After years of working with Safety Instrumented Systems (SIS) and conducting SIL verifications for major oil and gas operators, I’ve come to a controversial conclusion: exSILentia and similar expensive SIL calculation software are approaching their natural death.
I’ve got 30 years in the industry and at some point TUV certified so I’ve tested this approach below, it works.
What exSILentia Actually Does
Strip away the fancy UI and database-driven architecture, and exSILentia performs one core function: calculating PFDavg (Average Probability of Failure on Demand) using basic Boolean logic.
At its heart, every SIL calculation reduces to simple fault tree analysis:
- OR gates (series configuration): If any component fails, the system fails
- Formula: PFD_total ≈ PFD₁ + PFD₂ + PFD₃ (for small probabilities)
- AND gates (parallel/redundant configuration): All components must fail for system failure
- Formula: PFD_total = PFD₁ × PFD₂ × PFD₃
AI Can Generate Calculations On-Demand
With Claude, you could do the following stepwise
- Request custom SIL calculation templates
- Generate fault trees automatically
- Create audit-ready documentation in minutes
- Verify calculations against IEC 61511 formulas
Want a 2oo3 voting logic calculation with diagnostic coverage? Ask Claude. It’ll give you the formula, explain the logic, and even the results in excel
The Only Real Value: Historic Failure Rate Data (FEED only)
I’ll give exSILentia credit where it’s due – their certified database of failure rates (λDU, λDD values) from OREDA, PDS, and exida SERH is valuable. Having pre-certified failure rates for transmitters, solenoid valves, and logic solvers saves time during FEED and conceptual design phases.
But here’s the problem: This data advantage evaporates during detailed design.
In Detailed Design, We Use Actual Vendor Data
When you’re doing detailed SIL verification for construction, you don’t use generic database values. You use:
- Manufacturer-certified FMEDA reports from the actual vendors (Emerson, Honeywell, Siemens, ABB)
- Specific PFD values for the exact model numbers being installed
- As-installed configurations with actual proof test procedures and intervals
At this stage, exSILentia becomes an expensive calculator performing additions and multiplications you could do in Excel – or even better, automate with a simple Python script or have Claude generate for you.
Maybe during FEED you might not have actual data, but as the title says during verification, at the Detail Design Stage could you not have done it using an expensive peice of Kit.
A few prompts in an hour and you would have fixed it . You might need some prep work to have an actual PFD on a piece of Excel. Even Claude can read multiple datasheets and gather PFD for you.
Why Claude , we have tried everything Claude , Chat Gpt and Grok, Gemini , Claude seems to do a better job just in this instance with the chat interface and reading excel and understanding contents
The AI Advantage: Few-Shot Prompting for SIL Calculations
Here’s where it gets interesting. When I mentioned AI can replace exSILentia, I don’t mean just asking “calculate my SIF” and hoping for the best. The real power comes from few-shot prompting – giving Claude a few examples of your exact calculation methodology, and it will replicate it perfectly across your entire SIL verification workbook. Its easy actaully because as above its just boolean algebra. If you dont know boolean algebra you should not be reading this
What is Few-Shot Prompting?
Just give it worked examples before and after the calculations for a few examples using actualls . It will do the rest easily after that. Link below
https://www.promptingguide.ai/techniques/fewshot
Conclusion
Use AI , Claude is recommended for SIL verification work. There is simply so many benefits of doing so, you get to calogue all the actuall PFD nicely onto an excel sheet and calculate. And you can change and check like even things like proof testing etc