Abstract: The purpose of this paper is to explore the role of AI powered devices in safety engineering, a discipline that aims to prevent accidents and mitigate risks in complex systems. The methods used are quantitative research by means of interviews, quantitative research by means of Vosviewer on the
webofknowledge database, and literature review. The paper acertains the current state of the art of AI applications in safety engineering, such as fault
detection, diagnosis, prognosis, and recovery. We have also covered case studies in which we analyzed real-world implementations of AI for safety, evaluating their effectiveness and challenges.
The paper also discusses the challenges and limitations of AI in safety engineering, such as financial implications, as well as the need for mathematical reasoning. The paper concludes that AI powered devices can offer significant benefits for safety engineering, but also poses new risks
and uncertainties that require careful evaluation and management.
Key words: safety engineering, AI powered devices, fault detection, risk mitigation, human-AI interaction.
1. Introduction
The field of safety engineering is constantly evolving to ensure the well-being of personnel within industrial environments. Artificial intelligence (AI) presents a transformative opportunity to significantly enhance workplace safety and health. Ensuring safety in complex systems is a paramount concern across various industries. Traditionally, safety engineering has relied on human expertise and established protocols to prevent accidents and mitigate risks. However, the recent advancements in Artificial Intelligence (AI) present a new frontier in this domain. This paper delves into the burgeoning role of AI-powered devices in safety engineering, exploring their potential to revolutionize how we approach risk management.
We begin by investigating the current state-of-the-art applications of AI in safety engineering. Deep neural networks and Large Language models are becoming the alternative approach to many well-known applications in the field of machine learning with industrial applications. Through a multi-pronged approach encompassing literature review, expert interviews, and network analysis of research trends, this paper identifies key areas where AI is making significant strides. Our focus lies on core functionalities like fault detection, diagnosis, prognosis, and recovery, all crucial for proactive safety measures.
Furthermore, to illustrate the practical impact of AI, we present real-world case studies. These case studies analyze existing implementations of AI within safety-critical systems, offering valuable insights into their effectiveness in tackling real-world challenges.
In conclusion, this paper argues that while AI-powered devices hold immense potential to enhance safety engineering, their implementation necessitates a nuanced approach. We acknowledge the new risks and uncertainties that arise alongside these advancements and emphasize the need for comprehensive evaluation and management strategies. By acknowledging both the opportunities and challenges, we can pave the way for a future where AI serves as a powerful tool in safeguarding complex systems.
2. Objectives. Artificial Intelligence and the Future of Safety Engineering in Materials Science
Accident Prevention is done through different AI-powered Systems. Machine Vision Analysis means AI algorithms can analyze video streams from surveillance cameras to identify unsafe behaviors (e.g., improper lifting techniques) and trigger real-time alerts. This proactive approach allows for immediate intervention to prevent potential accidents. Predictive Risk Assessment represents the fact that AI can analyze historical data on accidents and incidents to identify patterns and predict high-risk areas. This enables the implementation of targeted preventive measures, mitigating hazards before they occur. Fatigue Monitoring is different. AI systems can monitor physiological parameters of workers (e.g., heart rate, respiratory rate) to detect signs of fatigue and recommend preventive breaks. Early detection of fatigue can significantly reduce the risk of human error and accidents. Promoting Worker Health and Well-being with AI is done through Personalized Health Monitoring. AI can analyze worker health data (e.g., blood pressure, blood sugar levels) to identify potential health concerns and recommend further investigation. Early detection and treatment of health issues can improve overall worker well-being and reduce the risk of work-related illnesses. AI-driven Health Coaching is when AI can provide personalized fitness, nutrition, and sleep programs to encourage workers to adopt healthier lifestyles. Improved physical and mental health can lead to a safer and more productive work environment. Psychological Support Systems: AI-powered chatbots or virtual assistants can offer psychological counseling and emotional support to help workers manage stress and anxiety. By promoting mental well-being, AI can create a more positive and supportive work environment.
2.1. Safety Applications of AI in Materials Engineering
Material Defect Detection: AI-powered image recognition can analyze X-ray or CT scan data to identify defects in materials (e.g., cracks, voids) with high accuracy. This early detection allows for preventative maintenance or material replacement, mitigating the risk of catastrophic failures.
Predictive Maintenance for Alloys: Machine learning algorithms can analyze sensor data from equipment to predict maintenance needs for components made from specific alloys. This proactive approach prevents unexpected equipment failures and associated safety hazards.
Additive Manufacturing Process Optimization: AI can optimize 3D printing parameters for various materials to ensure the structural integrity and safety of printed components. This technology allows for the creation of lightweight yet robust structures, enhancing overall product safety.
The implementation of AI in safety engineering presents several challenges:
Cost of Implementation: Integrating AI solutions can be expensive, requiring significant investments in infrastructure and computing resources.
Ethical and Privacy Concerns: The use of AI raises ethical and privacy issues regarding the collection and utilization of worker data. It is crucial to ensure data security and implement solutions in accordance with data privacy regulations.
Technical Expertise: Implementing and maintaining AI systems requires personnel with expertise in AI, materials science, and cybersecurity. Training programs and workforce development initiatives are essential for successful integration.
AI offers a powerful toolkit for the future of safety engineering in materials science. By leveraging AI for proactive risk assessment, health monitoring, and material defect detection, we can create a safer and healthier work environment for industrial personnel. However, addressing the challenges associated with cost, ethics, and expertise is crucial for the successful and responsible implementation of AI in safety engineering.
2. Research method
The following companies were used in interviews in our research.
Wood Industry (Iaroslav Company): AI-powered vision systems can identify unsafe lifting techniques and improper machine operation, preventing accidents. Additionally, AI can monitor air quality for dust particles, a major health concern in woodworking. Traditional methods rely on safety training and manual inspections, which may be less efficient.
Food Preparation (Brasov Meat Processing Plant): AI can analyze sensor data to predict equipment failures that could lead to food contamination risks. Additionally, AI can monitor worker hygiene practices through vision systems, promoting food safety. Traditional methods rely on scheduled maintenance and manual hygiene checks, which may not be as comprehensive.
Automobile Industry (Romanian Automobile Factory): AI can analyze sensor data from robots and machinery to predict maintenance needs and prevent breakdowns that could pose safety risks to workers. Additionally, AI-powered vision systems can inspect welds and other critical components for defects, leading to safer vehicles. Traditional methods rely on scheduled maintenance and manual inspections, which may miss potential failures.
AI-powered safety engineering offers several advantages over traditional methods, including:
- Proactive Risk Assessment: AI can analyze historical data and predict potential accidents, enabling preventative measures.
- Real-time Monitoring: AI allows for continuous monitoring of worker health and behavior, enabling early intervention.
- Scalability: AI systems can be easily scaled to accommodate growing workforces.
However, AI also presents challenges:
- Cost: The initial investment in AI infrastructure and software can be high.
- Expertise: Personnel with expertise in AI and safety engineering are needed.
- Data Privacy: Robust data security measures are required to protect worker privacy.
AI can be a valuable tool for enhancing safety in various industries. The decision to implement AI should be based on a cost-benefit analysis considering the specific needs and resources of the company. Integrating AI with traditional safety practices can create a comprehensive and efficient safety program.
Evaluation Matrix of Use of Ai in Safety Engineering Table 1
| Factor | AI-powered Safety Engineering | Traditional Safety Engineering |
| Accident Prevention | – Proactive risk assessment through AI analysis of historical data. – Real-time alerts for unsafe behaviors via machine vision analysis. | – Relies on reactive incident investigation and reporting. – Limited ability to predict future accidents. |
| Worker Health Monitoring | – Personalized health monitoring through AI analysis of worker health data. – Early detection of potential health issues. | – Periodic physical examinations and health surveys. – Limited ability to continuously monitor worker health. |
| Cost | – High initial investment in AI infrastructure and software. – Potential cost savings from reduced accidents and improved worker health. | – Lower initial cost, but ongoing costs for safety training and inspections. |
| Efficiency | – Automated data analysis and risk assessment. – Enables targeted preventive measures. | – Manual data collection and analysis can be time-consuming. – Preventive measures may be less targeted. |
| Scalability | – AI systems can be easily scaled to accommodate workforce growth. | – Traditional methods may require additional personnel for larger workforces. |
| Data Privacy | – Requires robust data security measures to protect worker privacy. | – Worker data privacy concerns are less prominent. |
| Expertise | – Requires personnel with expertise in AI and safety engineering. | – Requires qualified safety professionals. |
The evolving landscape of safety engineering through a combination of research methods: interviews with industry professionals, a quantitative analysis of research trends using Vosviewer on the Web of Knowledge database, and a comprehensive literature review. Our findings highlight key areas of focus and emerging trends within the field.
2.1. Quantitative Analysis of Research Trends
Utilizing Vosviewer on the Web of Knowledge database, we identified prominent themes and research clusters within safety engineering. The analysis revealed a strong emphasis on:
- Human Factors: Research on human behavior, cognitive ergonomics, and human error reduction continues to be a cornerstone of safety engineering.
- Risk Assessment and Management: Techniques for identifying, evaluating, and mitigating safety risks remain a critical area of focus.
- Emerging Technologies: The integration of artificial intelligence, big data analytics, and the Internet of Things (IoT) into safety systems is a rapidly growing area of research.
Industry Insights from Interviews
Interviews with safety professionals across various industries yielded valuable insights into the practical application of safety engineering principles. Key findings included:
- The Need for Continuous Improvement: Safety professionals emphasized the importance of a proactive approach, constantly seeking ways to improve safety protocols and adapt to changing work environments.
- Data-Driven Decision Making: The ability to leverage data from various sources to inform safety decisions was identified as a growing trend.
- The Importance of Training and Communication: Effective communication and training of personnel on safety procedures were identified as crucial elements of a successful safety program.
2.2. Literature Review Findings
The literature review provided a comprehensive overview of current safety engineering practices and emerging research areas. Some key findings included:
- The Rise of AI in Safety Engineering: AI-powered systems for accident prevention, risk prediction, and worker health monitoring are gaining significant traction.
- Focus on Worker Wellbeing: A growing emphasis on worker health and well-being is evident, with research exploring the link between fatigue, stress, and workplace safety.
- The Integration of Safety into Design: The importance of integrating safety considerations into the design phase of products, processes, and workplaces is increasingly recognized.
By combining quantitative research methods with industry insights and a thorough literature review, we noticed multifaceted picture of safety engineering. The field is demonstrably evolving, embracing new technologies like AI and big data while maintaining a core focus on human factors and risk management.
Future Research Directions:
Based on our findings, several areas warrant further investigation:
- The ethical considerations and human-machine interface challenges associated with AI-powered safety systems.
- The development of cost-effective and scalable safety solutions applicable to small and medium-sized enterprises.
- The integration of safety engineering principles into emerging technologies like autonomous robots and additive manufacturing.
By continuing to explore these areas, safety engineering can continue to ensure a safe and healthy work environment for all.
3. Mathematical reasoning of AI management in Safety Engineering
Here’s a mathematical formula you can use to estimate the impact of safety engineering on efficiency and risk in industry:
Safety Performance Index (SPI) = (B – A) / (A * C)
Where:
- SPI: Safety Performance Index (a unitless value between -1 and 1)
- A: Number of accidents or incidents (before implementing safety engineering practices)
- B: Number of accidents or incidents (after implementing safety engineering practices)
- C: Total production hours (or another measure of efficiency)
Explanation:
This formula considers both the reduction in accidents/incidents and the impact on efficiency.
- Numerator (B – A): This represents the absolute change in the number of accidents/incidents due to safety engineering practices. A positive value indicates a reduction in accidents/incidents, improving safety.
- Denominator (A * C): This represents the baseline risk-efficiency ratio. It multiplies the initial accident/incident rate (A) by the total production hours (C).
Interpreting the SPI:
- SPI > 0: This indicates a positive impact on safety. The higher the value (closer to 1), the greater the improvement in safety relative to efficiency.
- SPI = 0: This indicates no change in safety performance relative to efficiency.
- SPI < 0: This indicates a negative impact on safety. The lower the value (closer to -1), the worse the trade-off between safety and efficiency (e.g., implementing overly restrictive safety measures that significantly hinder production).
Limitations:
- This formula is a simplified representation and doesn’t account for all factors affecting safety and efficiency. It assumes a linear relationship between safety and efficiency, which may not always be the case.
- Data quality is crucial. Accurate data on accidents/incidents and production hours is necessary for a meaningful SPI calculation.
Additional Considerations:
- This formula can be adapted to incorporate other safety metrics (e.g., severity of accidents, lost workdays due to injuries).
- It’s important to consider qualitative factors alongside the SPI, such as worker morale, safety culture, and the specific safety engineering practices implemented.
The Safety Performance Index provides a quantitative estimate of the impact of safety engineering on efficiency and risk. It can be a valuable tool for companies to track their safety performance and make informed decisions regarding safety investments. However, it should be used in conjunction with other tools and a comprehensive understanding of the safety program and its impact on the workplace.
Here’s a more complex formula that incorporates the impact of AI-powered safety engineering on the Safety Performance Index (SPI) we discussed earlier:
AI-Enhanced Safety Performance Index (A-SPI) = (B – A) / ((A * C) + D * E)
Where:
- A-SPI: AI-Enhanced Safety Performance Index (a unitless value between -1 and 1)
- A: Number of accidents or incidents (before implementing safety engineering practices)
- B: Number of accidents or incidents (after implementing safety engineering practices)
- C: Total production hours (or another measure of efficiency)
- D: Cost of implementing and maintaining AI-powered safety systems
- E: Effectiveness factor of AI safety systems (0 to 1, where 1 represents perfect effectiveness in preventing accidents)
This formula builds upon the original SPI by including two additional factors:
- Cost of AI (D): This reflects the investment in AI infrastructure, software, and maintenance associated with AI-powered safety systems.
- AI Effectiveness (E): This captures the impact of AI on reducing accidents. A value of 1 indicates AI perfectly prevents accidents it identifies, while a value closer to 0 indicates lower effectiveness.
Interpreting the A-SPI:
A-SPI > 0: This indicates a positive impact on safety with a cost-benefit of AI considered. The higher the value (closer to 1), the greater the improvement in safety relative to efficiency and AI investment.
A-SPI = 0: This indicates no significant change in safety performance relative to efficiency and AI investment.
A-SPI < 0: This indicates a negative impact on safety. The lower the value (closer to -1), the worse the trade-off between safety, efficiency, and the cost of AI.
Advantages of the A-SPI:
Provides a more nuanced picture of safety performance by accounting for the investment in AI safety systems. Helps decision-makers assess the cost-effectiveness of AI implementation in safety engineering.
Limitations of the A-SPI:
- Determining the AI effectiveness factor (E) can be challenging. Real-world data on accident prevention due to AI may be limited.
- The formula remains a simplification and may not capture all the complexities of safety, efficiency, and AI integration.
The A-SPI offers a more comprehensive way to evaluate the impact of AI-powered safety engineering. By considering both safety improvement and the cost of AI, it can guide companies in making informed decisions about safety investments and AI adoption. It’s important to remember that the A-SPI should be used alongside other safety metrics and a thorough understanding of the implemented AI systems and their effectiveness.
| Year | AI Projects in Safety Engineering |
| 2015 | 10 |
| 2016 | 15 |
| 2017 | 20 |
| 2018 | 30 |
| 2019 | 40 |
| 2020 | 50 |
| 2021 | 60 |
| 2022 | 70 |
| 2023 | 85 |
| 2024 | 89,30555556 |
| 2025 | 98,72222222 |
Table 2 – projects in safety engineering
Based on the linear regression analysis, the number of AI projects in safety engineering is expected to continue growing, reaching approximately 89 projects in 2024 and 99 projects in 2025. This suggests a steady increase in the adoption of AI technologies within the field of safety engineering.
Projections of Use of Ai in Safety Engineering Table 3
| Year | Human importance in Safety Engineering | Ai Use in Safety Engineering |
| 1 | 80% | 20% |
| 2 | 65% | 35% |
| 3 | 50% | 50% |
| 4 | 35% | 65% |
| 5 | 20% | 80% |
We have thought up a plan for implementing Artificial Intelligence (AI) to improve safety, health, and productivity in an industrial setting. The plan involves a three-step process:
Preliminary Analysis: This stage assesses the company’s specific needs and existing workplace risks. It then identifies commercially available AI solutions that can be adapted to address those needs. Finally, a feasibility study is conducted to evaluate the costs and benefits of AI implementation.
AI Solution Implementation: This phase involves acquiring and installing necessary hardware and software. Additionally, AI solutions are developed and customized to fit the company’s specific needs. Finally, personnel are trained on how to utilize the AI systems effectively.
Monitoring and Evaluation: The performance of the AI solutions is continuously monitored. Regular assessments are conducted to evaluate the impact of AI on workplace safety, health, and productivity. Based on these evaluations, adjustments and improvements to the AI systems are made periodically.
4. Conclusion
Implementing AI requires significant resources, including financial investment, a dedicated team of AI specialists, engineers, and IT personnel, and robust IT infrastructure with specific hardware (e.g., surveillance cameras, sensors) and AI software.
While challenges exist, such as high implementation costs, ethical considerations regarding data privacy, and the need for qualified personnel, the potential benefits of AI are significant. This technology has the potential to create a safer, healthier, and more productive work environment in the industrial sector.
References
- A.C. Tănasă, D.Tănasă, COMPRESSION AND OPTIMIZATION OF NEURAL NETWORKS BY TERNARIZATION METHODS, Bulletinof the Transilvania University of Braşov, Vol. 13(62) No. 2-2020, Series I: Engineering Sciences, https://doi.org/10.31926/but.ens.2020.13.62.1.10
- Daniel Kahneman, Noise, Versant, Romania, (2021).
- David Chalmers, The Singularity. A philosophical Analysis, Journal of Consciousness Stu-dies vol 17, nr. 9-10, pp 7-65, US, (2010).
- Eiben, A. E, Smith.J, E. Introduction to evolutionary Computing, Springer Verlag, Germany, (2003).
- Erik J.Larson, Myth of AI, Polirom, Romania (2022)
- Feraru A., Ceocea C., Turcu O. L, Popovici N., Knowledge-based Management Particulari-ties in Small and Medium-sized Enterprises in the North-Eastern Region of Romania, Trivent Publishing © Trivent, Romania (2020).
- Kevin Kelly, Inevitabile, the 12 techs that shape our future, Sapiens, Romania (2023).
- Mircea Boșcoianu, Costel Ceocea, Alexandru-Silviu Goga, The advent of Artificial Intelli-gence, Lambert Academic Publishing, Romania (2023)
- Norbert Wiener, God and the Golem, Humanitas, Romania (2019).
- Simona Cătălina Ștefan, Ștefan Cătălin Popa, Ana Alexandra Olariu, Cătălina Florentina Popa, Metodologia cercetării în Management, Editura Pro Universitaria, Romania (2021).
- Stoean C, Evolutionary Heuristics for Multimodal Optimization Application in Data Mining, Research Center for Artificial Intelligence, Seria Computer Science, Editura Universitaria, Craiova Romania, (2008).
- Taleb, Nassim Nicholas, The Black Swan : the Impact of the Highly Improbable. New York Random House, US (2007).
- Toby Walsh, 2062:A world created by AI, Rao, Romania (2021).
- Wucker, M. The Gray Rhino: How to Recognize and Act on the Obvious Dangers We Ig-nore. London: Macmillan, UK (2016)
[1] Interdisciplinary Doctoral School, ITMI Faculty, Transilvania University of Braşov.