When a wind turbine or inverter starts to fail, management wants answers. Are they failing because of manufacturing problems? Is it due to bad operations and maintenance practices? Or is the design to blame? After doing root cause analysis the next stage of evolution poses the question what is the likely number of failures this year? What should be our budget for gearbox failures or major inverter repairs?
One of the most widely regarded methods for accurately predicting operational life and failure rate is statistical analysis of a component’s or device’s failure data. Though there are many statistical distributions that could be used, including the exponential and lognormal, the Weibull distribution is particularly useful because it can characterize a wide range of data trends, including increasing, constant, and decreasing failure rates, a task its counterparts cannot handle.
Weibull analysis is a common methodology to forecast machine health and reliability. A Weibull distribution is similar to a Normal/Gaussian or Poisson distribution: it is a probability distribution curve to describe the likelihood of an event. “Grading on a curve” falls into this category.
Many management decisions involving life-cycle costs and maintenance can be made more confidently from reliability estimates generated by Weibull analysis. For example, Weibull analysis can reveal the point at which a specific percentage of a population (such as a production run) will have failed; a valuable parameter for estimating when specific items should be serviced or replaced. Additionally, this analysis helps determine warranty periods that prevent excessive replacement costs as well as customer dissatisfaction.
Weibull analysis can be particularly helpful in diagnosing the root cause of specific design failures, such as unanticipated or premature failures. Anomalies in Weibull plots are highlighted when items uncharacteristically fail compared to the rest of the population. Engineers can then look for unusual circumstances that will help uncover the cause of these failures, which could include a bad production run, poor maintenance practices, or unique operating conditions, even when the design is good.
As with most analytical methods, the accuracy of a Weibull analysis depends on the quality of the data. For valid Weibull analysis, and to interpret the results, there are several requirements for the data:
Engineers often shy away from Weibull analysis because they believe it is too complex and esoteric. Although it’s true that an understanding of statistics is helpful, engineers can reap the benefits of a Weibull analysis without a strong statistical background.
Now that we know that a Weibull distribution is useful, lets say we draw up the same for key components of the solar plant and wind farm. The analysis helps us for the following reasons:
What next? Does the curve provide a good idea on the remaining useful life of the component under study? Let’s look at one analysis and see what we figure out. For example, if we know that 2 gearboxes have a 92% chance of failing in 2016, from a farm having 200 turbines that has a lot of value for budget development. But, do we know which 2 gearboxes and when?
When the ultimate goal is preventing failures, not just forecasting them and reacting afterwards, probability distribution does not help us identify and mitigate issues. To impact the cost of operations and maintenance, we will need more information. Algo Engines’ machine learning algorithms assist with Weibull analysis of probabilities of failure, but when combined with our diverse array of notifications we can help you pinpoint issues, and much more.