...
- Methods for developing reliability requirements for safetyfor safety-critical systems and barriers, with basis in risk analyses
"Safety integrity level (SIL) is a key reliability performance measure used for safety-critical systems. The SIL requirements are identified in an extension of the risk analysis, using methods often refered to as SIL allocation, SIL targeting and SIL classification. Key methods like Layers of protection analysis (LOPA), risk graph, and minimum SIL are presented and discussed." - Extension of methods for quantifying the reliability of safety-critical functions - analytical approaches & dynamic modeling approaches.
"In TPK 4120, some analytical formulas were introduced to calculate the average probability of failure on demand (PFD). It was also shown how the average PFD may be calculated using Markov methods and fault tree analysis. This reliability measure is of high importance in relation to SIL, as a relationship is established between a SIL requirement and the maximum PFD tolerated for a safety function."
In this course, we go one step further and:- Introduce some other methods for quantifying the average PFD:
- The analytical formulas presented in a standard called IEC 61508 (in part 6), which builds on slightly different assumptions than the analytical formulas from TPK 4120.
- The analytical formulas in the PDS method
- Petri Nets, which is an alternative and more flexible approach than e.g., Markov method state models.
- Study reliability of "high demand systems", where another reliability measure, the average system failure rate (called PFH), is recommended rather than the average PFD. One example of a high demand safety system is a machine that carry out safety-critical functions. Also PFH is linked to SIL.
- Introduce some other methods for quantifying the average PFD:
- Monitoring and maintaining SIL performance in the operational/use phase.
"The reliability of a safety-critical function is influenced over time after the system has been put in operation. Just like if you buy a car: We may think that the car has some kind of inherent reliability performance in light of what it costs, the type of engine, manufacturer reputation, safety systems installed with the car and so on. Once you start to drive it, its performance may change over time depending on your driving habits, how much you drive, where you drive, how often you send it to the garage for maintenance and checks and os on. You may collect some data about the car's performance, such as how often it does not start "on demand", milage, and how often some of the safety-features fail, and based on this (often limited information as you should not have much failures) you may try to estimate the reliability. In fact, you are trying to estimate the reliability as it has been up till a certain point in time. "
It is the same thing we would like to do with a safety-critical system: With rare data we would like to estimate the reliability using the information that we have. If the performance is not sufficient (in light of e.g. the SIL requirement), we need to do something. This "something" is also discussed as part of this topic.
...