Fault Tree Analysis (FTA) is the cornerstone of functional safety system development. However, the process of identifying and connecting thousands of failure causes is an arduous task that demands immense time and effort from engineers.
Today, we would like to introduce a method that leverages VisualPro, VWAY’s safety analysis platform, to support FTA execution by combining AI technology with the PSC (Primary, Secondary, Command) technique—a standard method for
1. The Challenge of FTA: Battling Complexity
FTA, which logically identifies the root causes of system failures, has been the standard for proving safety for decades. However, as systems become increasingly sophisticated, engineers must connect thousands or even tens of thousands of gates. This complexity often leads to human errors and an exponential increase in analysis time.
Existing versions of VisualPro have provided convenience through an intuitive UI for gate editing and tree construction. Now, we take a step further by presenting a way to utilize Generative AI as an analysis 'Assistant.'
2. Core Solution: AI-Powered Cause Inference Based on PSC
To prevent logical omissions during FTA development, the PSC technique is a widely used methodology. Instead of simply asking the AI model to "find the next gate," we query it through a structured PSC framework. Below are examples of causes identified by AI based on the PSC criteria:
- Primary (Primary Failure): Failure due to the durability limits or inherent defects of the component itself.
- Secondary (Secondary Failure): Failure caused by excessive environmental stress (e.g., heat, vibration).
- Command (Command/Signal Failure): Signal errors that cause the system to fail to operate at the intended time or to malfunction.
3. Mitigating Hallucinations: Leveraging 'Cut Set' Technology
The primary concern when applying AI to engineering is the 'hallucination' phenomenon. To resolve the issue of AI generating plausible but incorrect data, we utilize VisualPro’s Cut Set data.
We do not merely ask for recommendations for lower-level gates; we provide the AI with the context of upper-level gates and the currently connected Minimal Cut Sets (MCS) information. This allows the AI to understand the logical flow of the entire system and filter only the sophisticated failure causes that fit the context before presenting them to the engineer.
4. Building a Secure AI Environment: On-Premise & Frontier Models
VWAY supports two AI integration methods tailored to our clients' data security policies and needs:
- Integration with Frontier Models: We provide top-tier performance by integrating with the latest high-performance AI models (e.g., GPT-4o, Claude) via API.
- Internal On-Premise Deployment: For clients in sensitive sectors such as defense, automotive, and aerospace where data leakage is a concern, we establish security-enhanced sLLMs (small Large Language Models) on the client's internal servers to support AI functions safely without external communication.
5. Conclusion: The Future of Safety Analysis with VisualPro
Supporting various safety analysis techniques such as FTA, FMEA, STPA, and TARA, VisualPro is evolving beyond a simple 'tool' into an 'Intelligent Partner' for engineers.
Delegate repetitive tasks to AI, and let your engineers focus on more critical safety design and verification. VWAY will build the most convenient and secure AI analysis environment optimized for your data infrastructure.
Fault Tree Analysis (FTA) is the cornerstone of functional safety system development. However, the process of identifying and connecting thousands of failure causes is an arduous task that demands immense time and effort from engineers.
Today, we would like to introduce a method that leverages VisualPro, VWAY’s safety analysis platform, to support FTA execution by combining AI technology with the PSC (Primary, Secondary, Command) technique—a standard method for
1. The Challenge of FTA: Battling Complexity
FTA, which logically identifies the root causes of system failures, has been the standard for proving safety for decades. However, as systems become increasingly sophisticated, engineers must connect thousands or even tens of thousands of gates. This complexity often leads to human errors and an exponential increase in analysis time.
Existing versions of VisualPro have provided convenience through an intuitive UI for gate editing and tree construction. Now, we take a step further by presenting a way to utilize Generative AI as an analysis 'Assistant.'
2. Core Solution: AI-Powered Cause Inference Based on PSC
To prevent logical omissions during FTA development, the PSC technique is a widely used methodology. Instead of simply asking the AI model to "find the next gate," we query it through a structured PSC framework. Below are examples of causes identified by AI based on the PSC criteria:
3. Mitigating Hallucinations: Leveraging 'Cut Set' Technology
The primary concern when applying AI to engineering is the 'hallucination' phenomenon. To resolve the issue of AI generating plausible but incorrect data, we utilize VisualPro’s Cut Set data.
We do not merely ask for recommendations for lower-level gates; we provide the AI with the context of upper-level gates and the currently connected Minimal Cut Sets (MCS) information. This allows the AI to understand the logical flow of the entire system and filter only the sophisticated failure causes that fit the context before presenting them to the engineer.
4. Building a Secure AI Environment: On-Premise & Frontier Models
VWAY supports two AI integration methods tailored to our clients' data security policies and needs:
5. Conclusion: The Future of Safety Analysis with VisualPro
Supporting various safety analysis techniques such as FTA, FMEA, STPA, and TARA, VisualPro is evolving beyond a simple 'tool' into an 'Intelligent Partner' for engineers.
Delegate repetitive tasks to AI, and let your engineers focus on more critical safety design and verification. VWAY will build the most convenient and secure AI analysis environment optimized for your data infrastructure.