The Application Of HSE Management System And Expert System On Environment Safety In Petrochemical Corporation

As the development of the society, the needs for the resources are increasing. Meanwhile, the care for the living condition is getting more and more. In such a situation, chemical corporations have to make choices between enlarging the production scale and the protection of employees’ health, the equipment and the environment. The appearance of Health, Safety and Environment Management System （HSE as short） makes it possible to solve this problem properly.HSE management is an effective management mode, which analyzes the risks beforehand and make certain of the possible danger and sequent from its activities. So it takes effective protections and controls to prevent the happening of the possible danger and sequent in order to reduce the possible casualty of people, the loss of properties and the pollution of the environment. In the 21^st century, the oil companies all over the world have higher regards for the HSE management which has become the cosmopolitan trend and theme. The set up and the persistent improvement of the HSE management system have become the main direction of the management of the health, the safety and the environment in international oil companies. The auditing the HSE management system is standardizing. And the HSE management system is incorporating the quality management system. The environmental legislation in the world is more systemic and the environmental standard is stricter.After reviewing the development and changes of the HSE, according to some problems easily aroused from the current HSE management of the domestic chemical corporations, the environment safety expert system is proposed to be fused into the HSE management system, which can review and look through the policies and the activities during the implement of the HSE of the chemical corporation. That transforms the outer drive into the internal one, so it can keep on improving the HSE management of the chemical corporations.The general design and the architecture of the system are given in the following chapter, in which we also describe the whole structure of the expert system and the functionality and the effect of each component. The environment safety expert system has three possible inputs, which are the input of the administrator, the input of the sensors and the DCS, and the input of the experts, and one output, which is the output for the common users. The administrator input some historical records and some information related to the corporation that can’t be gained by the sensors or the DCS. The input of the sensors and the DCS is the dynamic change of the productive environment during the production. The input of the experts records the judgments of special cases by the experts, in the way of which the system gains the potential expert knowledge indirectly. The expert system records and learns from some historical records input by the administrator and the judgments of cases by the experts through some general intelligent models, and forms the knowledge library. Then, to the input data from the sensors or the DCS during the corporation production, the expert system can make a judgment with the built knowledge library and the model library in order to keep the production safe. Or else, the input environmental indexes of the corporation productive areas are judged and the environmental quality is estimated. These results are fed back to the common users to let them realize the current safety status of the production or the quality of the productive areas. That benefits the managers to discover the current problems of the corporation during the implementation of the HSE management system and improve the HSE management of the corporation further.The proposed expert system has the following main modules: the collection and the procession of data, the inference engine, the interpreter, the model library, the knowledge library, and the user interface. The collection and the procession of data is the procedure that the system gets the data from the sensors and the DCS at some interval. The procession work mainly includes the data filtering and the data smoothing. The inference engine consists of a set of programs, which are the scheduler, the executor, and the coordinator. The scheduler controls the whole flow of the inference. The executor executes the action selected by the scheduler and takes charge of reading the knowledge from the knowledge library and the model information from the model library. The main function of the coordinator is to modify the gained result in order to keep the consistence of the results after getting new data or new hypothesis. The interpreter is the program which interprets the procedure of the inference. It can record the questions from the users, and the inputs, the inference and the solution procedure of the system to construct the log of the expert system, which is convenient for the check, the maintenance, and the fix of the expert system. The model library includes five intelligent models commonly used in the expert system and the Boosting fusion model aiming at these intelligent models. What’s more, there are estimation models of the environment quality on the petrochemical corporation productive areas. The knowledge library is the set of the domain knowledge that can settle the problems, including the basic facts, rules, and some other related information. At last there is the inferface for the communication between the system and the users, the user interface.Developing the expert system on the environment safety makes use of the artificial intelligence techniques in the information science for references. The thesis summarizes five common intelligent methods in the field of artificial intelligence: the decision tree model, the neural networks model, the K-nearest neighbor model, the Bayesian model, and the Support Vector Machine model. The decision tree model is a method that approximates the discrete function. It’s robust to the noisy data and able to learn the disjunction expression. The neural networks consist of a series of simple densely inter-connected unit, which has a mount of real value inputs and produces a single real value output. The K-nearest neighbor model is a learning model based on the samples. It puts off the learning task to the time of the classification. That’s why sometimes it’s called ’the lazy learning method’. The Bayesian model is based on the following assumption: the variant to be examined follows some probability distribution, and we can make the inference and the optimal decision according to these probabilities and the observed data. The Support Vector Machine model is built on the VC theory of the statistical learning theory and the base of the minimum of the structural risk. It seeks the best trade off between the complexity of the model and the learning ability according to the limited sample information in order to obtain the best generality. These methods have advantages and disadvantages of their own while solving problems. A good consideration and a proper choice have to be taken when developing the expert system on the environment safety. For this sake, the proposed Boosting fusion algorithm gains the weights of each method during the fusion through iterative computations. It exerts the advantages of these methods while avoiding their disadvantages. The experiment shows that the performance of the fused system is better than any other system adopting only the single model.In the HSE management system, the evaluation of the environment quality is an important work. So a common integrative index method is first proposed to evaluate the environment quality of the production quarter. An index system for the production quarter is set up, which is the basis of the fuzzy integrative evaluation method. The key issue of the integrated index method is the fixing of the weights, which affects the accuracy of the integrated index method. The evaluation of the production quarter is a typical multiple rule decision problem. An evaluation model based on AHP-FUZZY incorporates the qualitative and the quantitative analysis, the evaluation of experts and scientific computation. With the instances, differences between the method based on AHP-FUZZY and the integrated index method are analyzed, which shows the advantages of the evaluation method based on AHP-FUZZY.At last, facing the chemical corporation, the demand is analyzed from three aspects, which are the explanation, the functionality, and the requirement of the system. The detailed description and the corresponding algorithm of the models, which haven’t been introduced particularly in the former chapters, are given. Meanwhile, some running dynamics of the system is shown.