A decorative picture containing text Predictive Analytics, screenshot, graphics, light

Can humans really deal with knowing the future?

Trust. It’s all about trust. Effective Predictive Maintenance will not tell you with 100% certainty what will happen in the future, but it can provide a probabilistic or risk-based forecast of what could happen based on current knowledge and previous experience. But how do we develop the confidence to trust what a Predictive Maintenance solution is telling us?

There are many publications, papers and articles that explore the Human interaction with Artificial Intelligence, Smart Technology and Predictive Analytics, and how ‘Trust’ forms a key component of successful relationships between them. For the purposes of this article, we will explore very high-level implications of exploiting Predictive Maintenance data to the point where we can trust what the outcomes might be in the future if we take action on what that data is trying to tell us.

Knowledge of the future can allow us to change the future by taking an alternate course of action. This may be considered by some as fantasy when it comes to predicting ‘our future’, but there can be a degree of reality if we have effective condition monitoring and associated prognostics in a Predictive Maintenance system that allows the failure propagation of an asset to be estimated, based on that current knowledge and previous experience. Essentially, by learning about a potential failure earlier, we can take action to avoid potentially more severe consequences that might occur if such action were not taken.

The real measure of success is ‘does the Predictive Maintenance solution produce an effective output most of the time, and as a result, are operational savings being made’? If the answer is yes, then a level of trust will develop, and the operational culture will adapt to rely on Predictive Maintenance as a critical source of actionable information. By inference the Predictive Maintenance system will then be enabling the humans to predict and influence the future in terms of maintenance activities and operational efficiency.

So, the first key ingredient to ‘trusting Predictive Maintenance’ is high quality condition monitoring data (current knowledge) supported by relevant maintenance information (including previous experience). We described in our earlier blog, “The Climb to Prescriptive Maintenance”, the importance of reliable and relevant condition monitoring data.

Knowledge associated with the current and historic health (or condition) of the asset is essential. Collecting vast amounts of ‘Big Data’ in itself is not going to bring value. Value can only be achieved if the data contains useful information; Relevant Data.

To build trust in this data it is essential that the following steps are included as explained in our “Predictive Maintenance Guide“:

Data Reasonableness/quality checks.Remember‘rubbish in = rubbish out’.
Reliable condition indicators must be supplemented by the contextual information of situational awareness (operational context).
Maintenance informationmust be available to createthe essential feedback loop.

Knowing when to act is key, and that can only be determined with a level of trust in the system doing the thinking for us. This takes us to the second key ingredient – ‘confidence to act’.

A lot of the time the Predictive Maintenance notification will be sent to busy people, with responsibility for lots of high value assets, and without the time to dive deep into data analytics each time a maintenance decision needs to be made. They need trustworthy, accurate and timely information that they can act on with confidence. The whole point of Predictive Maintenance is to make the overall operation more efficient.

The right Predictive Maintenance solution will present the user with meaningful information, in the right place and at the right time, with much of the workload being taken care of by intelligent data management tools.

SmartVision™ is a flexible solution to create actionable information, which enables the operator to exploit the data collected from the rolling stock and tracks to optimise maintenance procedures. Data pre-processing and condition indicators are incorporated to provide high quality, relevant data. These condition indicators give trustworthy, accurate and timely information that you can act on with confidence.

Read more news