Risk analysis is an integral part of working in the pharma and biotech industry. The ICH Q9 guideline describes exactly what it is used for and what risk analysis tools are proposed. Some of the most popular are FMEA, RRF, PHA, Ishigava diagram, HACCP etc. A few months ago, for a consulting project, I started using another risk analysis tool that is not described in ICH Q9: Fault tree analysis. And then I fell in love with this tool. Why? Today’s post is about that.
Fault tree analysis – a brief history.
Fault Tree Analysis (FTA) is not a new technique. It was developed in 1962 by H.A. Watson working then for Bell Laboratories. This company was involved in the development of the Minuteman I intercontinental missile. The tool was quickly adopted by Boeing and AVCO in subsequent military projects. Starting in 1966, Fault tree analysis began to be applied to civil aircraft projects. Very quickly the use of FTA expanded beyond Boeing and by the 1970s the use of the technique was already widespread in aviation. FTA was also used by NASA although initially with little interest. This was because the probability of success of the APOLLO mission based on FTA was unacceptably low. So it was decided to use other tools, e.g. FMEA. The situation changed after the Challenger disaster. It was then that the effectiveness of the FTA technique was realized and its application to improve space flight safety.
In the 1980s FTA was adopted by the chemical industry as a multifunctional tool for system and process evaluation.
How to conduct a Fault tree analysis?
Fault tree analysis is a top-down risk assessment. The problem is solved by starting from the final effect and breaking the system into smaller and smaller components until the root events underlying the original source of the problem are reached. The graphical representation of the problem allows for quick interpretation. FTA can be used qualitatively and quantitatively. It is also very suitable for finding the causes of non-conformance, deviations and problems with the production process.
The process begins with drawing a tree of events leading to failure. The tree is a graphical representation of events leading to the problem under consideration. The events are related to each other by logical operators.
There are three types of events:
- Primary event – the original cause
- Intermediate event – an indirect event resulting from another event
- Conditioning event – an event that is not a failure, but the occurrence of which determines the occurrence of a failure
- Normal events – Expected events that are not failures
- Transfer Event – an event connecting different failure trees.
Events are related to each other by logic gates. There are five basic types of logic gates:
- And – all conditions must be met for an event to occur
- Or – at least one condition must be met for the event to occur
- Inhibit and – the occurrence of the conditional event inhibits the occurrence of the event
- Priority and – conditions must occur in a certain sequence for the event to occur
- Excluding or – The event will occur when only one condition occurs. The occurrence of more conditions does not trigger the event.
After the development of the tree comes the turn to calculations (although not necessarily, but about this later). In this analysis we can accurately determine the probability of failure or breakdown. Of course, as long as we have the appropriate quantitative input data.
The first thing you should know is that the calculations in the Fault tree analysis are related to the time unit. First, we need to assume the time interval over which we are considering our system. For example, if we are considering the probability of some deviation in the production process we can do it on the scale of one production cycle or, for example, on the scale of one year.
We begin the calculations in the opposite way to drawing the tree. First, we assign probabilities for basic events. We can do this, for example, on the basis of information about the frequency of occurrence of these events. In subsequent steps, we calculate the probabilities of intermediate events, until we obtain the final result. The way of calculating the probability depends on the gate through which the events are connected to each other.
As you can see, the more complex the tree and the more possible events, the more complicated the calculation. For simple trees excel should be able to handle it perfectly, for more complex ones you need dedicated software or online tools.
Application of Fault tree analysis
Fault tree analysis can be performed in two ways: qualitative and quantitative (calculating the probability of failure). Thanks to the graphical representation of the problem under consideration and the systematic approach, we can use the tool for many purposes such as:
Root cause analysis – identification of all events and conditions leading to an unexpected event
Risk assessment – calculation of probability of unexpected event, identification of critical elements
Safety assessment design – demonstrating compliance with requirements, identifying control sites, etc.
Fault tree analysis has a wide range of applications. I most often find myself use it as a tool to find the root cause of some event, whether already occurring or hypothetical. I have implemented this tool in my routine work of developing analytical methods. I make a tree early in the development to identify possible causes of measurement errors and address them appropriately during laboratory work. I repeat the error tree analysis after the method development is complete. It later serves as a ready tool to solve deviation problems during routine analysis. I recommend this application not only for analytical methods, for which the tree is relatively simple, but especially for production processes.