Evaluation of Human Factors Risk and Management in Process Safety in Engineering
Introduction
With the growth of the economy and the development of science and technology, there has actually been a significant decrease in the number of deaths and accidents. However, the safety situation is still grim. In an analysis of accident-induced factors, human factors led to the vast majority of accidents. This chapter through investigation found that some 70%-80% of aviation accidents, 60% of petrochemical accidents, 90% of iron and steel metallurgy accidents, and 90% of road traffic accidents are caused by human factors (Cai et al., 2008). Therefore, human factors are the main factors leading to accidents. Many scholars have applied a wide variety of research methods to analyse and study human factors resulting in accidents. Early human factors are confined to the individual’s perspective. Farmer and Chamber introduced the concept of the accident-prone tendency, suggesting that some people have personal traits such as clumsiness or carelessness that easily lead to accidents (Tarlor et al., 2001). In 1931, Heinrich's book Industrial Accident Prevention he expounded his accident causation theory. Heinrich suggested that human ancestry and the social environment caused people to make errors. A mistake by a person can lead to an unsafe action, which leads to the occurrence of an accident (Heinrich et al., 1980). In 1972, Wigglesworth proposed a new chain of accident causation. He believed that, because of a lack of knowledge and education, people will make mistakes and cause accidents. Subsequently, Bode and Roose-fort introduced the concept of “management” into the cause of an accident (Fu, 2013). Stewart (2002. 2011) divided “safety management” into two levels, which are more specific given the cause and root causes of an accident. From the development course of an accident cause chain, the human factor has evolved from an individual level to an organizational level. At the same time, scholars have also studied a range of individual human factors. Research by Kumar (2016) shows that if there are no emergency rescue measures, emergencies will be difficult to deal with. Harvey (2016) believes that if companies do not develop safety rules and regulations, then employee awareness of safety will be weak, thereby increasing the probability of an accident. Konstantin (2010) believes that a good safety culture can improve employee safety awareness and safety skills, thereby reducing the incidence of accidents. Nie et al. (2016) claims that appropriate psychological counselling and a reasonable amount of rest time can effectively improve the working status of employees. Although the previous research has been of great help in improving the risk of accidents involving people, it is still confined to the study of individual factors.
Assessment of Human Factors
With the increasing amount of research on human factors, a series of human factors analysis models have been developed. Among them, the Human Factor Analysis and Classification System (HFACS) is the most widely used and accepted method. The HFACS was developed by Shappell and Wiegmann based on Reason’s “Swiss- Cheese” model and was applied in aviation accident analysis (Shappell and Wiegmann, 2001,2003, 2004; Wiegmann and Shappell, 1997,2001a,b,c, 2003). The framework is valuable in the analysis and classification of the human factors involved in accidents. Currently, HFACS is applied in different fields and has been applied effectively. For example, Celik and Cebi (2009) investigated the human factors in ship accidents using the HFACS. Similarly, Daramola (2014) used the framework to investigate air accidents in Nigeria during the period from 1985 to 2010. Australian scholars, Patterson and Shappell (2010), conducted an analysis of 508 human-induced mine accidents in the Queensland area using the HFACS. Michael used the HFACS to analyse 263 coal mine accidents in Australia between 2007 and 2008 (Lenne et al., 2012). Ruth Madigan applied the HFACS to line incidents in the UK rail system (Madigan et al., 2016) and Christine Chauvin analysed collisions at sea (Chauvin et al., 2013). Soner et al. (2015) used the HFACS-FCM in fire prevention modelling on board ships. Chen et al. (2013) used the HFACS-MA framework for maritime accident investigation and analysis. Zhan et al. (2017) used the Human Factors Analysis and Classification System-Railway Accident (HFACS-RAs) framework to identify and classify human and organizational factors involved in railway accidents. It is clear, therefore, that the HFACS has made an important contribution in the analysis of accidents and has provided a range of important research results. However, from the application point of view, most of the studies involve a qualitative analysis and occur after the accident. The safety management of human factors is an important means to prevent accidents. Methods for realizing risk assessment and management of human factors in daily production has still not been addressed. In view of this, a new method for human risk assessment and management was proposed in this chapter. Based on the HFACS framework, a human factors risk assessment model was established. Using the assessment model and the set pair analysis method, this case study calculates the connection number to get the safety score, risk level and risk development interval of each factor. The risk development trend of each factor is studied by using the partial connection number. Using the established SPA-Markov chain risk prediction model, the future development of risk is predicted. The ABC analysis and “S-O-R" model were used for human risk management to address the safety management link. The application results show that this method can effectively improve the safety of human factors. Finally, the common unsafe human factors and accident paths are summarized, and effective safety stimulus measures for different types of human factors were introduced.
Human Factors Risk Assessment Model
The HFACS framework is an important factor analysis method in accident investigation and is now used in aviation, power, mining and other areas of accident investigations. It was developed from Reason’s “Swiss-Chess” model of human error (Reason, 1990) and provides an organizational framew'ork for accident analysis. Therefore, the FIFACS addressed human errors and divided them into four different levels (Shappell and Wiegmann, 2003). Level 1 (Organizational Influences) includes three factors: resource management, organizational climate and organizational process. Level 2 (Unsafe Supervision) includes four factors: inadequate supervision, planned inappropriate operations, failed to correct a known problem and supervisory violations. Level 3 (Preconditions for Unsafe Acts) includes seven factors: physical environment, technological environment, physical/mental limitations, adverse mental states, adverse physiological states, crew resource management and personal readiness. Level 4 (Unsafe Supervision) includes four factors: decision errors, skill-based errors, perceptual errors and violations. In modern society, most enterprises place greater emphasis on safe production, and on taking many safety measures. Therefore, in such contexts unsafe acts may not lead to accidents. For example, if the unsafe operation of pow'er systems leads to leakage, the relay protection system will be quickly shut down to protect the whole power system. In the chemical industry, when a tank leaks, there are often emergency measures to prevent accidents. However, when the final defense system fails (emergency failure), this usually results in an accident. Based on this case, the revised accident model was developed. The model is described diagrammatically in Figure 18.1. According to the revised accident model and the HFACS framework, the human factors risk assessment model was established. This model includes 5 levels (organizational influences, unsafe supervision, preconditions for unsafe acts, unsafe acts and emergency failure) and 25 human factors, as shown in Figure 18.2.
Set Pair Analysis (SPA)
The set pair analysis method is a system research method proposed by Zhao (1989). In this method, the certainty and uncertainty between systems are studied from three angles of “identity-discrepant-contrast”. It has been applied to several fields. Chong et al. (2017) applied it to coal mine hazards. Su et al. (2009) used the SPA method to assess urban ecosystem health. Li et al. (2016) conducted a risk assessment of water pollution based on integrated к-means clustering and the setpair analysis method. Wei et al. (2016) used it to predict the analysis model of integrated carrying capacity. Tao et al. (2014) applied it to perform the multifunctional assessment and for the
FIGURE 18.1 Accident causation model (Reproduced with permission from Xuecai & Deyong 2018 Copyright © Elsevier)
zoning of a crop production system. In a system, two sets, A and B, which have a certain relationship, form a set pair H (A. B). Assuming the set has N characteristics, where 5 is the total number of identical characteristics, P is the total number of contrary characteristics and F = N-S-P is the total number of discrepancy characteristics. The connection degree “p” was given by Zhao (2000a,b) as follows:
Definition 1.
A = SIN, b = FIN, c = PIN represent the degree of “identity”,
“discrepancy" and “contrast”, respectively, and a + b + c = 1. In this connection degree, / is the uncertainty coefficient, / e [-1,1], and j is a contrast coefficient that usually equals -1 (Zhao, 2000 a,b).
Definition 2.
When considering the weight, the contact degree is as follows:
FIGURE 18.2 The human factors risk assessment model (Reproduced with permission from Xuecai & Deyong 2018 Copyright © Elsevier)
where W is the weight coefficient vector, R is the “identity-discrepant- contrast’’ assessment matrix, and E is the connection degree matrix (Zhao, 2000 a,b; Chong et al., 2017).
If the bj terms expand, then the multiple contact number is
When n = 3, the five-element connection number is
In the application process, the five-element connection number can be associated with the level of risk; a represents the ratio of risk V, b represents the ratio of risk IV, c represents the ratio of risk III, d represents the ratio of risk II, and e represents the ratio of risk I.
Definition 3.
W = (w„ w2, . . wn) represents the factor vector matrix. U = {p,, ц2> • • •. pn} represents the risk factor set. According to Formula (18.2), the five- element connection number expression formula is obtained as follows (Zhao, 2000 a,b; Zhou and Zhang, 2013)
Definition 4.
Since /g [-1,1],у = -1, when / = 1, the system safety is the highest; when / = -1, the safety of the system is the lowest. The degree of contact of the system is p g [a - b - c, a + b - c]. In this case study, according to the “sharing principle”, the author divided the risk into 5 grades (Table 18.1).
Risk I represents the highest risk, and risk V represents the highest safety.
Definition 5.
Risk levels l-V were given the scores 1, 3, 5, 7, and 9 points (Table 18.1).
The safety score is U = 9a + 7b + 5c + 3d + 1 e.
Risk Trend Analysis
In the SPA method, the partial connection number is an adjoint function of the connection number. Using the partial connection number, the development trend of the system can be predicted (Zhao, 2005; Wu. 2009).
TABLE 18.1
Risk Classification Table
|
Risk level |
I |
II |
III |
IV |
V |
|
Risk change |
|
|
|
|
|
|
Contact degree |
-l |
-0.6 |
-0.2 |
0.2 |
0.6 |
|
Score |
1 |
3 |
5 |
7 |
9 |
|
Scoring interval |
[1.0. 2.6] |
[2.6,4.2] |
[4.2. 5.8] |
[5.8.7.4] |
[7.4. 9.0] |
(Adapted with permission from Xuecai et al. 2018 Copy right © Elsevier)
Definition 6.
For multiple connection numbers p= a + + b2i2 + ... + b„i„ + cj, its 1st
order partial connection number is as follows:
where da = ——,5. =—-—,ды =———= ^"~2 a + b, b{+b2 ' b2 + by ' b„_ 2 + c
Definition 7.
For multiple connection numbers p= a + b,/, + b2i2 + ... + bnin + cj, its 2nd order partial connection number is as follows:
u з9 CY/ ■> ЭЬ| -o. db„.3
where u =-,<9"b: =-!—,...,S‘b„_3 =-—-
ca + db дЬ+дЫ db„.y+Bb„-i
Definition 8.
For multiple connection numbers p= a + b^+ b2i2+ . . . + b„i„+ cj, its n - 1-order partial connection number is as follows:
ля—2
where d"~'a = —2
d"-2a + d"-
SPA–Markov Risk Prediction Method
In the production process, human factors are uncertain and change dynamically. Therefore, to predict the human factor risk, a prediction model must be chosen that can handle the characteristics of humans. The Markov chain has the discrepancy of time and state. This characteristic is consistent with the changing law of human factors and can be used to predict the risk of human factors. In this case study, a human factor evaluation model based on the SPA method and the Markov chain was proposed and applied.
Definition 9.
Assume at time T that the contact pojytfjhgdegree is
Where
After taking safety measures, the safety status change during time [t, t+T], where T is a safety management cycle. In this period, A(/,) holds the a level, A(f2) becomes the b level. A(t}) becomes the c level, A(r4) becomes the d level, and A(r5) becomes the e level.
Then, the transfer matrix vector in time [t, t+1] is:
Similarly, the transfer vector matrix of b(t), c(t), d(t), and e(t) can be obtained. Thus, the transfer matrix of the system during the time [t, t + T] is
Based on the state transition probability matrix determined above, the evaluation model of human factors is established by the degree of association. That is, at t + T, the safety factor for human factors is
Assuming that the state transition probability matrix of the safety evaluation index is the same between each change period T. that is, the state transition probability matrix P is a constant matrix, the safety assessment value of the human factor after n change periods is
The formula satisfies the C-K equation (C-K, Chapman-Kolmogorov). According to the trail of the Markov chain, P',T will stabilize as the management cycle increases. Therefore, the steady-state values of human factor risk can be obtained from the following equations:
