危險化學品泄漏源的定位研究
發(fā)布時間：2019-01-23 15:09
【摘要】：泄漏源的定位是危險化學品事故應急救援的基礎與關(guān)鍵。論文基于監(jiān)測模式、擴散模式以及源強反算算法研究進行設計，以 定位模型建立——模型優(yōu)化——定位模型驗證‖為研究主線，開展危險化學品泄漏源定位的相關(guān)科學研究。主要完成了以下工作： （1）以優(yōu)化方法為模型框架建立泄漏源定位的優(yōu)化反算模型，率先提出利用模式搜索法進行毒氣泄漏源的定位。利用事故現(xiàn)場數(shù)據(jù)和擴散模式將泄漏源定位轉(zhuǎn)化為優(yōu)化問題求解，利用模式搜索法逐步更新尋找計算濃度與監(jiān)測濃度的最優(yōu)匹配。另一方面，模式搜索法提供了領域空間的搜索思想，為嵌入其他全局搜索法提供理論基礎，提高定位準確性和有效性。進而，通過設計混合算法結(jié)構(gòu)以及混合時機的選取等角度分析，利用基于鑲嵌型的混合優(yōu)化算法進行源強反算試算，結(jié)果表明混合算法顯著提高了反算精度。 （2）建立基于貝葉斯推理和優(yōu)化算法相結(jié)合的泄漏源參數(shù)識別方法，將模型參數(shù)的先驗信息、最終反算結(jié)果都通過概率分布來描述。在貝葉斯推理的基礎上，利用MCMC抽樣方法對后驗概率分布進行抽樣，得到參數(shù)的估計值。為了改善MCMC抽樣過程的計算效率，提出基于優(yōu)化算法的初始化過程，在抽樣之前利用優(yōu)化算法進行全局最佳化采樣，使得混合的算法既能保持貝葉斯方法對不確定性問題的求解性能，又提高計算效率。 （3）建立基于元胞自動機方法的氣體擴散模式，實現(xiàn)事故物質(zhì)濃度在空間中的實時動態(tài)分布預測。通過優(yōu)化模型方法或貝葉斯推理方法，將元胞自動機模型與實際觀測數(shù)據(jù)相結(jié)合進行泄漏源的反演。仿真結(jié)果表明該模型方法能夠提高反演結(jié)果精度，降低錯誤識別的概率。 （4）通過仿真模擬與外場試驗驗證相結(jié)合，驗證泄漏源定位方法的有效性。在理論模型研究的基礎上，通過仿真數(shù)據(jù)在簡單環(huán)境中進行驗證；再利用外場試驗進行實證檢驗。建立外場試驗平臺，通過固定監(jiān)測網(wǎng)絡與移動監(jiān)測的結(jié)合，有效獲取事故物質(zhì)的濃度信息。結(jié)果表明優(yōu)化方法在復雜環(huán)境中若擴散模型選取不當將造成較大誤差，而貝葉斯推理由于考慮到觀測誤差以及模型誤差，不論是仿真驗證還是實證檢驗都能夠獲取相對較好的結(jié)果。 創(chuàng)新之處主要體現(xiàn)在：1）引入了源強反算思想，率先提出利用模式搜索法進行源強反算研究。在優(yōu)化模型框架下，建立了不同監(jiān)測模式下的泄漏源定位方法。2）將貝葉斯推理與優(yōu)化方法相結(jié)合，，利用優(yōu)化算法獲取MCMC抽樣的初始抽樣點，既保證貝葉斯方法對不確定性問題的求解性能，又提高了計算的效率和精度。3）將元胞自動機方法引入用于毒氣擴散的建模與求解，更準確的描述物質(zhì)在空間中的動態(tài)分布。
[Abstract]:The location of leakage source is the foundation and key of emergency rescue of hazardous chemical accident. Based on the research of monitoring mode, diffusion mode and inverse calculation algorithm of source strength, this paper carries out scientific research on the location of hazardous chemicals leakage source, taking the establishment of location model, optimization of model and verification of location model as the main line of research. The main works are as follows: (1) the optimal inverse calculation model of leak source location is established based on the optimization method, and the method of pattern search is first used to locate the gas leak source. The leakage source location is transformed into an optimization problem by using the accident site data and diffusion mode. The pattern search method is used to update the method to find the optimal match between the calculated concentration and the monitoring concentration. On the other hand, the pattern search method provides the domain space search idea, provides the theoretical basis for embedding other global search methods, and improves the accuracy and effectiveness of location. Furthermore, by designing the structure of the hybrid algorithm and selecting the timing of the hybrid algorithm, a hybrid optimization algorithm based on mosaic is applied to the inverse calculation of the source strength. The results show that the hybrid algorithm improves the accuracy of the inverse calculation significantly. (2) based on Bayesian reasoning and optimization algorithm, the identification method of leakage source parameters is established. The prior information of model parameters and the final inverse results are all described by probability distribution. On the basis of Bayesian reasoning, the posterior probability distribution is sampled by MCMC sampling method, and the estimation of parameters is obtained. In order to improve the computational efficiency of MCMC sampling process, the initialization process based on optimization algorithm is proposed. The hybrid algorithm can not only preserve the performance of Bayesian method in solving uncertain problems, but also improve the computational efficiency. (3) establish a gas diffusion model based on cellular automata method, and realize the real-time dynamic distribution prediction of the concentration of accident matter in space. By means of optimization model method or Bayesian reasoning method, the cellular automata model is combined with the observed data to retrieve the leakage source. Simulation results show that the model method can improve the accuracy of inversion results and reduce the probability of error recognition. (4) the effectiveness of the leak source location method is verified by the combination of simulation and field test. On the basis of the theoretical model research, the simulation data is used to verify the model in simple environment, and the empirical test is carried out by using the field test. An outfield test platform was established to obtain the concentration information of the accident material effectively through the combination of fixed monitoring network and mobile monitoring. The results show that if the diffusion model is not selected properly in complex environment, the optimization method will cause a large error, while Bayesian reasoning takes into account the observation error and model error. Both simulation and empirical tests can obtain relatively good results. The main innovations are as follows: 1) the idea of inverse calculation of source strength is introduced, and the method of pattern search is firstly proposed to study the inverse calculation of source strength. Under the framework of the optimization model, the leak source location methods in different monitoring modes are established. 2) the Bayesian reasoning and optimization methods are combined to obtain the initial sampling points of MCMC sampling. It not only guarantees the performance of Bayesian method for solving uncertain problems, but also improves the efficiency and accuracy of calculation. 3) the cellular automata method is introduced into the modeling and solution of gas diffusion, and the dynamic distribution of matter in space is described more accurately.
【學位授予單位】：北京化工大學
【學位級別】：博士
【學位授予年份】：2013
【分類號】：X937
本文編號：2413932
[Abstract]:The location of leakage source is the foundation and key of emergency rescue of hazardous chemical accident. Based on the research of monitoring mode, diffusion mode and inverse calculation algorithm of source strength, this paper carries out scientific research on the location of hazardous chemicals leakage source, taking the establishment of location model, optimization of model and verification of location model as the main line of research. The main works are as follows: (1) the optimal inverse calculation model of leak source location is established based on the optimization method, and the method of pattern search is first used to locate the gas leak source. The leakage source location is transformed into an optimization problem by using the accident site data and diffusion mode. The pattern search method is used to update the method to find the optimal match between the calculated concentration and the monitoring concentration. On the other hand, the pattern search method provides the domain space search idea, provides the theoretical basis for embedding other global search methods, and improves the accuracy and effectiveness of location. Furthermore, by designing the structure of the hybrid algorithm and selecting the timing of the hybrid algorithm, a hybrid optimization algorithm based on mosaic is applied to the inverse calculation of the source strength. The results show that the hybrid algorithm improves the accuracy of the inverse calculation significantly. (2) based on Bayesian reasoning and optimization algorithm, the identification method of leakage source parameters is established. The prior information of model parameters and the final inverse results are all described by probability distribution. On the basis of Bayesian reasoning, the posterior probability distribution is sampled by MCMC sampling method, and the estimation of parameters is obtained. In order to improve the computational efficiency of MCMC sampling process, the initialization process based on optimization algorithm is proposed. The hybrid algorithm can not only preserve the performance of Bayesian method in solving uncertain problems, but also improve the computational efficiency. (3) establish a gas diffusion model based on cellular automata method, and realize the real-time dynamic distribution prediction of the concentration of accident matter in space. By means of optimization model method or Bayesian reasoning method, the cellular automata model is combined with the observed data to retrieve the leakage source. Simulation results show that the model method can improve the accuracy of inversion results and reduce the probability of error recognition. (4) the effectiveness of the leak source location method is verified by the combination of simulation and field test. On the basis of the theoretical model research, the simulation data is used to verify the model in simple environment, and the empirical test is carried out by using the field test. An outfield test platform was established to obtain the concentration information of the accident material effectively through the combination of fixed monitoring network and mobile monitoring. The results show that if the diffusion model is not selected properly in complex environment, the optimization method will cause a large error, while Bayesian reasoning takes into account the observation error and model error. Both simulation and empirical tests can obtain relatively good results. The main innovations are as follows: 1) the idea of inverse calculation of source strength is introduced, and the method of pattern search is firstly proposed to study the inverse calculation of source strength. Under the framework of the optimization model, the leak source location methods in different monitoring modes are established. 2) the Bayesian reasoning and optimization methods are combined to obtain the initial sampling points of MCMC sampling. It not only guarantees the performance of Bayesian method for solving uncertain problems, but also improves the efficiency and accuracy of calculation. 3) the cellular automata method is introduced into the modeling and solution of gas diffusion, and the dynamic distribution of matter in space is described more accurately.
【學位授予單位】：北京化工大學
【學位級別】：博士
【學位授予年份】：2013
【分類號】：X937
【參考文獻】
相關(guān)期刊論文 前10條
1 朱江;汪萍;;集合卡爾曼平滑和集合卡爾曼濾波在污染源反演中的應用[J];大氣科學;2006年05期
2 蘇芳,邵敏,蔡旭輝,曾立民,朱彤;利用逆向軌跡反演模式估算北京地區(qū)甲烷源強[J];環(huán)境科學學報;2002年05期
3 丁信偉,王淑蘭,徐國慶;可燃及毒性氣體泄漏擴散研究綜述[J];化學工業(yè)與工程;1999年02期
4 吳維平;港口油氣污染擴散及源強計算方法的探討[J];交通環(huán)保;1999年06期
5 關(guān)磊;魏利軍;吳宗之;劉驥;桑海泉;;基于人工神經(jīng)網(wǎng)絡的儲罐區(qū)泄漏源實時定位技術(shù)初步研究[J];中國安全生產(chǎn)科學技術(shù);2006年04期
6 吳宗之;張圣柱;張悅;石超;劉寧;楊國梁;;2006-2010年我國危險化學品事故統(tǒng)計分析研究[J];中國安全生產(chǎn)科學技術(shù);2011年07期
7 郭少冬;楊銳;翁文國;;基于MCMC方法的城區(qū)有毒氣體擴散源反演[J];清華大學學報(自然科學版);2009年05期
8 陳軍明,徐大海,朱蓉;遺傳算法在點源擴散濃度反演排放源強中的應用[J];氣象;2002年09期
9 張建文;王煜薇;鄭小平;王正;;基于混合遺傳-Nelder Mead單純形算法的源強及位置反算[J];系統(tǒng)工程理論與實踐;2011年08期
10 呂武軒,施小丁,趙寶元,劉慧;危險氣體擴散實時仿真的關(guān)鍵技術(shù)──“反求源強”方法初探[J];中國安全科學學報;1997年S1期
相關(guān)碩士學位論文 前1條
1 王煜薇;基于混合遺傳-Nelder Mead單純形算法的源強及位置反算[D];北京化工大學;2011年
本文編號：2413932
本文鏈接：http://sikaile.net/kejilunwen/anquangongcheng/2413932.html
最近更新
教材專著
