RISK ASSESSMENT OF HAZARDOUS EMISSIONS

Abstract

Since the very first incarnation of human beings, risk has been a constant companion. Even more important fact is that risk has always been assessed using whatever tools were available with the humans. Daily risks have always been getting addressed using common sense. When their nature got complicated, along with the social complexities, a need for rational and scientific approach was felt for prospective and retrospective speculations. This dissertation is concerned with assessment of human health risk due to some untoward incident in process plants leading to an accidental release of toxic materials into the atmosphere. The Bhopal tragedy may be a typical example. Risk is commonly defined as some measure of the expected frequency of an untoward industrial incident: its probability of occurrence and the expected consequences. A chemical accident occurs all of a sudden, leaving no room for graduated response. It may be accompanied by one or more of the factors such as fire, explosion, release of toxic material and spillage of hazardous substances in storage, processing, transportation and general handling. Commonly, the causes could be equipment failure, design deficiencies, unsafe acts caused by human error, corrosion, abnormalities in operation or maintenance, and fire or other emergency in the vicinity. The health risk assessment of chemical emissions from industrial facilities requires mathematical modelling of a variety of processes including transport and fate of pollutants in and between various environmental media, population exposures to these chemicals, and their associated dose and health effects. (ii) Any framework for risk assessment requires at least : • Identification of potential for risk. • Estimation of exposure. • Characterization of risks. Risk and uncertainty are always interlinked. Another step for completeness of risk assessment is the expression, characterization and propagation of uncertainty in the analysis. Risk assessment plays an important role in environmental law. It can structure the regulatory decision making process by permitting agencies to set priorities and compare alternative decisions. It can also help focus public policy decisions about the appropriate levels of acceptable risks. The objective of this dissertation is the overview and advancement of risk assessment tools and evolution of a framework incorporating uncertainty for further decision making of environmental health' and ethics. The probabilistic aspects of the •• risk assessment have been adhered to. The thesis consists of six chapters. Chapter 1 contains general information. Some major accidents in the process industry have also been compiled. In probabilistic risk assessment (PRA), the very first step is the estimation of frequency of some untoward incidence, e.g. rupture of a vessel storing some hazardous material. Various methods for this have been discussed in chapter 2. Two models have been evolved for safety characterization of the system - a model for fault tree synthesis and the other for fault tree analysis. Fault tree analysis is one of the most advanced technique providing objectivity in the assessment process. The process becomes rather complicated as the complexities of design and operation of a process plant increase. Construction of fault tree from the system piping and instrumentation diagram is rather difficult, (iii) and even if tried, many important events may be omitted/Further, one of the goals achieved by the safety analysis of the system is-the increased understanding of the system. This chapter presents an automated methodology for fault tree construction using the graph theory. It is not a difficult process to construct the digraph of the system from its piping and instrumentation diagram, but it requires the detailed aspects of layout and working of the system, and hence provides the increased understanding of the system to the analyst. The algorithm presented in this chapter analyses the digraphs for tracing the feedforward and feedback loops and then constructs the fault tree by using Breadth First Search of the digraph. It provides a non trivial and objective solution. Analysis of fault trees requires data on the probability of failure of components in the system. Precise estimation of data on the probability of failure is not available. It makes the system amenable to the treatment by fuzzy set theory, which has been used in the algorithm for analysis of fault tree with replicated and complementary events. The algorithm uses hybrid approach - top down as well as bottom up. Top down approach has been used for repetitive events while the bottom up, for non repetitive events. Puff models for dispersion of pollutants provide a method for handling both types of releases - routine or accidental. Apart from the inability to effectively handle the accidental releases, other dispersion models available in literature suffer from two basic drawbacks - either they use rather restrictive assumptions (e.g. Gaussian models), or they are computationally inefficient (e.g. Higher Order Closure schemes, Large Eddy Simulation models). Complexities of atmospheric turbulence spectrum are not amenable to treatment by empirical approaches. However, with the development in the atmospheric boundary layer theory, new avenues may be explored for objective modelling of the dispersion (iv) phenomena. Chapter 3 of the thesis deals with the parameterization of atmospheric boundary layer and couples the same with a puff algorithm for passive pollutants discharged into the atmosphere. Anew model 'UNIPUFF' has been developed onthe same lines, for simulation of dispersion of pollutants due to regular/accidental releases. In this model, the stability of the atmosphere has been described as a function of physical height and Monin-Obukhov length scale (L). L, in turn has been determined by an implicit scheme using the lapse rate and wind velocity at the reference height. The puff is assumed to be of ellipsoidal shape. Six tracer particles located on the surface of the puff have been used to obtain its diffusion in a Lagrangian frame attached with the puff. The diffusive and advective properties have been obtained using the parameterization schemes of the atmospheric boundary layer for stratified flows. The effect of wind shear has also been incorporated. Rise of each puff is considered to be dynamic. Under certain assumptions, equations have been derived from the volume balance, and conservation of momentum and heat. The algorithm assumes the puff to remain in quasi-equiblirium during each time step. For deposition and chemical transportation exponential decay terms have been used. The results obtained are compared with field observations reported in literature and are found to be rather promising. Many materials in the process industries are stored under pressure and/or at lower temperatures. Consequent to the rupture of vessel, they may form heavier-thanair clouds. Dispersion in this case is entirely different as compared to the methods discussed in chapter 3. These clouds may be negatively buoyant and spread due to gravity in the initial phases. Various models available for such situations have been outlined in chapter 4. In such cases, even the incorporation of complicated mathematical descriptions do not always yield substantial improvement in the (v) prediction efficacy. Hence a simple model has been developed. The model evolves from the first principles and provides analytical and physically realistic treatment to the phenomena which have been empirically treated in most of the literature. The salient features of dispersion under these conditions are the gravity spread, entrainment, diffusion and heat transfer. The characteristics of the gravity front are normally assumed to be described by the densimetric Froude number (Fr), which is generally taken as constant; while some experiments indicate that this does not hold good, Hence, the velocity of the front has been determined using non stationary control volume approach. Entrainment has also been taken in parameterized form. Expressions have been derived for turbulence inside the cloud, its density and height. The turbulence has been determined by using atmospheric boundary layer theory, while its density and height, from thermodynamic considerations. A rational basis for determination of mode of heat transfer from ground to cloud has also been derived using atmospheric boundary layer theory. The model yields good agreement with the field data. After the fate of pollutants in the atmosphere has been characterized, the next step is to assess its toxic effects on human beings and more generally to the ecosystem. From biological perspectives, characterization of the effects is rather difficult and an uncertain process - errors of the order of thousand times are normal. In the first part of chapter 5 (which has two parts) various toxicological models have been discussed and finally a new approach named 'Fuzzy Probits' has been proposed. This approach is based on fuzzy regression of toxicological data. It obtains the optimal solutions rather than the deterministic ones. In the second part of this chapter the process of risk assessment has been discussed. After discussing various issues, including perception and acceptability aspects of risk, a framework has been proposed for propagation and characterization of uncertainty and for risk (vi) assessment. The framework uses the fuzzy numbers. Due to media coverage of major chemical accidents, the probability theory has suffered a serious setback, hence the possibility theory (Fuzzy Set Approach) has been used. The relative susceptibility of different age groups of population has been categorized using the expert opinion of medical professionals. The dissertation emphasizes on the use of fuzzy set theory in place of any other method for decision making under uncertainty. All the algorithms have been programmed using C++ programming language with object oriented paradigms. It achieves the modularity, flexibility and ease of revision/modifications. The last chapter of the dissertation (chapter 6) contains conclusions, recommendations and scope of future work.