Risk Analysis of Occupational Hazards in Confined Spaces in the Petrochemical Industry Using the HORshe Method

2.1 Occupational Health and Safety risk management in the petrochemical industry

OHS risk management in the petrochemical industry is a systematic process that involves hazard identification, risk assessment, and risk control to establish a safe working environment, particularly in industries with high-risk characteristics owing to the use of hazardous chemicals, extreme pressure and temperature conditions, and complex operational processes [8, 14, 15]. One of the most critical areas is confined space work, which is characterized by limited ventilation, the potential for oxygen deficiency, and the accumulation of toxic gases that may result in fatal accidents [2, 8]. Conceptually, a hazard is defined as a source with the potential to cause injury, whereas risk is the combination of the likelihood of an event occurring and the severity of its consequences [14, 15]. The risks associated with confined spaces include physical, chemical, biological, ergonomic, and psychosocial factors, which often occur simultaneously [8, 11, 13, 21]. Therefore, effective risk management requires a comprehensive approach supported by the implementation of Process Safety Management (PSM), the adoption of modern technologies such as sensors and inspection robots, and the strengthening of safety culture through communication and worker participation [2, 22, 23].

2.2 HORshe

The HORshe method is an extension of the HOR model,  specifically developed to provide a structured framework for OHS risk analysis [14, 15]. HORshe integrates the SHELL model (Software, Hardware, Environment, and Liveware), the principles of the Hierarchy of Control, and a weighting system based on the Analytical Hierarchy Process (AHP) into three main phases: risk identification and assessment, prioritization of PAs, and evaluation of residual risk using the NARP [14, 15]. The integration of the SHELL model enables a systemic analysis of interactions among human factors, equipment, procedures, and the working environment, whereas the application of the Hierarchy of Control facilitates the prioritization of proposed PAs according to their relative effectiveness. The residual risk evaluation phase represents one of the principal advantages of HORshe, as it enables continuous measurement of risk reduction until risks are reduced to an acceptable level in accordance with the ALARP principle [14, 15].

2.3 Previous studies

Numerous previous studies have demonstrated that methods such as HIRA, JSA, FMEA, FTA, and resilience engineering-based approaches have been widely applied in confined space risk analysis [11, 19, 24-28]. However, most of these methods remain partial in nature and are not capable of comprehensively integrating the complex relationships among multiple risk factors. In contrast, the HOR method and its subsequent developments have shown considerable advantages in systematically identifying risk agents and measurably determining mitigation priorities [29, 30]. The latest development, HORshe, offers a more holistic approach by incorporating systemic interactions and residual risk evaluation [14, 15]. Nevertheless, the application of HORshe in the context of confined space work in the petrochemical industry remains limited, particularly in integrating the SHELL model for systematic risk event classification, incorporating the Hierarchy of Control into PA prioritization, and evaluating projected residual risk through HORshe Phase 3. Therefore, the novelty of this study lies in integrating the SHELL model with the HORshe framework to systematically identify risk events, applying the Hierarchy of Control through the Effectiveness-to-Difficulty Ratio (ETD_k) for prioritizing PAs, and evaluating projected residual risk through post-mitigation expert reassessment within the context of confined space work in ammonia plants. This integrated approach provides a structured framework for identifying, prioritizing, and evaluating occupational risks in high-risk confined space operations.

4.1 General overview of the research site

Plant A is located in East Kalimantan, Indonesia, and is one of the largest producers of ammonia and urea fertilizers in Indonesia. The company’s production facilities include five ammonia, five urea, and several NPK plants.

4.2 Data collection

Data were collected through direct field observations conducted during the plant turnaround period, during which all reactors, vessels, and tanks were opened for inspection and maintenance activities. In addition, a Focus Group Discussion (FGD) was conducted involving nine certified General OHS practitioners from four ammonia companies in Indonesia.

In this study, the identification of risk events and risk agents constituted a crucial initial step prior to conducting further risk assessment. Risk events were identified by integrating data obtained from all the data collection methods described previously. This process was carried out at Plant A, specifically at the ammonia plant operating under Haldor Topsoe technology licensing. The primary confined space activities observed in this study are presented in the table of main confined space observation objects (Table 6). In addition to these major units, the researcher observed numerous smaller confined spaces, such as separators and heat exchangers, between the main vessels listed in Table 6.

Table 6. Main confined space observation objects

4.3 HORshe Phase 1—Risk identification and assessment

4.3.1 Mapping of risk events based on the SHELL model

Based on the collected data, risk events were mapped using the SHELL model (Table 7), which consists of software (S), hardware (H), environment (E), and liveware (L) components [10]. The results identified 14 risk events, of which seven risk events (50.0%) were associated with Liveware-Environment interactions, six risk events (42.8%) were associated with Liveware-Hardware interactions, and one risk event (7.1%) was associated with Liveware-Liveware interactions.

Based on Table 7, the risks associated with confined space work are diverse and arise from interactions between the Liveware component and the Environment, Hardware, and other Liveware components of the SHELL model. The majority of the identified risk events belonged to the Liveware-Environment category, accounting for 7 of the 14 risk events (50.0%), including E₁ (workers exposed to toxic gases), E₂ (workers experiencing oxygen deficiency), E₃ (workers exposed to explosions or fires caused by flammable gases), E₅ (workers falling from a height), E₆ (workers slipping or tripping due to slippery conditions and confined space structures), E₁₀ (overheating/heat stress in confined spaces), and E₁₂ (entrapment caused by narrow confined space access structures). These findings indicate that hazardous atmospheric conditions, restricted access, and adverse environmental conditions are the dominant sources of occupational risks.

Table 7. Risk event mapping

Liveware-Hardware interactions account for 6 risk events (42.8%), comprising E₄ (workers suffering electric shock), E₇ (delayed evacuation), E₉ (loss of communication during emergencies), E₁₁ (entrapment in substances such as pall rings or catalysts), E₁₃ (workers being pinched, punctured, struck, or cut by work equipment), and E₁₄ (noise exposure).

These findings highlight the importance of equipment integrity, electrical safety, communication systems, emergency response facilities, and safe work equipment in preventing occupational accidents in the petrochemical industry. Finally, only one risk event (7.1%), E₈ (secondary victim incidents resulting from attempts to rescue the first victim), was classified as a Liveware-Liveware interaction, emphasizing the importance of effective communication, coordination, and strict adherence to rescue procedures to prevent additional casualties during confined space emergencies. Collectively, these findings demonstrate that environmental hazards represent the dominant source of occupational risk in confined space work, followed by equipment-related hazards, while human interaction contributes a smaller but equally critical risk, requiring effective communication and coordinated emergency response (Figure 2).

image009.png

Figure 2. Distribution of risk events based on the SHELL model

4.3.2 Determination of risk event severity

Severity assessment was conducted based on the greatest impact on human aspects.

As shown in Table 8, the identified risk events exhibit varying levels of severity according to their potential consequences. Six risk events (42.9%) were assigned a severity score of 5, indicating catastrophic consequences that may result in multiple fatalities. These events include exposure to toxic gases, oxygen deficiency, explosions or fires caused by flammable gases, electric shock, delayed evacuation during emergencies, and secondary victim incidents resulting from attempts to rescue the first victim. Furthermore, four risk events (28.6%) were assigned a severity score of 4, indicating hazards that may result in a single fatality or major injury, including falls from height, loss of communication during emergencies, entrapment in catalysts or packing materials, and entrapment caused by restricted confined space access. The remaining four risk events (28.6%) were assigned a severity score of 2, representing hazards that may result in injuries requiring medical treatment, including slips or trips, heat stress, injuries caused by work equipment, and noise exposure. Overall, the Severity assessment indicates that confined space work is characterized by a substantial proportion of high-consequence hazards, with nearly half of the identified risk events classified at the catastrophic severity level. These findings highlight the importance of implementing effective preventive and protective measures to reduce occupational risks and improve worker safety (Figure 3).

Table 8. Severity of risk events for HORshe Phase 1

image010.png

Figure 3. Distribution of severity scores for risk events (HORshe Phase 1)

4.3.3 Determination of risk agent occurrence

Occurrence assessment was conducted for each risk agent using a scale of 1–5. A total of 53.85% of the risk agents (A₃, A₄, A₅, A₇, A₈, A₁₀, and A₁₁) were assigned an occurrence score of 5, indicating a very high likelihood of occurrence, while 23.08% were assigned an occurrence score of 4, and the remaining 23.08% were assigned an occurrence score of 3.

As shown in Table 9, the likelihood of risk occurrence in confined space work varies across five rating levels, with the majority, representing 53.85% (A₃, A₄, A₅, A₇, A₈, A₁₀, and A₁₁), classified at an occurrence scale 5 (very high) owing to factors such as exposure to toxic gases, flammable gases, high-voltage electricity, and differences in elevation. Furthermore, 23.08% (A₁, A₂, and A₁₂) were classified as occurrence scale 4 (high), influenced by inadequate worker knowledge, the absence of periodic atmospheric testing, and limited access. Another 23.08% (A₆, A₉, and A₁₃) were classified at occurrence scale 3 (moderate, approximately 40–60%), resulting from heat stress, improper use of personal protective equipment (PPE), and the use of mechanical equipment. Overall, more than half of the identified risks exhibited a very high likelihood of occurrence in the absence of PAs, with a distribution of seven risk agents (53.85%) at scale 5, three risk agents (23.08%) at scale 4, and three risk agents (23.08%) at scale 3, underscoring the importance of implementing effective controls to reduce the probability of occupational accidents (Figure 4).

Table 9. Occurrence of risk agents for HORshe Phase 1

image011.png

Figure 4. Distribution of occurrence scores for risk agents (HORshe Phase 1)

4.3.4 Determination of risk agent detection

Detection assessment was conducted through discussions with experts using a scale of 1-5 [14, 15]. Risk agent A₅ (inert gas) received the highest detection score (5) because gases such as nitrogen are odorless and colorless, making them extremely difficult to detect without specialized instruments. In contrast, risk agents A₁₁ and A₁₂ were assigned a detection score of 1 because these risks could be identified visually.

The detection assessment was conducted using a scale of 1-5, ranging from risks that are very easy to detect to those that cannot be detected at all. In general, the results presented in Table 10 indicate that most risks associated with confined space work were assigned detection scores of 2 and 3, meaning that they can be identified through observation or condition inspections. This includes A₁, A₆, and A₉ (score of 2) and A₂, A₃, A₄, and A₇ (score of 3), which require detection instruments. In contrast, A₅ was assigned a detection score of 5 because inert gases, such as nitrogen, are colorless and odorless, making them extremely difficult to detect. Meanwhile, A₁₁ and A₁₂ were assigned detection scores of 1 because these risks could be predicted visually.

Overall, the distribution consisted of one risk agent (7.7%) at scale 5, four risk agents (30.8%) at scale 3, six risk agents (46.2%) at scale 2, and two risk agents (15.4%) at scale 1. These findings indicate that although most risks can still be detected, several critical hazards require more advanced detection and mitigation systems to ensure worker safety (Figure 5).

Table 10. Detection of risk agents for HORshe Phase 1

image012.png

Figure 5. Distribution of detection scores for risk agents (HORshe Phase 1)

4.3.5 Calculation of Aggregate Risk Potential

The ARP value was calculated using the following formula:

$A R P_j=O_j \times D_j \times \sum\left(S_i \times R_{i j}\right)$   (4)

An example of this calculation is presented below.

$\begin{aligned} & A R P_1=4 \times 2 \times[ (5 \times 9)+(5 \times 9)+(5 \times 9)+(5 \times 9) \\ &+(5 \times 9)+(4 \times 9)+(2 \times 9)+(5 \times 9) \\ &+(5 \times 9)+(4 \times 9)+(2 \times 9)+(4 \times 9) \\ &+(4 \times 9)+(2 \times 9)] \\ & A R P_1=4,104\end{aligned}$

Based on the ARP calculation (Table 11), the highest-priority risk agent was A₅ (inert entry) because of its severe consequences, association with multiple risk events, and very low detectability owing to the characteristics of inert gases, such as nitrogen, which are colorless and odorless. The second priority was A₁ (workers’ lack of knowledge regarding confined space hazards), which exhibited a high ARP value because it was related to nearly all identified risk events and had a high occurrence score, despite being relatively easier to detect. The third priority was A₄ (work environments containing flammable gases such as methane and hydrogen), which is highly relevant to ammonia plant processes, is associated with several risk events, and requires gas detectors to ensure that gas concentrations remain below the established explosive limits.

Table 11. Aggregate Risk Potential (ARP) values for HORshe Phase 1

image016.png

Figure 6. Pareto analysis of Aggregate Risk Potential (ARP)

4.3.6 Pareto analysis and selection of priority risk agents

The Pareto diagram was used to assist the researchers in prioritizing interventions for the most dominant contributing factors, although the 80/20 principle is not rigid and may be adjusted according to the context of the data [32, 33]. The Pareto analysis result indicated that 88.4% of the cumulative risk agent values contributed significantly to the occurrence of risk events (Figure 6); therefore, these risk agents were designated as priorities for mitigation. The determination of this threshold was further validated through expert discussions, particularly considering the extensive use of electrical equipment in confined space work, where the severity level may reach scale 5, indicating the potential for multiple fatalities (Table 12). Consequently, nine priority risk agents were selected for further analysis in HORshe Phase 2.

Table 12. Results of risk agent mapping based on Pareto analysis

4.4 HORshe Phase 2—Determination and prioritization of Identification of Preventive Actions

4.4.1 Identification of Preventive Actions

Based on the nine priority risk agents, 13 PAs (PA₁–PA₁₃) were identified and classified according to the Hierarchy of Controls [12, 13, 27]. These actions consisted of elimination (PA₁, PA₅, and PA₈), substitution (PA₉ and PA₁₁), Engineering Controls (PA₂, PA₆, PA₇, and PA₁₀), Administrative Controls (PA₃, PA₁₂, and PA₁₃), and Personal Protective Equipment (PPE) (PA₄) (Table 13).

Table 13. Effectiveness to Difficulty Ratio (ETD_k) Values and Priority of Preventive Actions (PAs)

4.4.2 Calculation of Total Effectiveness (TE_k)

The Total Effectiveness (TE_k) value was calculated using the following formula:

$T E_k=S C_k \times \sum\left(R_{k j} \times A R P_j\right)$   (5)

An example of this calculation is presented below.

$\begin{aligned} T E_1=9 \times((9 \times & 4,250)+(1 \times 4,104)+(9 \times 3,600) \\ & +(9 \times 3,030)+(3 \times 2,775) \\ & +(3 \times 2,436)+(0 \times 2,322) \\ & +(0 \times 2,040)+(0 \times 1,995) \\ T E_1=1,058,913 & \end{aligned}$

PA₅ achieved the highest Total Effectiveness (TE_k) value of 2,150,712, indicating its strong effectiveness in eliminating the risk associated with worker incompetence.

4.4.3 Calculation of Effectiveness-to-Difficulty Ratio and prioritization of Preventive Actions

The ETD_k was calculated using the following formula:

$E T D_k=\frac{T E_k}{D_k}$   (6)

An example of the calculation for PA₁ is presented as follows:

$\begin{aligned} & E T D_1=\frac{1,058,913}{4} \\ & E T D_1=264,728\end{aligned}$

Figure 7 shows that PA₅ (mandatory certification for confined space OHS technicians) obtained the highest ETD_k value (537,678), followed by PA₁ (blind isolation and lockout/tagout [LOTO], 264,728), and PA₁₃ (awareness training, 238,968).

image021.png

Figure 7. Effectiveness to difficulty ratio (ETD_k) values for Preventive Actions (PAs)

4.5 HORshe Phase 3—Residual risk evaluation

The implementation of the proposed PAs was carried out during the two-week preparation period before the turnaround and continued throughout the first two days of turnaround execution. Initially, awareness training on safe confined space work practices was provided to workers who had previously obtained certification as Confined Space OHS technicians. This was followed by coordination meetings involving the OHS, Project Planning, Process Management, Operations, Electrical and Mechanical Maintenance Departments to ensure that all PAs were integrated into the turnaround schedule, including their implementation sequence and assigned responsibilities.

Before confined space entry was authorized, safety inspectors from the OHS Department conducted field verification to confirm that all PAs had been fully implemented in accordance with the predetermined control plan. In addition, all workers were granted Stop Work Authority, enabling them to suspend or postpone work whenever any required control measure was satisfactorily implemented.

Following field verification, the Severity, Occurrence, and Detection parameters were reassessed through structured Focus Group Discussions (FGDs) involving expert consensus in accordance with the HORshe Phase 3 methodology. This reassessment was undertaken to estimate the projected residual risk under the implemented control measures rather than to demonstrate confirmed long-term operational safety performance, which would require extended operational monitoring and safety performance data. Accordingly, the post-mitigation severity values represent expert-based estimates of the expected residual consequences after the verified control measures had been implemented, rather than changes in the intrinsic severity of the original hazards.

For example, blind isolation prevents the ingress of inert, toxic, and flammable gases from interconnected process equipment into the confined space, thereby substantially reducing the expected consequences of hazardous atmospheric exposure. Likewise, engineering controls, such as rope access systems, full-body harnesses, and lifelines, reduce the expected consequences of falls from elevated work locations or into catalyst beds. Furthermore, replacing electrically powered tools with pneumatic equipment eliminates potential ignition sources in flammable atmospheres while simultaneously removing workers' exposure to electrical hazards. Consequently, the expected consequences of hazardous gas exposure, fire, explosion, falls, and electrical shock were considered substantially lower after implementing the verified control measures.

However, the implementation of PAs does not eliminate occupational risk because residual risk is inherently dynamic and may change over time owing to variations in workplace conditions, equipment integrity, operational activities, environmental factors, and human performance. Therefore, the reassessed severity values should be interpreted as expert-based estimates of the projected residual consequences under the implemented control measures at the time of assessment, rather than as permanent reductions in risk severity. Continuous monitoring, periodic reassessment, and ongoing implementation of risk controls remain essential to ensure that residual risks are maintained at an acceptable level throughout the operational lifecycle.

Based on the post-mitigation reassessment, the occurrence values decreased for all risk agents, with five risk agents (55.6%) classified as Scale 1 and four risk agents (44.4%) as Scale 2. Likewise, all Detection values improved to Scale 1 (100%), indicating that the implemented control measures substantially enhanced hazard detectability. The reassessed severity values showed that nine risk events (64.3%) were classified as Scale 2 and five risk events (35.7%) as Scale 1, representing the projected residual consequences following the implementation of PAs.

The NARP was calculated using the same formula as follows:

$A R P_j=O_j \times D_j \sum\left(S_j \times R_{i j}\right)$   (7)

An example of this calculation is presented below:

$\begin{aligned} A R P_1=2 \times 1 \times[ & (2 \times 9)+(2 \times 9)+(2 \times 9)+(2 \times 3) \\ & +(2 \times 9)+(2 \times 9)+(2 \times 9)+(2 \times 9) \\ & +(1 \times 9)+(1 \times 3)+(1 \times 9)+(1 \times 9) \\ & +(1 \times 9)+(2 \times 9)]\end{aligned}$

$A R P_1=414$

This calculation was also performed for several priority risk agents, namely A₃, A₁₀, A₉, A₅, A₈, A₄, A₂, and A₇.

The results presented in Table 14 indicate that the risk agent related to workers' lack of knowledge regarding confined space hazards (i.e., inadequate competence) (A₁) became the highest-priority residual risk following the HORshe Phase 3 evaluation, whereas the highest-priority risk agent before implementation of the PAs was the work environment containing inert gases, such as nitrogen and carbon dioxide (A₅). Nevertheless, the ARP value for risk agent A₁ was estimated to decrease by 89.91%, whereas the ARP value for risk agent A₅ was estimated to decrease by 96.24%.

Furthermore, the largest estimated reduction in ARP was observed for the risk agent associated with the use of electrical systems exceeding 50 VAC and 120 VDC in confined spaces (A₇), with an estimated reduction of 98.80%. Overall, the HORshe Phase 3 evaluation estimated that the implemented PAs could reduce the ARP by an average of 94.75% compared with the initial risk condition. This estimated reduction was derived from the post-mitigation reassessment of the Severity, Occurrence, and Detection parameters through a structured expert consensus process conducted during the turnaround period. Therefore, the reported reduction should be interpreted as an expert-based estimate of the projected residual risk under the implemented PAs at the time of assessment, recognizing that occupational risks remain dynamic and require continuous monitoring and periodic reassessment.

Table 14. Aggregate Risk Potential (ARP) values in HORshe Phase 3—Comparison before and after mitigation

4.6 Advantages of the HORshe method in confined space risk analysis

The HORshe method offers significant advantages over the JSA and HIRA methods, currently used at the research site.

HORshe encompasses a more comprehensive set of parameters, including additional components such as Detection, the SHELL model, the degree of implementation difficulty, preventive action evaluation, and the analysis of relationships between risk agents and risk events (Table 15). The integration of the SHELL model enables a holistic examination of the interactions among workers, equipment, systems, and environmental conditions of the work environment. By incorporating implementation difficulty through the ETD_k, HORshe produces risk control recommendations that are more practical and can be prioritized rationally based on a balanced consideration of both effectiveness and ease of implementation.

Table 15. Comparison of HORshe, Job Safety Analysis (JSA), and Hazard Identification and Risk Assessment (HIRA) at the research site

4.7 Managerial implications

The results of the risk analysis using the HORshe method indicate that PAs should be established as mandatory prerequisites for all confined space work, particularly during plant turnaround activities. During this period, risk levels increase significantly because numerous process equipment units are opened for inspection, cleaning, and repair; multiple tasks are carried out simultaneously; large numbers of workers and contractors are involved; and all activities are performed under strict schedule constraints. Therefore, PAs must be fully integrated into turnaround planning, including the work breakdown structure, project schedules, and Stop Work Authority mechanisms. Work should not commence until all critical controls have been verified, including energy isolation, atmospheric testing, ventilation, gas detectors, personal protective equipment (PPE), standby personnel, rescue team readiness, and worker competency. The principle of Stop Work Authority serves as a fundamental safeguard to ensure that safety is maintained without compromising the project objectives.

From a managerial perspective, HORshe provides a quantitative framework for prioritizing risk-control measures. The ARP is used to identify critical risk agents, the ETD_k is used to select PAs that are both effective and feasible, and the NARP is used to evaluate residual risks following mitigation. Consequently, safety-related decisions are based not only on experience and professional judgment but also on measurable and systematic analyses. Effective implementation requires a clear allocation of responsibilities among project management, operations, maintenance, and OHS departments to ensure that turnaround activities are conducted safely, systematically, and in support of achieving zero incidents.