(cè)試以及對(duì)孤立節(jié)點(diǎn)進(jìn)行處理來加邊補(bǔ)充貝葉斯初始結(jié)構(gòu)得到完全結(jié)構(gòu),利用改進(jìn)的爬山搜索算子對(duì)完全結(jié)構(gòu)進(jìn)行搜索直到得出最優(yōu)結(jié)構(gòu)。將該算法與爬山(HC)算法、MIHC算法、SHC算法進(jìn)行比較,仿真結(jié)果表明,IHC算法能夠得到較高準(zhǔn)確率的模型,時(shí)間開銷最小而且產(chǎn)生的冗余邊數(shù)遠(yuǎn)遠(yuǎn)少于SHC算法產(chǎn)生的冗余邊數(shù)。最后基于IHC算法,結(jié)合某回轉(zhuǎn)窯數(shù)據(jù)進(jìn)行訓(xùn)練,得到了回轉(zhuǎn)窯工藝參數(shù)的故障診斷模型,對(duì)回轉(zhuǎn)窯的燒成帶溫度實(shí)現(xiàn)了較為準(zhǔn)確的故障診斷。
[Abstract]:Aiming at the problems of low precision, long time consuming and large number of redundant edges generated by the existing improved mutual information mountain climbing (MIHC) algorithm, a new structure learning algorithm, that is, the improved hill climbing (IHC) algorithm, is proposed to solve the problem of reducing the amount of redundant edges generated by the simplified hill climbing (SHC) algorithm. The Bayesian initial structure is obtained by computing the mutual information chain, and the complete structure is obtained by using conditional independence test and processing of isolated nodes to add edges to supplement the Bayesian initial structure. The improved mountain-climbing search operator is used to search the complete structure until the optimal structure is obtained. Compared with (HC) algorithm, MIHC algorithm and SHC algorithm, the simulation results show that the IHC algorithm can get a high accuracy model. The time cost is minimum and the number of redundant edges generated is far less than the number of redundant edges generated by the SHC algorithm. Finally, based on the IHC algorithm and the data of a rotary kiln, the fault diagnosis model of process parameters of rotary kiln is obtained, and the fault diagnosis of firing zone temperature of rotary kiln is realized accurately.
【作者單位】： 燕山大學(xué)信息科學(xué)與工程學(xué)院;燕山大學(xué)電氣工程學(xué)院;
【基金】：國(guó)家自然科學(xué)基金資助項(xiàng)目(51641609) 河北省自然科學(xué)基金資助項(xiàng)目(F2016203354)
【分類號(hào)】：TP18;TQ054

【相似文獻(xiàn)】

相關(guān)碩士學(xué)位論文 前1條

1 楊振;化學(xué)反應(yīng)優(yōu)化算法的理論研究與改進(jìn)[D];南通大學(xué);2014年

，

本文編號(hào)：2429861