The Full Lifecycle Evolution Model of Accidents: A Case Study of Underground Metal Mines in China

Abstract

Analyzing the mechanisms of accidents is essential for clarifying the accident evolution process, devising preventive measures, and achieving proactive accident management. To address the potential issues in existing accident causation theories, such as the unclear distinction between direct causes and incomplete accident evolution pathways in enterprise-level accident prevention analysis, this study systematically reviewed the elements involved in safety management activities and their interrelationships. We identified the central role of human factors in the accident evolution process and developed a full lifecycle evolution model for industrial accidents, which begins with hazard identification and follows a safety management logic as its primary framework. This model provides a clear pathway for constructing enterprise-level risk control lists and accident prevention schemes. The model’s effectiveness was validated through its application to China’s underground metal mining industry. Drawing on Chinese laws and regulations as well as accident investigation reports, this study identifies 11 common types of accidents in underground metal mines and maps their evolution pathways from a complex systems perspective. Quantitative data from 61 accident reports were used to pinpoint the core factors and critical pathways leading to these various accidents. The study also analyzes prevention strategies and proposes new countermeasures to control the propagation of accident risks. Practical applications of the model demonstrate that emphasizing human factors enhances the effectiveness and accuracy of enterprise-level accident analysis and risk management.

1. Introduction

Accident prevention is a core aspect of safety management in production enterprises. Both understanding the mechanisms behind accidents and effectively controlling causative factors are crucial for preventing and minimizing industrial accidents [1]. Accident causation theories and models are used to describe the relationships and interactions between accident causes and the occurrence of industrial accidents. These models provide a theoretical framework for tracing accident causes, understanding accident progression, formulating preventive strategies, and assessing accident risks [2,3]. Currently, the application of accident causation theories predominantly focuses on analyzing the causes of accidents and developing targeted improvement measures based on the analysis results [4,5]. For instance, Yali Wu et al. [6] utilized the 24Model to examine a major hazardous chemical explosion, identifying its primary causes and proposing improvement measures. Read et al. [7] employed AcciMap to identify and analyze 22 automation-related accidents in the aviation and rail industries. The study classified contributing factors using a network approach, revealing similarities and differences between traditional models and emerging automation-related traffic accidents. Accident causation theories have also been widely applied in accident prevention efforts across various industries, such as construction [8], transportation [9] and maritime shipping [10].

Accident causation theories can be categorized in several ways based on different classification standards [11]. In this study, accident causation theories are classified into two types, chain-based (modular) causation theories and system-based (non-modular) causation theories, depending on whether they explicitly categorize causative factors [12]. The chain-based accident causation theory views accidents as a linear chain of events, where each link represents an independent causal factor (such as unsafe acts of people or unsafe conditions of objects) acting in a specific sequence. In contrast, the system-based accident causation theory considers accidents as emergent phenomena resulting from the dynamic interaction of multiple elements within a complex system, emphasizing non-linear interactions such as management deficiencies, organizational factors, and technical interdependencies. These theories have different scopes of application. Qiao et al. [13] compared the two models in terms of their features, analysis processes, and results. They found that non-modular models provide more comprehensive and systematic analysis results, though their complexity limits their use in large-scale statistical analyses. In contrast, modular models are more structured, enabling the efficient statistical analysis of numerous accidents. The choice of causation model in safety research and practice should consider both adequacy and efficiency [14]. However, in practice, current accident causation theories face certain limitations, particularly in identifying widespread safety risks and developing preventive measures for enterprises [15,16].

As a whole, production enterprises may encounter a wide range of accident types and hazards. With the continuous advancement of production technology, the increasing use of new equipment and techniques further complicates the identification and analysis of enterprise-level accident risks [17]. System-based accident models require a high level of expertise and accurate data for analysis, yet data collection often presents significant challenges in real-world applications. Consequently, system-based models have not been widely applied in this context [18]. Chain-based models are more effective in general scenarios, and ongoing updates have refined the feedback relationships and dynamic evolution processes among the factors involved. However, applying chain-based models to enterprise-wide accident causation analysis can be difficult due to a lack of procedural guidance and real-world case studies, resulting in challenges in merging similar factors across different accident types [19]. This can lead to a redundancy in control factors, an increase system complexity, and ultimately reduce the accuracy of key factor analysis at the enterprise level [20]. These issues arise primarily from unclear distinctions between common factors across different accident types.

Most existing accident causation theories assume that organizational or managerial factors are indirect or higher-level causes of accidents [21], whilst unsafe human behaviors and unsafe conditions are considered direct causes. However, the relationship between human factors and unsafe conditions, as well as their roles in accidents, remains a topic of ongoing debate [6]. Heinrich [22] first proposed that accidents are caused by unsafe behaviors and conditions, attributing unsafe conditions to human error. The energy transfer theory, on the other hand, emphasizes the fundamental role of unsafe conditions of energy in accidents [23]. The loss causation model (LCM) views accidents as failures in control measures, describing the process of accident development [24]. However, it is important to note that humans, as the only active entities in the accident system, influence all other factors, including the environment, equipment, and management systems. Reason [25] emphasized the central role of human factors in the accident process with his Swiss cheese model (SCM). However, the SCM lacks descriptions of the role of material factors in accident evolution. To address this issue, some models have introduced improvements. The 24Model [26] considers both unsafe behavior and unsafe conditions as direct causes of accidents. Liu et al. [27] explored factors influencing the choice of safety analysis methods, while Grant et al. [28] extracted 15 core system thinking principles to analyze the relationships among accident causation factors. However, these improvements either apply to specific industries or still suffer from the lack of clear distinctions between human and material factors, leading to a redundancy in causative elements across accident categories. A standardized and comprehensive enterprise-level accident analysis framework is still lacking [29]. Therefore, it is necessary to re-evaluate the logical relationships and evolution processes between the various elements of accidents, with particular emphasis on the interaction between unsafe human behaviors and unsafe conditions.

Additionally, existing theories still have limitations in analyzing the consequences of accidents [30]. Many scholars have conducted in-depth research on the relationships between safety measures and accident consequences, revealing how deficiencies or inadequacies in safety measures at various stages accelerate accident escalation [31,32]. For example, Wang et al. [33] analyzed dynamic coupling in chemical processes and devices using the hazard and operability study model. They quantified the likelihood and severity of chemical accidents, validating the dynamic risk evolution trend with the BP refinery explosion in the U.S. Recal and Demirel [34] applied multiple machine learning algorithms to classify accidents of different severities, identifying measures to reduce accident severity. Joe-Asare et al. [35] used human factors analysis and classification system (HFACS) to classify and code accident causation in 701 mining accident reports from Ghanaian gold mines, identifying significant relationships between accident severity and unsafe behaviors. Wang et al. [36], Pinto et al. [37], and Zhang et al. [38] conducted studies on the potential impact of unsafe behaviors on the severity of accidents in maritime traffic, construction, and public transportation, respectively. However, current accident causation theories rarely delve into the relationship between various accident causes and the severity of outcomes. Typically, accidents are treated as singular events, with the outcomes often summarized as “Injuries”, as seen in the accident causation chain theory. Furthermore, the research on how different safety measures, influenced by multiple factors, jointly affect the development of accidents remains insufficient.

In summary, current accident causation models face limitations in analyzing the evolution pathways of accidents and identifying critical risks within large enterprises. Additionally, there is a lack of research on the staged progression of accident outcomes and the corresponding control measures once an accident has been triggered. Consequently, this study focuses on the hazard source as the initial point, using personnel safety management activities as the foundation to comprehensively analyze the potential evolution of industrial accidents. By thoroughly understanding the core principles and advantages of existing accident causation models, this study aims to establish accident evolution logic and clarify the pivotal role of personnel in corporate safety management. Furthermore, it re-evaluates the interactions among management systems, unsafe conditions, and human factors. The proposed model offers a phased analysis of accident-induced losses, resulting in a conceptual framework for accident development and a dynamic evolution pathway. This structured model provides theoretical guidance for improving accident prevention in enterprises and establishes a standardized, stable, and practical framework for developing enterprise-level safety management strategies grounded in accident causation theories.

The rest of this paper is structured as follows: Section 2 reviews and compares the core concepts and main characteristics of typical accident causation theories. It analyzes the interactions between control factors during accident evolution, thereby establishing the main components, hierarchical structure, and application pathways of the full lifecycle evolution model for production accidents. Section 3 uses the Chinese underground metal mining industry as a case study to explore the specific applications of this model in accident analysis and the formulation of safety measures. Section 4 analyzes the results of the case study, discusses existing problems, and provides insights into the model’s limitations and future directions for application.

2. Methodology

2.1. Safety Management Activity Analysis

2.1.1. Comparison of Major Accident Causation Theories

The development of accident causation theories over the past century has reflected the evolution of safety management concepts and control elements.

Table 1. Comparison of major accident causation theories.

Model	Proposer	Core Concept	Contributions	Elements	Representation of Consequences
Domino Theory	Heinrich [22]	Accidents result from a chain reaction of multiple factors, including heredity, social environment, personal defects, unsafe behavior or conditions, and injury.	Introduced the safety pyramid and chain accident causation model.	Social environment, human errors, and unsafe conditions.	Injuries.
Energy Transfer Theory	Haddon [23]	Accidents involve unwanted or abnormal energy transfer, and preventing accidents requires preventing unintentional energy release.	Defined the physical nature of accidents from an energy perspective.	Energy sources and transmission paths.	Injuries.
Loss Causation Model (LCM)	Bird and Loftus [24]	Accidents result from management errors leading to unsafe behaviors or conditions, which ultimately cause accidents and losses.	Highlighted the importance of management control in accident prevention.	Management failures.	Losses.
Swiss Cheese Model (SCM)	Reason [25]	Accidents result from multiple layers of defense failure, occurring when several defenses fail simultaneously.	Visually demonstrated the multi-factor nature of accidents and the layered defense structure, emphasizing organizational defects.	Organizational influence, unsafe supervision, and unsafe behavior.	Not explicitly reflected.
AcciMap	Rasmussen [39]	Analyzes accident causes from a multi-level socio-technical system perspective, focusing on interactions across layers.	Provided a socio-technical perspective, emphasizing the multi-level interactions that influence accidents.	Government policies, regulatory agencies, company management, technical and operational management, employees, and equipment.	Not explicitly reflected.
Human Factors Analysis and Classification System (HFACS)	Shappell and Wiegmann [40]	Analyzes accident causes from a human factors perspective, including organizational influence, unsafe supervision, and unsafe behaviors.	A systematic, layered model for analyzing accidents from a human factors viewpoint.	Organizational influence, supervision.	Not explicitly reflected.
Functional Resonance Analysis Method (FRAM)	Hollnagel and Goteman [41]	Accident causes result from variability in the functions of complex systems, where accidents are the outcome of system dysfunction.	Emphasized the interdependence of system functions and is applicable to complex systems analysis.	Functional dependencies between system functions.	Not explicitly reflected.
Systems-Theoretic Accident Model and Processes (STAMP)	Leveson [42]	Accidents occur due to failures in control structures and feedback mechanisms within a system.	Highlighted the role of control and feedback mechanisms in ensuring safety: suited for complex systems.	Control structures, feedback mechanisms, and system boundaries.	Not explicitly reflected.
24 Model	Fu et al. [43]	Divides accident causes into two types and four stages, covering human, material, managerial, and environmental aspects.	Integrated various theories, emphasizing the combined influence of human, material, and managerial factors on accidents.	Human factors, material factors, and management failures.	Not explicitly reflected.

summarizes the core ideas and contributions of the major accident causation theories [1,19].

2.1.2. Analysis of Accident Causation Factors

The evolution of accident causation theories has progressively broadened the scope of control factors, extending from unsafe behaviors and unsafe conditions to include management systems, societal influences, and legal regulations. Based on a comprehensive review of existing theories and practical case applications, this study organizes and categorizes accident causation factors, forming the logical structure shown in Figure 1.

Below is a detailed analysis of the factors and their relationships:

• Human factors: Since the accident proneness theory, human error has been regarded as a critical factor in accidents. Heinrich’s accident causation chain theory posits that both unsafe behavior and unsafe conditions stem from human errors. In the SCM and HFACS, human factors occupy a central position in accident causation. Human factors are crucial because humans, as the only active agents in the production environment, influence all other factors, including the operation of machinery, any improvement to the working environment, and the establishment of management systems.
• Unsafe conditions: This factor includes all physical entities involved in production activities, aside from personnel. These can be further subdivided into equipment, materials, and the environment. Equipment refers to machinery and systems used to enhance production efficiency. Materials are man-made objects, aside from equipment, that lack power. The environment includes both natural and man-made settings. Changes in environmental factors can affect human behavior and tool performance, potentially triggering accidents.
• Management factors: In this study, management factors mainly refer to enterprise-level systems. Enterprises establish safety management systems in compliance with relevant laws, regulations, and standards to prevent safety management deficiencies and foster a positive safety culture [44].
• Legal factors: Laws and regulations related to safety management are enacted by each country based on its specific circumstances. These provide binding guidelines for government, society, and enterprises and serve as the primary foundation for enterprises’ safety management systems.
• Government factors: Government agencies oversee enterprise safety by implementing policies and supervising production safety. They also provide resources and support in the event of major accidents.
• Societal factors: society plays a supervisory role in ensuring production safety through public scrutiny and media oversight.

2.1.3. Accident Causation Logic

(1)

Role Classification in Safety Management

Personnel involved in safety management can be divided into three categories based on their interaction with other factors: managers, supervisors, and operators.

• Operators: Primarily interact with elements that may pose unsafe conditions, including equipment, materials, and the production environment. They are responsible for implementing specific safety measures.
• Supervisors: oversee the actions of operators to ensure compliance with safety regulations and operational procedures.
• Managers: ultimately responsible for production safety, managers establish and revise safety management systems in accordance with laws and policies.

Supervisors can be hierarchical, with higher-level supervisors monitoring lower-level ones. The highest level of supervision lies with the managers.

(2)

Enterprise-Level Safety Management Logic

The safety management system is developed by managers in accordance with laws, industry operating procedures, and the specific needs of the enterprise, serving as the foundation for safety management activities within the enterprise. This system directly impacts the human factors by guiding, constraining, and standardizing supervisory and enforcement behaviors. Personnel manage hazards present in the production process by adhering to established regulations, procedures, and operating standards, thereby ensuring their own safety.

(3)

Accident Causation Logic

The energy transfer theory elucidates the physical nature of production accidents, positing that the source of these accidents lies in the release of energy from hazards present in the workplace, including environmental factors, equipment, and materials. Human errors or a failure to diligently adhere to management systems can result in these hazards being in an unsafe condition. If personnel fail to effectively control this unsafe condition, it may lead to the unintended release of energy from the hazards, ultimately resulting in production accidents. Consequently, the logical chain for tracing the origins of accidents can be expressed as follows: production accident → failure to manage specific hazards → human oversight → breakdown in supervision → gaps in the regulatory framework → external factors (government oversight, societal supervision, or legal loopholes).

2.2. Full Lifecycle Evolution Model of Accidents

Building on the aforementioned safety management and accident causation logic, along with the accident development process, this study constructs the full lifecycle evolution model (FLEM), illustrated in Figure 2. It is essential to note that the primary responsibility for accident prevention and safety management rests with enterprises, while external factors primarily offer oversight and emergency resources. Consequently, this model emphasizes how enterprises manage accident risks, intentionally excluding the roles of government and society at this stage.

2.2.1. Accident Evolution Process

(1)

Hazard Identification and Control

Production accidents originate from various hazards present in the production environment, which are managed by personnel in accordance with the safety management system. From a preventive standpoint [19], effective control methods for identified hazards include elimination (M01), substitution (M02), engineering control (M03), monitoring (M04), and warning (M05). A failure in the management system refers to inadequate or ineffective management practices, policies, or procedures that result in accidents or incidents. Such failures can occur at different stages of the execution process, including the design, implementation, and maintenance of control measures [45]. This also involves supervision by safety management personnel at various levels, as well as the evaluation and improvement of control measures.

(2)

Accident Development Stages

When supervisors discover through routine inspections that operators are not following established procedures, the parties responsible must take appropriate corrective actions. If these actions are inadequate, safety hazards may arise. If such hazards are not effectively addressed over time, and supervisory personnel fail to perform their oversight duties in subsequent inspections, the safety risks will continue to accumulate, eventually triggering the onset of a production accident.

When hazardous energy is released due to insufficient control, the first step is to determine whether the energy is coupled with personnel. If no such coupling occurs, the event is classified as a near miss (or no-injury incident). If the energy does interact with personnel, the next consideration is whether the protective measures in place are effective. If these protective measures fail to mitigate the harmful energy, injury becomes inevitable, escalating the event into an actual accident. The severity of the accident’s consequences will depend on the effectiveness of the on-site response and emergency rescue measures. The accident will conclude once the hazardous energy is either fully dissipated or brought back under control through appropriate actions.

(3)

Accident Consequence Analysis

Currently, the analysis of accident consequences typically focuses on three areas: economic losses, personnel casualties, and social reputation. Accidents can result in significant direct and indirect economic losses, and even near-miss incidents can reduce productivity. Personnel casualties are a critical metric for evaluating accident severity and distinguishing near-misses from confirmed accidents. Social reputation is also impacted, as major accidents often damage an enterprise’s public image, eroding public trust and market share. The analysis of accident outcome severity requires a comprehensive assessment of several factors, such as the magnitude of the hazardous energy, the number of individuals exposed at the time of release, and the effectiveness of emergency response measures. Different types of accidents may require distinct logical frameworks for determining their outcomes, and further specification is needed based on the specific accident scenario during practical application.

2.2.2. Prerequisite Conditions for Unsafe Behaviors

Unsafe behaviors of personnel can lead to the failure of hazard control measures [18]. This study categorizes the causes of unsafe behaviors into two main factors: psychological (subjective) and physiological (objective).

(1)

Psychological Factors:

• Fluke mentality: workers may hold the mistaken belief that occasional non-compliance with safety procedures will not necessarily lead to accidents, leading to unsafe behaviors.
• Complacency: overfamiliarity with the work environment and processes can cause a gradual decline in vigilance, making workers blind to potential safety risks.
• Herd mentality: when workers observe others engaging in unsafe behaviors without consequences, they may imitate these actions.
• Rebellious mentality: workers may develop resistance to safety management regulations or supervisors’ demands, deliberately engaging in unsafe behaviors as an act of defiance.
• Adventurous mentality: in pursuit of excitement or to complete tasks quickly, some workers may choose to take dangerous risks [46].

(2)

Physiological Factors:

• Fatigue: prolonged work hours or high-intensity labor can lead to physical fatigue, slower reaction times, and reduced attention, increasing the risk of unsafe behaviors.
• Physical discomfort: factors such as lack of sleep, hunger, or illness can impair a worker’s focus and job performance.
• Environmental impacts: high humidity, elevated temperatures, loud noise, and other environmental factors can negatively affect workers.
• Lack of safety skills or competence: insufficient safety training may leave workers without the skills to handle emergencies or operate equipment correctly, increasing the likelihood of errors during hazardous tasks.

2.2.3. Responsibility of Managers

At the highest level of accident analysis within an enterprise is the institutional layer. All preconditions that may lead to unsafe behaviors by personnel stem from either poor design of the safety management system or ineffective implementation of safety policies. In this model, organizational and management factors are not further subdivided, as safety systems are established specifically to ensure the safety of personnel and control hazards. Thus, when the analysis reaches this level, the causative factors manifest as either incomplete or inadequately enforced safety systems. The fundamental solution is to revise these systems to improve the execution and maintenance of safety controls.

2.3. Application of the FLEM

FLEM primarily has two applications: accident cause analysis and the formulation of enterprise-level preventive measures. The application process and interaction relationships are illustrated in Figure 3.

Application method I: Accident traceability based on investigation reports is the first application method, which involves identifying the hazard sources that led to the accident, categorizing the control measures, and analyzing their failure mechanisms. Additionally, this approach allows for the evaluation of the response at various stages of accident progression, further tracing the preconditions of unsafe behavior and ultimately identifying targeted control measures and recommendations for improvements in management systems.

Application method II: systematic analysis of enterprise-wide accident evolution pathways, with the following primary steps:

(1)

Hazard source identification: identify potential accident types and the specific categories of hazard sources that could lead to such accidents within the enterprise, based on case studies, the relevant literature, or industry consensus.

(2)

Pathway mapping and construction: Based on legal regulations, industry standards, and operational procedures, determine the specific control measures for each hazard source and establish corresponding control systems at different stages of accident evolution. This forms the enterprise’s accident evolution pathway, where non-compliance with any part of the system represents a potential hazard.

(3)

Accident case statistical analysis: Map the causes identified in the individual accident analysis process to their corresponding points in the enterprise’s accident evolution pathway. Through statistical analysis of the frequency at each node, identify the critical weak points that require focused management and develop improvement measures or institutional adjustments to enhance the overall safety management level of the enterprise.

(4)

Model refinement and improvement: continuously improve the enterprise’s accident evolution pathway by repeating steps (1)–(3) as new production accidents occur during the implementation of new control measures.

Method I primarily focuses on post-accident causality analysis, which aligns with the primary application of existing accident causation theories for analyzing single accident types. Method II, however, adopts a proactive approach by mapping enterprise-wide risk factors from the source and is more geared towards developing comprehensive accident risk control plans for enterprises.

3. Case Study

3.1. Overview of Underground Metal Mines

This study uses underground metal mining as a case study to demonstrate the practical application of FLEM. Underground metal mining entails the extraction of economically valuable minerals from deposits located beneath the Earth’s surface. The mining process comprises a series of activities, including geological exploration, mining design, tunneling, and the establishment of ventilation and drainage systems. This sector is characterized by complex working environments and the high frequency of accidents. Figure 4 illustrates the structure of an underground metal mining production system.

3.2. Accident Evolution Process in Underground Metal Mines

Following Application II of FLEM, the main types of accidents and their associated hazards in underground metal mines were identified. This study focuses on the underground environment, excluding surface facilities like tailings storage, processing plants, and surface buildings. Additionally, only mines in the stable production phase were considered. The accident types were determined based on the Classification Standard for Work-Related Injuries in Enterprises (GB/T 6441-1986) [47], which lists 20 types of production accidents. After reviewing accident cases and the relevant literature, 11 common accident types for underground metal mines were identified: roof and rib falls, inundations, object strike, transportation, mechanical injury, electrocution, fire, falls from heights, explosions, explosive handling, and gas poisoning. For each accident type, the corresponding hazards were analyzed and listed in

Table 2. Hazard control measures code.

Accident	Hazard Source	Code	Measure Type	Behavior Requirement	Basis
Roof and rib falls (T01)	Rock with high ground stress (S01)	M03-R01	Control measure	The roadway support and stope roof management can ensure the safety of the workplace.	RIMSL-2.10.4
		M03-R02	Control measure	The mine roadway layout leaves enough ore pillars or rock pillars to ensure production safety.	RIMSL-2.10.5
		M03-R03	Control measure	It is prohibited to mine across layers and boundaries.	RIMSL-3.13
		M03-R04	Control measure	The nature and scale of larger faults and fracture zones should be made clear.	RIMSL-2.5.1
Object strike (T02)	Floating stone (S01)	M03-R01	Control measure	Underground mining operations must manage the roof and ribs according to the regulations of the operation procedures.	RIMSL-3.17
Inundation (T03)	Groundwater/Empty-area water (S01)	M03-R01	Control measure	The nature of aquifers and aquitards, the hydraulic connections between aquifers, and between surface water and groundwater should be made clear.	RIMSL-2.5.2
		M03-R02	Control measure	The distribution range, mining depth, and water accumulation situation of the original small kilns and old kilns within the mine design range should be made clear.	RIMSL-2.5.3
		M03-R03	Control measure	The drainage equipment should meet the national safety standard or industry safety standard.	RIMSL-3.14.1
		M03-R04	Control measure	The mine enterprise should mine within the scope approved by the mining license and is prohibited to mine across layers and boundaries.	RIMSL-3.13
	Rainwater (S02)	M03-R01	Control measure	There should be ground and underground waterproof and drainage systems and measures to prevent surface water from flowing into the stope.	RIMSL-2.10.7
Gas poisoning (T04)	Toxic and harmful gases (S01)	M03-R01	Control measure	Each mine has an independent ventilation system using mechanical ventilation to ensure sufficient air volume in the underground workplace.	RIMSL-2.10.2
		M03-R02	Control measure	The concentration of toxic and harmful substances in the air of the mine workplace should not exceed the national standard or industry standard.	RIL3SM-3.16
		M04-R01	Monitoring measure	Each mine is equipped with a sufficient number of ventilation detection instruments and toxic and harmful gas and underground environment detection instruments.	RIMSL-2.10.19
Fire (T05)	Combustibles (S02)	M02-R01	Substitution	The fireproof material should meet the national safety standard or industry safety standard.	RIMSL-3.14.3
	Ignition source (S02)	M01-R01	Elimination	No one is allowed to carry tobacco and ignition tools down the mine.	RIMSL-3.18
		M03-R01	Control measure	There is a possibility of spontaneous combustion of minerals and dust explosion, and effective preventive measures for spontaneous combustion should be taken.	RIMSL-2.5.4
		M03-R03	Control measure	There are fire prevention and extinguishing facilities and equipment.	RIMSL-2.10.10
		M03-R02	Control measure	When conducting dangerous operations such as blasting, hoisting, hot work, and temporary electricity, specialized personnel should be arranged for on-site safety management.	WSL-2.43
Explosive handling (T06)	Gunpowder, detonators in non-blasting operation sites (S01)	M03-R01	Control measure	The place where the mine stores blasting materials meets the relevant national regulations.	RIMSL-2.10.16
Explosions (T07)	Materials or equipment related to blasting (S01)	M03-R01	Control measure	The blasting materials should meet the national safety standard or industry safety standard.	RIMSL-3.14.3
		M03-R02	Control measure	The blasting operations and the manufacturing, storage, transport, test, and destruction of blasting materials in the mine must strictly implement the relevant national regulations.	RIMSL-3.24
Electrocution (T08)	Electrical equipment with electricity (S01)	M03-R01	Control measure	The ground and underground power distribution systems meet the relevant national regulations.	RIMSL-2.10.11
		M03-R02	Control measure	Motors, transformers, distribution cabinets, electrical switches, and electrical control devices should meet the national safety standard or industry safety standard.	RIMSL-3.14.2
		M03-R03	Control measure	Non-duty electrical personnel are not allowed to conduct electrical operations.	RIMSL-3.15
Mechanical injury (T09)	Mechanical equipment other than vehicles (S01)	M01-R01	Elimination	The state has a system of phasing out technologies and equipment that seriously endanger production safety.	WSL-2.38
		M03-R01	Control measure	The mine enterprise should regularly inspect and repair its mechanical and electrical equipment and its protective devices and safety detection instruments and establish a technical archive to ensure safe use.	RIMSL-3.15
		M03-R02	Control measure	Non-responsible equipment operation personnel are not allowed to operate the equipment.	RIMSL-3.15
Transportation (T10)	Vehicle equipment (S01)	M01-R01	Elimination	The state has a system of phasing out technologies and equipment that seriously endanger production safety.	WSL-2.38
		M03-R01	Control measure	The selection of motor cars, overhead lines, and tracks can meet the safety requirements.	RIMSL-2.10.12.4
		M03-R02	Control measure	The mechanical equipment for transporting people has reliable safety protection devices.	RIMSL-2.10.12.5
Falls from heights (T11)	Lifting equipment (S01)	M03-R01	Control measure	The mine lifting and transportation equipment, devices, and facilities meet the following requirements: The wire rope, connection device, lifting container, and safety chain have sufficient safety coefficients.	RIMSL-2.10.12.1
		M04-R01	Monitoring measure	The lifting and transportation equipment has a sensitive and reliable signal device.	RIMSL-2.10.12.6
	Other places that make people far from the ground (S02)	M03-R01	Control measure	The chute mine has safety measures to prevent and handle blockages.	RIMSL-2.10.8
		M05-R01	Warning measure	According to the requirements of different working places, set up mine safety signs.	RIMSL-4.40

.

Following the steps in Application Method II, different levels of control measures for various hazard sources were determined, primarily based on the Work Safety Law of the People’s Republic of China (WSL), the Mine Safety Law (MSL), and the Regulations for the Implementation of the Mine Safety Law (RIMSL). This resulted in the development of accident risk control processes, as shown in Table 2,

Table 3. Accident evolution process codes.

Code	Evolution Stage	Behavior Requirement	Basis
E01-R01	Hidden Hazard Management	Take timely measures to deal with the accident hidden dangers in the mine.	RIMSL-4.29.7
E02-R01	Personnel Coupling	Production and operation units should set up obvious safety warning signs on production and operation places and relevant facilities and equipment with large risk factors.	WSL-2.21.35
E02-R02		When production and operation units conduct dangerous operations, such as blasting, hoisting, hot work, and temporary electricity, specialized personnel should be arranged for on-site safety management to ensure the compliance of operation procedures and the implementation of safety measures.	WSL-2.21.43
E03-R01	Personnel Protection	Production and operation units must provide labor protection articles that meet the national standard or industry standard for employees and supervise and educate employees to wear and use them according to the use rules.	WSL-2.21.45
E04-R01	On-site Disposal	The mine enterprise should make each employee familiar with the mine disaster prevention and emergency plan and organize at least one mine disaster rescue drill every year.	RIMSL-4.40
E05-R01	Emergency Mechanism	The mine enterprise should compile a mine disaster prevention and emergency plan every year; at the end of each quarter, they should modify the plan according to the actual situation and formulate corresponding measures.	RIMSL-4.40
E05-R02		Production and operation units should formulate their own production safety accident emergency rescue plans.	WSL-5.81
E05-R03		Establish an emergency rescue organization.	WSL-5.82
E05-R04		Prepare necessary emergency rescue equipment, materials, and supplies and conduct regular maintenance and repair to ensure normal operation.	WSL-5.82

and

Table 4. Prerequisite conditions and regulatory requirements codes.

Code	Category	Requirement	Basis
H01-S01-R01	Fluke	Production and operation units should pay attention to the physical status, psychological status, and behavior habits of employees. They should strengthen the psychological counseling and spiritual comfort of employees and strictly implement the post-safety production responsibility to prevent accidents caused by abnormal behavior of employees.	WSL-2.21.44
H01-S02-R01	Complacency
H01-S03-R01	Herd
H01-S04-R01	Rebellious
H01-S05-R01	Adventurous
H02-S01-R01	Fatigue	The mine enterprise union lawfully safeguards the legitimate rights and interests of employees in production safety.	RIMSL-4.23
H02-S02-R01	Physical discomfort	When there is a situation that endangers the safety and health of employees, the mine enterprise should immediately correct or limit the time to solve it.	RIMSL-5.44
H02-S03-R01	Environmental impacts	In the intake air flow of the mining face, according to volume calculation, the oxygen content should not be less than 20%, the carbon dioxide content should not exceed 5%, and the air temperature at the underground working place should not exceed 28 °C.	RIMSL-3.22
H02-S04-R01	Insufficient skills	According to the regulations of this regulation, conduct safety education and training for employees.	RIMSL-4.29.5
O-R01	Manager	Seriously implement the provisions of the “Mine Safety Law” and this regulation as well as other laws and regulations related to mine production safety.	RIMSL-4.29.1
O-R01		Seriously implement the provisions of the “Mine Safety Law” and this regulation as well as other laws and regulations related to mine production safety.	RIMSL-4.29.1
O-R02		Formulate the enterprise’s production safety management system.	RIMSL-4.28, RIMSL-4.29.2
O-R03		Implement a double prevention mechanism of safety risk grading control and hidden danger investigation and treatment.	WSL-1.4
O-R04		Ensure the effective implementation of the enterprise’s safety production investments.	WSL-2.21.4

.

3.3. Accident Data Statistics and Quantification

Due to the sensitivity of mining accident data, this study uses publicly published accident analysis reports as the data source. The accident case sources include accident investigation reports publicly available on websites of the Ministry of Emergency Management and the Mining Safety Supervision Bureau. These accident reports typically contain information such as the basic details of the accident-prone enterprises, casualties and economic losses resulting from the accidents, the sequence of events leading to the accidents, the direct and indirect causes, and the outcomes of the incident response. In this study, a total of 61 accident cases were collected, with the distribution of various types of accidents illustrated in Figure 5.

Table 5. Accident information statistics (N/P represents quantity/proportion).

Accident Type	Hazard Source Code (N/P)	Violated Measure (N/P)	Personal Protection (N/P)	On-Site Disposal (N/P)	Emergency Situation (N/P)	Prerequisite Conditions Code (N/P)
T01	S01 (27/100%)	M03-R01 (26/96.3%)	Not Involved (27/100%)	Not Involved (27/100%)	Not Involved (23/85.2%) Defective Emergency Process (4/14.8%)	H01 (20/74.1%) H02-S04 (7/25.9%)
		M03-R02 (1/3.7%)
		M03-R04 (3/11.1%)
T02	S01 (3/100%)	M03-R01 (3/100%)	No Protection (3/100%)	Not Involved (3/100%)	Not Involved (3/100%)	H01 (3/100%)
T03	S01 (14/87.5%)	M03-R01 (9/56.2%)	Not Involved (16/100%)	Not Involved (16/100%)	Not Involved (16/100%)	H01 (12/75.0%)
	S02 (2/12.5%)	M03-R02 (7/43.8)				H02-S04 (4/25%)
T04	S01 (4/100%)	M03-R02 (2/50.0%)	Not Involved (3/75.0%)	Not Involved (3/75.0%)	Not Involved (2/50.0%)	H01 (4/100%)
		M03-R03 (2/50.0%)	No Protection (1/25.0%)	On-site Measures Ineffective (1/25.0%)	Emergency Process Defective (2/50.0%)
T05	S02 (3/100%)	M03-R02 (3/100%)	Not Involved (1/33.3%)	On-site Measures Ineffective (3/100%)	Not Involved (2/66.7)	H01 (3/100%)
			No Protection (2/66.7)		Emergency Process Defective (1/33.3%)
T06	S01 (1/100%)	M03-R01 (1/100%)	No Protection (1/100%)	Not Involved (1/100%)	Not Involved (1/100%)	H01 (1/100%)
T07	S01 (1/100%)	M03-R02 (1/100%)	Not Involved (1/100%)	Not Involved (1/100%)	Not Involved (1/100%)	H02-S04 (1/100%)
T10	S01 (1/100%)	M01-R01 (1/100%)	Not Involved (1/100%)	Not Involved (1/100%)	Not Involved (1/100%)	H01 (1/100%)
T11	S01 (3/60.0%) S02 (2/40.0%)	M03-R01 (3/60.0%) Not Applicable (2/40.0%)	Not Involved (2/40.0%) No Protection (3/60.0%)	Not Involved (5/100%)	Not Involved (4/80.0%) Emergency Process Defective (1/20.0%)	H01 (4/80.0%)
						H02-S01 (1/20.0%)
						H02-S04 (1/20.0%)

presents the statistical information on accident cases categorized by the accident causation codes we have outlined. Each column in the table corresponds to a specific stage in the evolution of the accident, providing a detailed breakdown of the occurrences at each phase.

4. Discussion

4.1. Analysis of Results

The analysis of accident cases revealed several key control points for different types of accidents:

• Roof and rib falls: The majority (96.3%) of these accidents resulted from inadequate support and roof management. Furthermore, 14.8% of accidents were exacerbated by deficiencies in emergency response mechanisms. Unsafe behaviors, particularly insufficient safety awareness among workers, were identified as significant factors. These accidents often involved instantaneous energy release, making on-site response and protection measures less effective. The focus should be on improving roof support management and emergency preparedness, while enhancing workers’ safety awareness.
• Inundations: most accidents were due to inadequate hydrogeological investigations.
• Falls from heights: Some fall accidents were not linked to other hazards and could only be mitigated through personal protective measures. Increasing workers’ awareness of fall protection is crucial.
• Fires: These accidents were primarily caused by the failure of on-site response measures, highlighting the importance of emergency response training to mitigate the consequences of fire accidents.
• Gas poisoning: the serious consequences of these accidents were mostly due to improper rescue measures, leading to further escalation of the incident.
• Object strikes: Injuries from falling rocks were primarily due to the failure to wear protective equipment. Therefore, stricter enforcement of penalties for not wearing protective equipment during operations can significantly reduce the severity of such accidents.

The accident data used in the case analysis are sourced from publicly available accident investigation reports, which originate from mining enterprises of varying types, scales, extraction methods, and management levels. This leads to certain limitations in the application of the case study:

• Most accident reports analyze the causes of accidents primarily based on direct and indirect factors, often lacking corresponding information on safety measures, on-site responses, emergency situations, and the preconditions for unsafe behaviors.
• Although all accident cases can be linked to corresponding causative nodes within the accident evolution pathways, the organization of these pathways primarily references relevant laws without constructing a more detailed accident evolution network based on safety regulations and operational standards, resulting in an overly broad analysis.
• The analysis of the preconditions for unsafe behaviors lacks a detailed classification of psychological factors, mainly because the publicly available accident cases are all related to fatalities, which prevents a thorough investigation of psychological aspects.
• The application of different accident data from various mines does not include a breakdown of responsible personnel.

These issues can be addressed through comprehensive analyses based on specific production enterprises’ hazard assessments, near-miss incidents, and detailed investigations of accidents.

4.2. Comparison with Existing Accident Causation Models

Accident causation tracing and the formulation of control measures are typical application scenarios for accident causation models. In this study, we comprehensively review the relationships among the factors that may be involved in the evolution of production accidents, clarifying the core role of personnel as the only factor with subjective initiative in the accident evolution process. We use the failure modes of personnel in controlling accident hazards and subsequent response behaviors as the basis for tracing the genesis and development of accidents. This approach effectively distinguishes between the unsafe behaviors of individuals and unsafe conditions in accident causation analysis. Prior to the initiation of an accident, the emphasis is primarily on the unsafe conditions associated with various factors, whereas after the initiation, attention shifts to the unsafe behaviors of individuals [28]. This model effectively resolves issues related to incomplete element analysis and the unclear distinctions between unsafe conditions and unsafe behaviors that may arise in chain-type accident causation models, resulting in a coherent and clear analytical framework. Constructing this framework facilitates more accurate identification of potential accident risks and provides a solid foundation for formulating effective control measures.

This model is also systematic. The systematization of the accident causation model is reflected in the abstract description of system elements and the organization of their hierarchical structure and complex feedback relationships [10]. For example, the FRAM abstracts each element involved in the analysis into functional modules characterized by six attributes: inputs, preconditions, resources, time, constraints, and outputs [6]. The elements in the model discussed in this study are derived from a synthesis of existing modular accident causation models. During the analysis phase, we conduct an in-depth examination of the possible interactions among these elements, ensuring the comprehensiveness and systematic nature of both the elements and their interactions. Furthermore, this model offers clear analytical logic and a specific framework, with risk information primarily sourced from laws, regulations, and standards. Consequently, researchers only need a certain level of industry experience and the ability to analyze data and extract information to conduct accident analysis or even trace the development paths of accidents within an organization [5]. This enhances the model’s operability, reducing both the time and complexity involved in the analysis process.

Visualization is a crucial method for clarifying accident development paths and enhancing the efficiency of analyzing accident evolution chains. This study has systematically organized and graphically depicted the potential evolutionary logic of accidents [42]. This graphical approach lays a strong foundation for utilizing information technologies to trace the evolution of accidents. By representing the accident development process visually, stakeholders can more intuitively grasp the key nodes and responsible parties involved in an incident. This clarity facilitates data mining to explore the causes of accidents at various levels of granularity. In accident investigations, the constructed accident evolution path model enables investigators to swiftly identify critical links and accountable individuals, thereby improving the overall efficiency of the investigation process.

4.3. Limitations and Future Research Directions

The application scope of this model has been intentionally limited to the internal environment of production enterprises, without adequately considering the impact of external factors on organizational safety. Specifically, the roles of external influences, such as government regulation, societal oversight, and external safety resources, in enterprise safety management have not been explored in depth. Additionally, the societal losses resulting from production accidents—such as impacts on wealth, the environment, and intangible assets—are not reflected in this model. Government regulation plays a crucial role in enterprise safety management; strict regulatory policies and effective enforcement can motivate companies to enhance their safety practices. Societal oversight can also exert pressure on enterprises, prompting them to prioritize safety concerns. Furthermore, external safety resources, such as professional safety consulting agencies and technical service providers, can offer valuable expertise and technical support for safety management. However, the exclusion of these external factors from analysis limits the comprehensiveness of this model [15].

Due to data constraints, this study’s analysis of key safety management controls focuses solely on accident case data, utilizing manual analysis and identification methods for statistical purposes. This approach is restricted by the volume of data and lacks the timeliness required for effective safety management. In practice, organizations need to be promptly informed about safety conditions to implement effective preventive and control measures. Manual analysis and identification of accident case data is not only time-consuming but also challenging when addressing large datasets. Moreover, relying solely on accident case data may overlook some potential safety hazards and risks. According to Heinrich’s accident theory, there exists a proportional relationship between safety hazards, near-miss incidents, and production accidents. Therefore, including all relevant data within the enterprise in the analysis is crucial for enhancing the precision of controls and the effectiveness of measures. However, this creates a conflict between the vast amount of data and the efficiency of analysis.

It is important to note that the collection of accident case data may be subject to certain limitations. Some accidents may be underreported or not reported at all, especially those with minor consequences or those that are considered normal operational variations. This could lead to a biased representation of the actual accident distribution and risk profile. Additionally, the accident analysis methods used in the collected accident reports may influence the data’s reliability and scope. If the methodology focuses predominantly on specific cause–effect chains, other contributing factors might remain unexplored. Conversely, there might be instances where numerous accidents are attributed to human error, possibly due to the ease of identification or the prevalence of such factors in certain contexts.

With the rapid advancement of big data analytics and large language model technologies, real-time monitoring and dynamic assessment of safety conditions within enterprises based on real-time hazard data have become feasible [48]. Big data analytics can process large volumes of data, extracting valuable insights to support decision-making in safety management. Meanwhile, large language model technology can quickly analyze and comprehend textual data, aiding organizations in better identifying safety hazards and risks. The accident analysis model proposed in this study provides a robust framework and foundation for risk assessment based on hazard sources. Future research will explore the applicability of this model in the context of intelligent accident prevention and dynamic real-time analysis of control measures.

5. Conclusions

As industrial production becomes increasingly complex, accident risks in modern industries continue to grow. Therefore, in-depth research on accident causation theory and the construction of comprehensive accident analysis models is of significant practical importance for ensuring safe production and the protection of employees’ lives. This study aims to address the limitations of existing accident causation theories in the practical application within complex systems, proposing an evolution model centered on human factors that covers the entire lifecycle of accidents. Through this research, we hope to further refine the analysis of accident causes and provide a theoretical basis for comprehensively preventing production accidents.

Through a comparative analysis of typical accident causation theories, we have clarified the essential role of personnel in safety management and redefined the relationships among various factors. Furthermore, we have introduced a systematic safety management approach grounded in this new model and conducted an empirical analysis via a case study of underground metal mining. This allowed us to identify key pathways and core risk factors that contribute to accidents, providing a foundation for enterprises to develop more targeted accident prevention measures.

However, there are still certain limitations in this research, such as the insufficient consideration of external factors (such as government regulation and societal oversight) and the need to improve the breadth and accuracy of accident data analysis. Notably, the proposed FLEM demonstrates promising potential for application in near-miss incident analyses, which could further enhance proactive risk identification capabilities. Future research will integrate technologies like big data analytics and large language models to further enhance real-time monitoring and dynamic assessment mechanisms for accident risks, promoting the realization of intelligent accident prevention. Additionally, the model will be validated across more industries to expand its applicability and support broader enterprise safety management efforts.