Literature Review on Industrial Safety, Occupational Health, and AI-Enabled Workplace Risk Management

Literature Review on Industrial Safety, Occupational Health, and AI-Enabled Workplace Risk Management

Sajithkumar S. K.

Lecturer in Automobile Engineering department Government Polytechnic College, Attingal Kerala, India

Abstract – Industrial safety is essential for preventing workplace accidents and protecting employee health. This review examines nine research papers on occupational hazards, workplace health, risk assessment, artificial intelligence, machine learning, and Safety 4.0 technologies. The studies highlight the importance of identifying workplace risks and using digital technologies for hazard monitoring, injury prediction, and safety management. However, challenges such as data quality, technology integration, and practical implementation remain. The review emphasizes the need for an integrated, technology- driven approach to improve workplace safety and promote proactive risk prevention.

Keywords – Industrial Safety, Occupational Health, Risk Assessment, Artificial Intelligence, Machine Learning, Safety 4.0, Workplace Safety, Injury Prevention.

1. INTRODUCTION

   Industrial safety is a multidisciplinary concern involving the prevention of workplace injuries, control of occupational hazards, protection of workers physical and psychological health, and continuity of industrial operations. The papers reviewed approach this issue from complementary perspectives: workplace hazard identification, quantitative risk assessment, cardiovascular workload, health-promotion programmes, machine-learning-based injury prediction, artificial intelligence (AI) for employee well-being, digital Safety 4.0 technologies, knowledge-graph construction, and regional industrial-safety assessment.

   Taken together, the studies indicate that effective industrial safety requires more than preventing immediate accidents. It also requires identifying underlying physical and psychosocial hazards, monitoring workers health, learning from incident records, and incorporating risk information into organizational and technical decisions [1][8]. However, the papers differ in scope and methodology: some provide systematic reviews, others propose computational methods, and others focus on risk theory or regional economic indicators. Their findings should therefore be integrated as complementary evidence rather than treated as directly equivalent measures of safety.

2. REVIEW OF THE LITERATURE

   1. Workplace hazards and occupational health

      Erazo-Chamorro et al. [1] examine workplace risk factors and the conditions required to create a healthy and safe work environment. Their analysis considers physical and psychological hazards alongside organizational safety climate and workplace design. Ergonomics receives particular emphasis, with the authors identifying inadequate ergonomic conditions as a critical factor in work-related illness prevention. This highlights the importance of designing workstations and tasks around workers capabilities rather than relying exclusively on protective equipment or post-incident corrective measures.

      The review by Dias et al. [2] extends occupational-health assessment to cardiovascular workload. It identifies work- environment factors, working hours, task demands, posture, repetitive movements, and individual worker characteristics as relevant factors in cardiovascular load. Relative heart rate is among the objective metrics commonly reported in the reviewed studies. The authors also identify inconsistent evidence regarding causal relationships between occupational exposures and cardiovascular disease, indicating a need for more consistent measurement and stronger research designs.

      These studies complement each other by showing that workplace risks can arise from both the immediate physical environment and the cumulative physiological demands of work. Ergonomic assessment and cardiovascular monitoring can therefore contribute to a broader occupational-health strategy. Nevertheless, the studies do not establish a single universal measure that captures all workplace hazards.

   2. Workplace health promotion and worker well-being

      Javanmardi et al. [3] systematically review workplace health- promotion programmes for industrial workers. Of 25,555 records initially identified, 39 studies were included. The review found generally positive but heterogeneous effects across interventions, while the overall evidence was limited by moderate study quality and a high risk of bias in many studies. Only one included intervention improved safety outcomes, illustrating that improvements in general health and improvements in occupational safety should not automatically be assumed to be the same.

      García-Madurga et al. [4] investigate AI applications in workplace well-being through a systematic review of 31

      articles published between 2018 and 2023. The applications include mental-health monitoring, emotional support, personalized well-being programs, identification of psychosocial risks, and employee training and development. The review also notes that AI-based monitoring can have adverse effects, including stress associated with surveillance technologies, and raises concerns about privacy, security, and responsible use.

      Together, these studies suggest that industrial safety programs should address both the prevention of physical injury and the maintenance of workers well-being. Health-promotion interventions may support healthier behaviors, while AI may assist with identifying psychosocial concerns and tailoring support. However, the effectiveness of specific interventions must be established through rigorous evaluation, and monitoring systems should be implemented with appropriate safeguards for worker privacy and autonomy [3], [4].

   3. Quantitative risk assessment and industrial safety management

      Makhutov et al. [5] present a risk-oriented approach in which risk is evaluated in terms of the probability of hazardous events and the resulting damage. Their framework considers harm to people, damage to industrial facilities, and environmental consequences. The authors discuss acceptable risk levels and the economic resources required to reduce risk, demonstrating the relationship between technical integrity, risk reduction, and safety-related expenditure. The approach is illustrated using nuclear power plants, so its application to other industries requires appropriate adaptation.

      Tymchenko [6] approaches industrial safety from a regional economic perspective. The proposed system uses nine indicators related to industrial and agricultural activity, energy efficiency, innovation, the technical condition of fixed assets, economic efficiency, and food stability. Its application to the Odesa, Mykolaiv, and Kherson regions of Ukraine produces different levels of assessed regional safety. Importantly, these indicators describe broader industrial and economic resilience rather than directly measuring individual workplace hazards or injury rates.

      The two studies offer different but complementary levels of analysis. Makhutov et al. [5] focus on the probability and consequences of hazardous events at industrial facilities, while Tymchenko [6] considers the wider conditions that influence regional industrial capacity and resilience. A comprehensive assessment framework could connect facility- level risk information with regional indicators, provided that the distinction between direct occupational hazards and broader economic-security measures is maintained.

   4. Machinelearning for occupational injury prediction

      Khairuddin et al. [7] investigate machine-learning methods for predicting occupational injury severity using 66,405 occupational injury records from the U.S. Occupational Safety and Health Administration (OSHA). The study compares Support Vector Machine, K-Nearest Neighbors,

      Naïve Bayes, Decision Tree, and Random Forest models. Random Forest achieves the strongest performance among the compared models. The analysis identifies the nature of injury, type of event, and affected body part as important predictive features. Hyperparameter optimization further improves the models performance for predicting hospitalization and amputation.

      This research demonstrates how historical incident data can support data-driven safety decisions. Instead of using accident records only for retrospective reporting, organizations may use predictive models to identify patterns associated with severe outcomes and inform preventive measures [7]. However, prediction performance alone does not establish that a model can prevent accidents in real operating conditions. Practical use requires suitable data quality, validation on relevant workplaces, and careful interpretation of model predictions.

   5. Safety 4.0 and digital workplace monitoring

      Haleem et al. [8] review Safety 4.0, which applies digital technologies to industrial safety management. The technologies discussed include artificial intelligence, robotics, the Internet of Things (IoT), blockchain, virtual reality, cloud computing, and data analytics. These technologies can support real-time monitoring, hazard identification, predictive maintenance, incident reporting, and worker alerts.

      For example, connected sensors can monitor equipment conditions, while wearable devices can alert workers to overexertion or hazardous areas. AI-based analysis can identify unusual patterns in operational data and help organizations respond before a potential incident develops. The authors also highlight implementation challenges, including cybersecurity, integration with existing systems, data management, and the need to train employees to use new technologies effectively.

      The findings of Khairuddin et al. [7] and Haleem et al. [8] are closely related. Machine learning provides analytical methods for learning from injury records, whereas Safety 4.0 offers a broader digital infrastructure through which safety information can be collected, analyzed, and communicated. Combining these approaches could support a transition from reactive incident management toward more proactive risk management. This potential, however, should not be interpreted as proof that all digital interventions automatically improve safety.

   6. Knowledge graphs for industrial safety knowledge management

      Yin et al. [9] address the challenge of extracting reusable knowledge from industrial safety reports, including Hazard and Operability Analysis (HAZOP) and Pre-Hazard Analysis (PHA) documentation. They propose an industrial named entity extraction method and a knowledge-graph construction framework. The framework represents safety information through entities and relationships and stores the resulting knowledge structure in Neo4j.

      In the reported experiments, the proposed extraction method improves accuracy by an average of 17 percentage points compared with the other key entity-extraction algorithms considered. In an application involving fluid transportation, the researchers construct 1,329 entities. These results demonstrate the potential of structuring unstructured safety documentation so that information can be reused and retrieved more effectively.

      The approach complements machine-learning-based injury prediction [7] and the integrated monitoring vision of Safety

      4.0 [8]. Injury prediction models learn from recorded outcomes, while knowledge graphs can organize information about hazards, equipment, causes, and safety measures. Integrating these forms of information could improve the traceability of risk assessments and support decision-making. However, the studys results do not establish that its knowledge graph can independently predict accidents or that it has been validated across all industrial sectors.

3. SYNTHESIS AND RESEARCH GAPS

   The papers collectively support a multi-layered approach to industrial safety. They cover different but related dimensions: worker health and ergonomics, injury risk, technical and environmental consequences, organizational interventions, digital monitoring, and regional industrial resilience.

   Several research gaps emerge from this synthesis.

   1. Integration of physical and psychosocial risks. Workplace ergonomics, cardiovascular load, psychological well-being, and injury severity are addressed through different approaches [1] [4], [7]. A more integrated assessment framework could help examine how these factors interact rather than treating them as separate problems.

   2. Combining real-time monitoring with predictive analytics. Safety 4.0 technologies can support data collection and continuous monitoring [8], while machine learning can identify patterns in historical injury data [7]. Further work is needed to validate how these capabilities can be combined in operational environments and whether the resulting alerts lead to measurable safety improvements.

   3. Connecting structured and unstructured safety information. Knowledge graphs can organize information extracted from safety reports [9], while injury datasets support statistical and machine-learning analysis [7]. Research could investigate how these sources can be linked without compromising data quality, interpretability, or traceability.

   4. Stronger evaluation of health and well-being interventions. The evidence on workplace health-promotion program remains heterogeneous and limited by study quality [3]. AI- based well-being applications also require evaluation of both their benefits and possible harms [4]. Future studies should use robust designs and assess outcomes over appropriate periods.

   5. Balancing technical performance with human factors. Digital systems may improve monitoring, but workers understanding, training, trust, privacy, and participation

      remain important [4], [8]. Safety technology should complement organisational responsibility and worker involvement rather than replace them.

   6. Linking facility-level safety with regional resilience. Quantitative risk assessment considers the probability and consequences of hazardous events [5], whereas regional indicator systems assess broader industrial and economic conditions [6]. Future frameworks could connect these levels while keeping their different meanings and measurement scales explicit.

   7. Demonstrating effectiveness in real workplaces. Several studies propose methods or synthesize existing research, but the effectiveness of an integrated safety system cannot be inferred merely by combining their individual findings. Field trials, external validation, and consistent outcome measures are needed to establish whether integrated approaches reduce injuries, improve health, or strengthen operational resilience.

4. CONCLUSION

The reviewed papers show that industrial safety is best understood as a combination of occupational-health protection, systematic risk assessment, evidence-based worker support, data-driven injury analysis, and technology- enabled safety management. Ergonomic design and physiological monitoring help identify risks to workers; health-promotion programs and well-being applications address broader health needs; quantitative risk assessment supports decisions about the probability and consequences of hazardous events; and machine learning, Safety 4.0,and knowledge graphs offer tools for analyzing and reusing safety information.

The main contribution of consolidating these studies is to demonstrate how their different approaches could complement one another within a unified safety-management framework. Nevertheless, the literature does not yet demonstrate that one fully integrated system incorporating all these dimensions is universally effective. Future research should focus on combining the approaches carefully, validating them in real industrial settings, protecting workers privacy, and measuring actual improvements in safety and health.

REFERENCES

1. V. C. Erazo-Chamorro, R. P. Arciniega-Rocha, N. Rudolf, B. Tibor, and

   S. Gyula, Safety workplace: The prevention of industrial security risk factors, Applied Sciences, vol. 12, no. 21, Art. no. 10726, 2022 doi: 10.3390/app122110726.

2. M. Dias, L. Silva, D. Folgado, M. L. Nunes, C. Cepeda, M. Cheetham, and H. Gamboa, Cardiovascular load assessment in the workplace: A systematic review, International Journal of Industrial Ergonomics, vol. 96, Art. no. 103476, 2023, doi: 10.1016/j.ergon.2023.103476.

3. S. Javanmardi, L. Rappelt, S. Zangenberg, L. Heinke, C. Baumgart, D. Niederer, and J. Freiwald, Effectiveness of workplace health promotion programs for industrial workers: A systematic review, BMC Public Health, vol. 25, Art. no. 168, 2025, doi: 10.1186/s12889- 025-21365-8.

4. M.-Á. García-Madurga, A.-I. Gil-Lacruz, I. Saz-Gil, and M. Gil- Lacruz, The role of artificial intelligence in improving workplace well-being: A systematic review, Businesses, vol. 4, no. 3, pp. 389 410, 2024, doi: 10.3390/businesses4030024.

5. N. Makhutov, V. Nadein, M. Gadenin, and D. Reznikov, Risk and industrial safety, Reliability: Theory & Applications, vol. 17, Special Issue no. 3 (66), pp. 138143, Jan. 2022.

6. K. Tymchenko, Methodological aspects for the formation of a system of indicators for assessing industrial safety at the regional level, Economics & Education, vol. 10, no. 4, pp. 717, Dec. 2025, doi: 10.30525/2500-946X/2025-4-1.

7. M. Z. F. Khairuddin, P. Lu Hui, K. Hasikin, N. A. Abd Razak, K. W. Lai, A. S. Mohd Saudi, and S. S. Ibrahim, Occupational injury risk mitigation: Machine learning approach and feature optimization for smart workplace surveillance, International Journal of Environmental Research and Public Health, vol. 19, no. 21, Art. no. 13962, 2022, doi: 10.3390/ijerpp92113962.

8. A. Haleem, M. Javaid, and R. P. Singh, Encouraging Safety 4.0 to enhance industrial culture: An extensive study of its technologies, roles, and challenges, Green Technologies and Sustainability, vol. 3, Art. no. 100158, 2025, doi: 10.1016/j.grets.2024.100158.

9. Z. Yin, L. Shi, Y. Yuan, X. Tan, and S. Xu, A study on a knowledge graph construction method of safety reports for process industries, Processes, vol. 11, no. 1, Art. no. 146, 2023, doi: 10.3390/pr11010146.