何葉榮 范志豪
摘要:為切實(shí)做好煤礦安全管理工作,確保煤礦安全高效生產(chǎn),從煤礦安全管理現(xiàn)狀入手,分析影響煤礦安全管理的風(fēng)險(xiǎn)因素;運(yùn)用粗糙集理論對(duì)收集的60個(gè)煤礦安全管理風(fēng)險(xiǎn)因素指標(biāo)進(jìn)行約減,精簡(jiǎn)出18個(gè)風(fēng)險(xiǎn)評(píng)價(jià)指標(biāo)和3個(gè)風(fēng)險(xiǎn)后果指標(biāo);通過(guò)現(xiàn)場(chǎng)訪談,結(jié)合專家咨詢和問(wèn)卷調(diào)查,選擇12個(gè)典型煤礦作為樣本煤礦;根據(jù)煤礦安全管理非線性、動(dòng)態(tài)時(shí)變性及不確定性等特點(diǎn),對(duì)結(jié)構(gòu)方程模型SEM(structural equation model)評(píng)價(jià)方法、模糊支持向量機(jī)FSVM(fuzzy support vector machines)評(píng)價(jià)方法,以及結(jié)構(gòu)方程模型和模糊支持向量機(jī)相結(jié)合的SEM-FSVM風(fēng)險(xiǎn)評(píng)價(jià)方法進(jìn)行研究,并對(duì)這3種評(píng)價(jià)方法評(píng)價(jià)的結(jié)果進(jìn)行對(duì)比。研究證明,將SEM 的路徑系數(shù)與FSVM的核函數(shù)進(jìn)行結(jié)合,所構(gòu)建的基于樣本和指標(biāo)雙權(quán)重的SEM-FSVM風(fēng)險(xiǎn)集成評(píng)價(jià)模型,評(píng)價(jià)過(guò)程科學(xué)合理,評(píng)價(jià)結(jié)果更加精確,能較好地為實(shí)現(xiàn)煤礦安全管理精細(xì)化、本質(zhì)化提供決策依據(jù)。
關(guān)鍵詞:雙權(quán)重;結(jié)構(gòu)方程模型;模糊支持向量機(jī);煤礦安全管理;評(píng)價(jià)方法
中圖分類號(hào):TD 76;X 936文獻(xiàn)標(biāo)志碼:A
文章編號(hào):1672-9315(2022)03-0600-07
DOI:10.13800/j.cnki.xakjdxxb.2022.0324開(kāi)放科學(xué)(資源服務(wù))標(biāo)識(shí)碼(OSID):
Research on? double-weight risk assessment method
of coal mine safety managementHE Yerong,F(xiàn)AN ZHihao
(School of Economic and Management,Anhui Jianzhu University,Hefei 230601,China)Abstract:In order to manage mine safety effectively and ensure efficient and safe production of coal mines,the risk factors affecting safety management were analyzed with the present situation in view.The collected 60 risk factors of coal mine safety management were reduced by using Rough Set Theory,in which 18 risk evaluation indicators and 3 consequences indicators were extracted.Twelve typical coal mines were selected as samples by field interview,expert consultation and questionnaire survey.Based on the characteristics of non-linear,dynamic time-varying and fuzzy information of coal mine safety management,the structural equation model(SEM),fuzzy support vector machine(FSVM)and the combination of these two risk assessment methods(SEM-FSVM) were studied with their results compared.The empirical research integrated? combining the path system number of SEM with the kernel function of FSVM,the indicates two-weighted risk? evaluation method is more reasonable,and the evaluation results are more accurate,which is more conducive to? the essential safety management of coal mine.
Key words:double-weight;structural equation model;fuzzy support vector machine;coal mine safety management;evaluation method
0引言
煤炭產(chǎn)業(yè)在中國(guó)一次性能源生產(chǎn)和消費(fèi)結(jié)構(gòu)中占主導(dǎo)地位,“十四五”時(shí)期,煤炭行業(yè)將進(jìn)入高質(zhì)量發(fā)展攻堅(jiān)期,煤炭將占到中國(guó)一次能源消費(fèi)一半以上。預(yù)計(jì)到2025年,煤炭消費(fèi)量在41億t左右,占比約為52%[1]。但是中國(guó)煤礦環(huán)境復(fù)雜,安全事故多發(fā)。國(guó)家礦山安全監(jiān)察局對(duì)2020年煤礦安全事故案例進(jìn)行了梳理,統(tǒng)計(jì)出了全國(guó)煤礦安全事故十大典型案例[2],見(jiàn)表1。
近年來(lái),中國(guó)煤礦安全法律法規(guī)體系進(jìn)一步健全,監(jiān)管體制機(jī)制不斷完善,安全管理越來(lái)越規(guī)范,煤礦安全形勢(shì)明顯好轉(zhuǎn),重特大事故明顯減少,2019年全國(guó)煤礦發(fā)生死亡事故170起、死亡316人,同比分別下降24.1%和5.1%;百萬(wàn)噸死亡率0.083%,同比下降10.8%[3]。2020年以來(lái),面對(duì)極其嚴(yán)峻復(fù)雜的國(guó)內(nèi)外形勢(shì),特別是新冠肺炎疫情嚴(yán)重沖擊,在黨中央、國(guó)務(wù)院的堅(jiān)強(qiáng)領(lǐng)導(dǎo)下,廣大煤礦企業(yè)認(rèn)真貫徹落實(shí)黨中央決策部署,統(tǒng)籌抓好疫情防控和安全生產(chǎn)工作,健全公共安全體系,完善安全生產(chǎn)責(zé)任制,提升安全事故的防范能力,使全國(guó)煤礦事故總量、重大事故數(shù)量、百萬(wàn)噸死亡率持續(xù)下降,煤礦安全生產(chǎn)形勢(shì)持續(xù)穩(wěn)定向好。但是,煤礦安全事故仍時(shí)有發(fā)生,與歐美一些國(guó)家相比,事故總量依然偏大 [4]。暴露出一些煤礦企業(yè)法律意識(shí)淡薄、違法違規(guī)行為屢禁不止、事故隱患較多等影響安全生產(chǎn)的突出問(wèn)題和薄弱環(huán)節(jié)尚未得到根本解決。5C8C7694-51B2-48E7-8957-15E1C567EE6E
據(jù)統(tǒng)計(jì),已發(fā)生的煤礦安全事故中,90%以上是由于人的因素所致[4],本質(zhì)上是由于安全管理所致。近些年來(lái),國(guó)內(nèi)外專家、學(xué)者針對(duì)煤礦安全管理風(fēng)險(xiǎn)問(wèn)題,開(kāi)展了大量研究,取得了豐富的成果[4-7],推動(dòng)了煤礦安全管理工作。然而,由于煤礦安全管理的動(dòng)態(tài)時(shí)變性和非線性,這些理論在實(shí)際應(yīng)用中受到一定的限制,實(shí)用性不強(qiáng)[7]。結(jié)構(gòu)方程模型(structural equation model,SEM)[7-10]可以對(duì)多個(gè)因變量同時(shí)處理,分析風(fēng)險(xiǎn)因子對(duì)風(fēng)險(xiǎn)結(jié)果的影響路徑,而且對(duì)煤礦安全管理多變的、相互演化的風(fēng)險(xiǎn)關(guān)系,通過(guò)路徑系數(shù)進(jìn)行分析,并精確計(jì)算出來(lái);模糊支持向量機(jī)(fuzzy support vector machine,F(xiàn)SVM)[11-15],是將模糊隸屬度添加于支持向量機(jī)二次規(guī)劃的懲罰參數(shù)中,能夠?qū)μ厥鈽颖具M(jìn)行模糊隸屬度賦值,消除數(shù)據(jù)差異的影響。文中在對(duì)國(guó)內(nèi)外相關(guān)研究進(jìn)行梳理的基礎(chǔ)上,結(jié)合中國(guó)煤礦安全管理現(xiàn)狀,擬采用模糊支持向量機(jī)(FSVM)和結(jié)構(gòu)方程(SEM)相結(jié)合的方法,對(duì)煤礦安全管理風(fēng)險(xiǎn)進(jìn)行評(píng)價(jià)[3]。
1理論基礎(chǔ)及程序
1.1理論基礎(chǔ)
運(yùn)用SEM進(jìn)行風(fēng)險(xiǎn)因素分析,構(gòu)建SEM風(fēng)險(xiǎn)因素結(jié)構(gòu)模型,計(jì)算風(fēng)險(xiǎn)因素路徑系數(shù),確定風(fēng)險(xiǎn)評(píng)價(jià)指標(biāo)權(quán)重,將該權(quán)重與FSVM的核函數(shù)進(jìn)行內(nèi)積運(yùn)算,建立特征加權(quán)核函數(shù),由此形成FSVM新的核函數(shù),以平衡指標(biāo)貢獻(xiàn)度對(duì)風(fēng)險(xiǎn)評(píng)價(jià)結(jié)果的影響。經(jīng)過(guò)處理后的模型既能簡(jiǎn)化樣本數(shù)據(jù)處理,又能很好地解決煤礦安全管理的動(dòng)態(tài)性、時(shí)變性等問(wèn)題[3]。這種將樣本權(quán)重與指標(biāo)權(quán)重同時(shí)考慮的雙權(quán)重風(fēng)險(xiǎn)評(píng)價(jià)方法,能大大提高煤礦安全管理風(fēng)險(xiǎn)評(píng)價(jià)的精度和效率。
1.2程序與步驟
運(yùn)用SEM和FSVM結(jié)合的方法,將SEM風(fēng)險(xiǎn)路徑系數(shù)值作為評(píng)價(jià)指標(biāo)特征權(quán)重向量,與GAUSS核函數(shù)進(jìn)行內(nèi)積運(yùn)算,構(gòu)建特征加權(quán)GAUSS核函數(shù),建立特征加權(quán)支持向量機(jī)[3]。具體程序如圖1所示。
2評(píng)價(jià)過(guò)程
2.1選擇樣本
通過(guò)現(xiàn)場(chǎng)訪談,結(jié)合專家咨詢和問(wèn)卷調(diào)查,選擇12個(gè)典型煤礦為樣本(7個(gè)正類,5個(gè)負(fù)類)。運(yùn)用粗糙集理論對(duì)收集的60個(gè)煤礦安全管理風(fēng)險(xiǎn)因素指標(biāo)進(jìn)行約減,提煉出18個(gè)風(fēng)險(xiǎn)評(píng)價(jià)指標(biāo)和3個(gè)風(fēng)險(xiǎn)后果指標(biāo)。數(shù)據(jù)采集期為2017—2019年,取3年的數(shù)據(jù)均值作為指標(biāo)分值。
2.2設(shè)計(jì)風(fēng)險(xiǎn)類別
風(fēng)險(xiǎn)等級(jí)區(qū)間采用1~10分,采用李克特五級(jí)量表,分為5級(jí)分值區(qū)間:[0-2]無(wú)風(fēng)險(xiǎn),(2-4]輕微風(fēng)險(xiǎn),(4-6]一般風(fēng)險(xiǎn),(6-8]較大風(fēng)險(xiǎn),(8-10]嚴(yán)重風(fēng)險(xiǎn);風(fēng)險(xiǎn)后果設(shè)置為±類:-1表示有風(fēng)險(xiǎn)后果,+1表示無(wú)風(fēng)險(xiǎn)后果[3]。
2.3數(shù)據(jù)處理
2.3.1數(shù)據(jù)預(yù)處理
2.3.2數(shù)據(jù)信度、效度分析
對(duì)預(yù)處理數(shù)據(jù)進(jìn)行信度和效度分析,分析結(jié)果總體量表的Alpha值為0.928,大于參照值0.7,指標(biāo)信度很好。效度是衡量測(cè)量結(jié)果與實(shí)際情況的符合程度的,通常運(yùn)用KMO和Bartlett球形檢驗(yàn)。文中風(fēng)險(xiǎn)評(píng)價(jià)指標(biāo)的KMO值為0.886,Bartlett球型檢驗(yàn)值為1 650.33,其概率Sig值為0,存在顯著差異,指標(biāo)信度和效度較好。
2.4模型構(gòu)建與風(fēng)險(xiǎn)評(píng)價(jià)
2.4.1模糊隸屬函數(shù)選擇及參數(shù)設(shè)置
2.4.2核函數(shù)構(gòu)建
2.4.3選擇訓(xùn)練樣本,構(gòu)建模糊訓(xùn)練集
2.4.4選定測(cè)試樣本進(jìn)行測(cè)試
3結(jié)果分析
3.1評(píng)價(jià)結(jié)果分析
4結(jié)論
由上述評(píng)價(jià)結(jié)果來(lái)看,總誤差和平均誤差均最小的是基于指標(biāo)和樣本雙權(quán)重的SEM-FSVM風(fēng)險(xiǎn)評(píng)價(jià)方法,評(píng)價(jià)精度最高,SEM和FSVM評(píng)價(jià)精度相對(duì)偏低。主要原因在于
1)樣本數(shù)據(jù)可能有野值點(diǎn)的存在,導(dǎo)致評(píng)價(jià)結(jié)果出現(xiàn)誤差;另外,指標(biāo)數(shù)據(jù)的處理方法也有可能影響評(píng)價(jià)結(jié)果,導(dǎo)致出現(xiàn)誤差。
2)SEM一般用于評(píng)價(jià)多因素多變量之間的影響關(guān)系,進(jìn)行多樣本風(fēng)險(xiǎn)評(píng)價(jià)比較困難,要逐個(gè)將樣本的指標(biāo)分值代入模型進(jìn)行計(jì)算,計(jì)算過(guò)程比較繁瑣,容易出錯(cuò)且耗時(shí)較長(zhǎng),評(píng)價(jià)精度不理想。
3)基于FSVM的評(píng)價(jià)模型,雖然引入了模糊隸屬度,對(duì)于孤立點(diǎn)賦予了很小的隸屬度,但是樣本指標(biāo)權(quán)值對(duì)評(píng)價(jià)結(jié)果還是有一定影響的。
4)基于特征和樣本雙重加權(quán)的SEM-FSVM風(fēng)險(xiǎn)評(píng)價(jià)模型,通過(guò)構(gòu)建SEM風(fēng)險(xiǎn)因素結(jié)構(gòu)模型,將SEM路徑系數(shù)與GAUSS核函數(shù)進(jìn)行集成,構(gòu)造特征加權(quán)GAUSS核函數(shù),對(duì)FSVM的核函數(shù)進(jìn)行了改進(jìn)。改造后的SEM-FSVM模型,一方面能注重樣本重要性,同時(shí)又充分考慮指標(biāo)貢獻(xiàn)度,降低采集數(shù)據(jù)成本。此方法適用于小樣本、復(fù)雜多因素評(píng)價(jià),評(píng)價(jià)精度較高,能為煤礦安全風(fēng)險(xiǎn)管理提供決策依據(jù)和理論指導(dǎo),從而有效降低煤礦安全事故。
參考文獻(xiàn)(References):
[1]中國(guó)煤炭新聞網(wǎng)[EB/OL].http://www.cwestc.com/newshtml/2019-12-19/593155.shtml
[2]國(guó)家礦山安全監(jiān)察局.[EB/OL].http://www.chinacoal-safety.gov.cn/xw/mkaqjcxw/.2021.01.25
[3]何葉榮.基于SEM-FSVM的煤礦安全管理風(fēng)險(xiǎn)評(píng)價(jià)[D].淮南:安徽理工大學(xué),2016.HE Yerong.Risk evaluation of coal-mine safety management based on SEM-FSVM[D].Huainan:Anhui University Science & Technology,2016.
[4]MA J Q,DAI H.A methodology to construct warning index system for coal mine safety based on collaborative management[J].Safety Science,2017,93(5):86-95.5C8C7694-51B2-48E7-8957-15E1C567EE6E
[5]曹慶仁,李凱.各種煤礦安全管理行為及其相互影響作用研究[J].安全與環(huán)境學(xué)報(bào),2015,15(1):6-10.CAO Qingren,LI Kai.Research on safety management behavior of coal mine and its interaction[J].Journal of Safety and Environment,2015,15(1):6-10.
[6]WANG L,CAO Q,ZHOU L.Research on the influencing factors in coal mine production safety based on the combination of DEMATEL and ISM[J].Safety Science,2018,103:51-61.
[7]田水承,唐凱,田方圓.基于DFSS理論的煤礦企業(yè)培訓(xùn)安全質(zhì)量管理體系構(gòu)建研究[J].煤礦安全,2020,51(6):251-254.TIAN Shuicheng,TANG Kai,TIAN Fangyuan.Quality management system construction of coal mine enterprises safety traning based on DFSS theory[J].Safety in Coal Mines,2020,51(6):251-254.
[8]YANG D,XIE K,OZBAY K,et al.Fusing crash data and surrogate safety measures for safety assessment:development of a structural equation model with conditional autoregressive spatial effect and random parameters[J].Accident Analysis and Prevention,2021(1):1-13.
[9]JONAS G,LEV?QUE,ROBERT? C B.A structural equation modeling approach to water quality perceptions[J].Journal of Environmental Management,2018,52(8):143-149.
[10]鐘茂華,田向亮,劉暢,等.基于結(jié)構(gòu)方程模型的地鐵乘客安全行為影響因素分析[J].中國(guó)安全生產(chǎn)科學(xué)技術(shù),2018,14(1):5-11.ZHONG Maohua,TIAN Xiangliang,LIU Chang,et al.Analysis of influencing factors of subway passenger safety behavior based on structural equation model[J].Journal of Safety Science and Technology,2018,14(1):5-11.
[11]LU R W,WANG X H,YU H,et al.Multiparty evolutionary game model in coal mine safety management and its application[J].Complexity,2018(3):1-10.
[12]ROBLAW Q W.Fuzzy support vector regression machine with Penalizing Gaussian noises on triangular fuzzy number space[J].Expert Systems with Applications,2010,37(12):7788-7795.
[13]YANG B.Dynamic risk identification safety model based on fuzzy support vector machine and immune optimization algorithm[J].Safety Science,2019,118:205-211.
[14]何葉榮,孟祥瑞.煤礦企業(yè)安全管理能力風(fēng)險(xiǎn)因素結(jié)構(gòu)模型[J].中國(guó)安全生產(chǎn)科學(xué)技術(shù),2015,11(3):135-140.HE Yerong,MENG Xiangrui.Structural model on risk factors of safety management capability in coal mine enterprise[J].Journal of Safety Science and Technology,2015,11(3):135-140.
[15]FURKAN B,AYSEN A.Hybrid fuzzy support vector regression analysis[J].Journal of Intelligent and Fuzzy Systems,2015,28(5):2037-2040.
[16]TAHEREH B,SAEEDEH G,SADOGHI H Y.Toward optimum fuzzy support vector machines using error distribution[J].Engineering Applications of Artificial Intelligence,2020,90:214-217.5C8C7694-51B2-48E7-8957-15E1C567EE6E