Our previous post concluded with the following statement: “Garbage in, garbage out” and highlighted that for predictive maintenance to be effective, you need refined, quality data. When data is “dirty” or fragmented you will not only have to spend a significant amount of time to turn it into useful information, but you will not be able to “prove” the desired operational impact for which your predictive maintenance strategy was established. The data component is often considered trivial – “just plug it in and it works”. Unfortunately, it is not that simple. Data is often either stored in various different places or is of poor quality or irrelevant to solving a specific problem. Businesses are therefore drowning in data that must be scrubbed to render it useable to the extent that it can be measured to support predictive maintenance decisions that are driven by accurate information. To achieve one single, complete view of IoT data requires accumulation from multiple sources and systems, which has put data quality squarely back in the spotlight.
However, before getting into how to achieve quality data, two further considerations bear mention. First, there are many different types of data: People movement, machine vibrations, temperature, current, voltage, power, speed, torque, depth, height, volume to name but a few. For all these different types of data, the first step is always clean it up and see what can be used. Only then can it be determined what data is needed (and usable) for accurate analysis. For example, if your goal is to measure temperature and vibration it might be required to add extra sensors. Secondly, the frequency of the data measurement should be kept in mind, that is, how often do we need to measure? Real-time, every minute, hourly, weekly, and so forth.
So how do organisations deliver an accurate, single view of data? We look at 3 steps to improve data quality:
Given these numbers, there has been an increased focus on reducing operational costs and asset downtime. One of the ways in which to achieve this is by using predictive maintenance. And, with Market Research Future predicting that the global predictive maintenance market is expected to grow to approximately $6.5 billion by 2022, it’s time to join the conversation on predictive maintenance and the benefits it boasts.
1. Eliminate data silos
Employing an analytics-driven approach to maintenance and incorporating data across different industrial control systems, sensors, and applications will see companies eliminate data silos. They can do so by securely automating the pushing of data to a central cloud-based platform that is flexible, scalable and secure.
2. Clean data properly
To maximise the accuracy of system data it is important that companies invest sufficient time to clean data. This means they have to ensure that the data is correct, consistent and useable by detecting any errors and correcting them to avoid a recurrence. The cleansed data then can be used to drive actionable insights.
3. Manage data governance
For companies to establish and enforce effective data governance processes and procedures, it is essential that they know where their data is, who has access to it and how is it being used.
In our next post, we will look at four reasons why predictive analytics is beneficial for growing your business.
If you enjoyed this post these topics might also interest you:
An introduction to predictive maintenance