SIL Calculation Question

[Sierladas]

Hello,

If I have a mission time for a "safety relay" and after 48 months it achieves a SIL 3, what is its failure rate?

[ctci]

Hi,

Since your safety relay got 3 SIL, it means the Probability of failure on Demand ranges 0.001 - 0.0001.

You can assume that the PFDaverage = (0.001+0.0001)/2 = 0.00055.

Then you have:

PFDaverage = (lamda*t)/2 --> lamda = 2*PFDaverage/t

in which: t = mission time = 48 months.
lamda = failure rate

[cosili]

Hy, for all of you.
Please pay attention, it is available only if you have exponential low, regarding the repartition of your equipment.
For example if you use Weibull it is necessary another approach.
Please contact me if you need more details.
Regards

[karzy]

Hello,

I would like to point you in the correct direction with regards to the SIL for the SIS.

The SIL is not applied to the single element within a SIS loop. however, the SIL is applied to the entire Loop such that each component within the loop weather it is a transmitter, relay, and valve solenoid etc.

The sum of the failure rates of the entire loop (ignoring 1oo2 or 2oo3 voting systems as these will reduce the PFDav) as well as factoring the Proof Test Frequency(From IEC 61511 part 3) (mission time as stated by some people).

For a new system you need to set a target test frequency i.e 1 year as a minimum this is then fed into the calculation to ensure that the loop meets the required SIL target from the risk assessment process and the safety allocation process. This is where it gets complicated as you can change the test frequency and the SIL of the loop can change from a SIL 2 to a SIL 1 depending on where you go with the testing and the component you use.

One note on failure rate data from manufacturers. The failure rate data which is sometimes used by manufacturers in the "Proven in Use", category should not be used as the way the data is generated is from manufacturers returns data and this data can be corrupted as the majority of returns would only be provided during the warranty period and would proved a bad data set to generate the failure rates, even when a certificate is present the rates a based on this data. You just need to be careful in this area and ensure that users data is used to generate the failure rate data.

Regard

Paul K
BEng, MSc, MIET