Failure types and probability
Thursday 12/2/10 time 2:38 PM - Andy
By collecting and analyzing failure data (if available) and fitting failure rate functions gives a feeling that this is a true engineering science. However, there are other types of product failures or faults, which may be much more important than probabilistic failures.
Different failure types: According to IEC 61508 the failure analysis and fault removal efforts should cover the complete set of failures, namely:
Probabilistic failures are often considered as 'inherent failure behavior' and probabilistic functions can be fitted to study these mathematically. They are often described with 'bathtub' curve with 1) early failures or 'infant mortality', 2) steady state failures and 3) wear-out failures or aging.
The biggest problem is that it is very difficult to get reliable failure data, especially for new product types and new technology, simply because the new technology doesn't exist 'until it's too late'. Probabilistic failures a presumed to be independent of each other, which presents a huge problem.
Systematic faults are deterministic by nature. Some examples of these types of failures are
Common cause faults: These faults present a category of dependent failures. In practice they are difficult to study with probabilistic means. However, because they are likely to induce a severe sequence of events leading to an accident, they are a very important category, especially in safety critical applications.
Reliability Academy Course 'Risk management Systems' discusses the practical techniques and procedures to control the product failure categories, especially the systematic faults.
System States: In order to control these 3 failures types it is important that the relevant system or product operational modes are clearly defined and studied (IEC 61508-2, p.33):
The system states alter the system reliabilityt structure and corresponding actions in these states. The effects of a failure may be different in different system states.