危险化学品泄漏源的定位研究

发布时间：2019-01-23 15:09

【摘要】：泄漏源的定位是危险化学品事故应急救援的基础与关键。论文基于监测模式、扩散模式以及源强反算算法研究进行设计，以定位模型建立——模型优化——定位模型验证‖为研究主线，开展危险化学品泄漏源定位的相关科学研究。主要完成了以下工作：（1）以优化方法为模型框架建立泄漏源定位的优化反算模型，率先提出利用模式搜索法进行毒气泄漏源的定位。利用事故现场数据和扩散模式将泄漏源定位转化为优化问题求解，利用模式搜索法逐步更新寻找计算浓度与监测浓度的最优匹配。另一方面，模式搜索法提供了领域空间的搜索思想，为嵌入其他全局搜索法提供理论基础，提高定位准确性和有效性。进而，通过设计混合算法结构以及混合时机的选取等角度分析，利用基于镶嵌型的混合优化算法进行源强反算试算，结果表明混合算法显著提高了反算精度。（2）建立基于贝叶斯推理和优化算法相结合的泄漏源参数识别方法，将模型参数的先验信息、最终反算结果都通过概率分布来描述。在贝叶斯推理的基础上，利用MCMC抽样方法对后验概率分布进行抽样，得到参数的估计值。为了改善MCMC抽样过程的计算效率，提出基于优化算法的初始化过程，在抽样之前利用优化算法进行全局最佳化采样，使得混合的算法既能保持贝叶斯方法对不确定性问题的求解性能，又提高计算效率。（3）建立基于元胞自动机方法的气体扩散模式，实现事故物质浓度在空间中的实时动态分布预测。通过优化模型方法或贝叶斯推理方法，将元胞自动机模型与实际观测数据相结合进行泄漏源的反演。仿真结果表明该模型方法能够提高反演结果精度，降低错误识别的概率。（4）通过仿真模拟与外场试验验证相结合，验证泄漏源定位方法的有效性。在理论模型研究的基础上，通过仿真数据在简单环境中进行验证；再利用外场试验进行实证检验。建立外场试验平台，通过固定监测网络与移动监测的结合，有效获取事故物质的浓度信息。结果表明优化方法在复杂环境中若扩散模型选取不当将造成较大误差，而贝叶斯推理由于考虑到观测误差以及模型误差，不论是仿真验证还是实证检验都能够获取相对较好的结果。创新之处主要体现在：1）引入了源强反算思想，率先提出利用模式搜索法进行源强反算研究。在优化模型框架下，建立了不同监测模式下的泄漏源定位方法。2）将贝叶斯推理与优化方法相结合，，利用优化算法获取MCMC抽样的初始抽样点，既保证贝叶斯方法对不确定性问题的求解性能，又提高了计算的效率和精度。3）将元胞自动机方法引入用于毒气扩散的建模与求解，更准确的描述物质在空间中的动态分布。
[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.
【学位授予单位】：北京化工大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：X937

【参考文献】