Risk Analysis using Monte Carlo Simulation instead of Scoring and Heat Map

Discussion in 'FMEA - Failure Modes and Effects Analysis' started by Katana Clarity, Oct 26, 2020.

Tags:
1. Katana ClarityNew Member

Joined:
Oct 18, 2019
Messages:
1
0
Trophy Points:
1
The other day I came across an article that mentioned how the financial industry managed risk using Monte Carlo simulations and the experts in that industry had a disdain for risk assessment done by typical scoring 1-10 likelihood, occurrence and severity method. The argument was that the latter does not have any evidence of being effective, rather it gives a false sense of confidence that risk is being managed appropriately.

I was wondering if anyone in Medical device Industry has ever employed Monte Carlo Simulations for Device Risk Assessment in replacement of Scoring method, or perhaps employed both in some combination.

I have no idea how MCS need to be customized to cater our needs but I love to see an example of this if anyone's made an attempt at it.

2. Bev DModeratorStaff Member

Joined:
Jul 30, 2015
Messages:
618
679
Trophy Points:
92
Location:
Maine
My organization doesn’t use occurrence or detection ratings at all. Only severity.
We DO perform characterization studies on teh high severity failure modes to understand the system and set specifications on the inputs that ‘guarantee’ good outputs and very low failures. We also perform validation experiments to determine the occurence levels (OQ/PQ)

We will occasionally use Monte Carlo - or better yet bootstrapping - to determine the failure rate on ‘marginal’ processes that don’t Comfortably meet specifications.

3. MinerModeratorStaff Member

Joined:
Jul 30, 2015
Messages:
597
518
Trophy Points:
92
Location:
Greater Milwaukee USA
Monte Carlo is an excellent tool for establishing the probability that a requirement will or will not be met. It is used frequently in Finance to establish risk, and can be used in design and reliability to do the same. However, it does require that you start with a mathematical model and are able to estimate the variation expected in the inputs and in the noise variables. If you are lacking the mathematical model or have no information on variation, you cannot use Monte Carlo analysis.

Atul Khandekar and Bev D like this.