I want someone to check the literature review of this paper. Proofread it and make it more appealing and understandable.

Introduction

In Oil & Gas Refinary Industry, decisions effects given by managers as can lead to immense profits or losses, are very broad and considerable. In other words, most of the equipment and resources used in the industry are enormously expensive, so if a situation that potentially could harm to human, environment, equipment, resource- property in general – happens, cost of compensation for that will be tremendous, therefore making and implementing decisions that can prevent or at least mitigate hazardous effects or save costs or decisions can lead to increased profits will be definitely valuable. Among the issues related to decision making in real world, uncertainty is one of the most important issues belong to that, thus risk and risk management have been developed. Various references have proposed several definitions for the word " risk " (Jannadi & Almishari, 2003)). As risk management principles and guidelines ( the project management body of knowledge or PMBOK and AS/NZS ISO 31000:2009 are most common guidelines in risk management(Dey, 2012) . However these guidelines describes overall tools can be used. Finding tools and techniques in order to risk identification, risk analysis and risk assessment are related to various factors and parameters. In this paper, Technique and Failure Modes and. Preliminary Hazard Analysis and its Potential Effects are two broadened tools and techniques in industries used for Risk analysis.

Preliminary Hazard Analysis Technique

Preliminary Hazard Analysis Technique was applied for the first time in the early 1950s A.D. in the United States to analyze the safety of liquid-propellant rocket. This technique was systematized by the country's aviation industry and was renamed by airline companies under its current name. Subsequently, the application of this technique was broadened to numerous industries such as chemical, nuclear, and other industries.

It is used when there is not any available information belong to detailed design and safety analysis for detecting hazards, factors caused them and evaluating effects and risk levels should be done.

Preliminary hazard analysis technique was known as a semi-quantitative analysis technique performed for the following objectives:

1. Identifying potential hazards and incidental events that may lead to an incident

2. Ranking the identified events based on their hazards

Determining the required controls for hazards and identifying corrective actions.

Failure modes and their potential effects

Failure modes and their potential effects are a systematic tool based on teamwork used to define, identify, evaluate, prevent, curb or control failures and their potential effects on a system(Gilchrist, 1993). Put differently, failure modes and their potential effects are an analytical method of risk assessment, attempting to identify and assess potential hazards existing in the areas where risk assessment and causes and effects. This is one of the most sophisticated and practical methods designed for identifying, classifying, analyzing failures, and assessing the leading hazards. By utilizing this method, failures can be eradicated and prevented from occurring.

Risk Assessment Matrix

The risk matrix is one of the structural approach for the risk management process which leads us to provide a response for measurement and controlling the main risks of the project while identifying the risks.(Paul R Garvey, 1998). In order to calculate Risk concerning, Risk Priority number(RPN) is used. The RPN consists of three factors: the probability of occurrence, the Impact of the consequence, and the frequency of occurrence. Each of these three factors is numbered from 1 to 10, and the higher number indicates more significant risk factors.

Factors Affecting Risk

The tables of these three factors are as follows:

Table 1_Probability Index List

Descriptive Expression	Probability Index
Permanent. Certainly	10
almost certain	9
Very likely	8
Likely	7
A little more than an equal chance	6
Equal chance	5
Less than equal chance	4
Unlikely	3
Very unlikely	2
Almost impossible	1

Table 2_ Impact Index Lists

Descriptive Expression	Intensity Index
Death	10
Permanent disability	9
Serious permanent disability	8
Partial permanent disability	7
Occupational absence for more than three weeks with recurrent disability	6
Occupational absence for more than three weeks and complete recovery afterward	5
Occupational absence between three days to three weeks and complete recovery afterward	4
Occupational absence for less than three days and complete recovery afterward	3
Slight damage without disability	2
No damage	1

So, the RPN would emerge as follows:

(1-1)

Fuzzy Logic and Fuzzy Inference system

Fuzzy logic is a newly approach. In this approach, the concept of membership function is defined for it instead of assigning an absolute value to a number, so that this makes it possible to model the method of non-explicit reasoning(Zadeh, 1988), which reduces the uncertainty of the real world and the dispersion of human information. Thus, fuzzy inference systems are defined and designed for that. Fuzzy inference system is a system by using fuzzy concepts such as membership degree etc. and using verbal variables, the descriptions and propositions of experts are converted to fuzzy variables. In addition to evaluating experts' propositions, inference and conclusion from the inputs are done.

1-3-1- Takagi–Sugeno fuzzy logic.

In this paper, the Takagi–Sugeno fuzzy logic is used in order to develop fuzzy inference systems. The introduction of Takagi–Sugeno fuzzy logic or Sugeno fuzzy logic was announced in 1985(Huo et al., 2001; Rezaee et al., 2020; Rouse & Billington, 2007) . Since Sugeno fuzzy logic is considered as the modified version of the Mamdani model since both of them involve the same fuzzification process for inputs, and output membership function (MF) is the only difference between them. For outputs of the Takagi–Sugeno type, the MFs are either linear or constant functions. The latter is known as a Sugeno fuzzy model of the zero order that can be defined as:

(1-2)

where and represent fuzzy sets describing the MFs of and , respectively. The constant k is a crisp value in the consequent. Thus, the output of each rule can be considered as a spike. Addition and multiplication are the operators of aggregation and implication, respectively. The rules below are formulated for a Sugeno fuzzy logic model of the first order:

(1-3)

where and represent fuzzy sets in the antecedent, and , and are all the constants. The input data are used to define the location. It results in the optimal performance of the system in the successful adaptation to any change in the input data. The models of higher order are possible, but they become more complicated and disadvantageous(Al RawiI & Kheder, 1990; Jang, 1993; Topcu & Sarıdemir, 2007; Topçu & Sarıdemir, 2008).

Stochastic process and Markov Discrete-Time Chain

A stochastic process is a model with probabilistic trait that have been variously developed for systems which evolving randomly in time. (Kulkarni, 2011). Based on observing continuously and non continuously, it is categorized to a continuous-time stochastic process.and a discrete-time stochastic process

A stochastic process is a Markov chain if we have for sequence of random variables , , , … , :

(1-4)

The possible values for make several sets, which are called state space.

The Markov chain is often represented by a sequence of directional graphs, in which N-graph ridges are labeled by the likelihood of going from a mode in time N to other modes in the time n + 1. The same information is inserted in the transfer matrix from time N to time N + 1 where the row i and column J element is the possibility of changing the status of i-1 mode to j-1 mode. The sum of elements of each row equals one but not necessarily the sum of elements of each column. In the matrix theory, if all elements of the matrix are unnegated and the sum of elements of each row is one, in this case, the matrix is called the Markov matrix. For such matrices, we can build a sample space so that the matrix elements, the possibility of changing the status of all the contingencies of the sample space, and then define a Markov chain for the sample space. However, Markov chains are often assumed to be uniformly in time; in this case, the graph and matrix are independent of N, and thus not provided a sequence.(Norris, 1998)