We’ve covered a short introduction to Condition Monitoring before – it’s great but has some shortcomings in that generally you need to be quite an expert in understanding the raw data to extract value for your predictive maintenance program.
Tradition doesn’t mean good
Effective condition monitoring is traditionally expensive due to its reliance on expert human interpretation, type ‘Condition Monitoring’ into Google Image Search and you’ll unfortunately get something that looks like this:
This might be fine if you’re only analysing one or two assets and know what to look for but increase that to many hundreds and it becomes a task ill-suited for a human. Although there are various condition monitoring packages available with different user-aids, the workflows all tend to look something like this:
It’s time-consuming, difficult work. The people employed to do it are typically skilled engineers – they don’t come cheap and unless they have deep experience with the assets being monitored, prognostics is unlikely.
The aerospace and defence industries have been pioneers of condition monitoring and prognostics, partly due to their insistence on safety and partly due to contracting methods which mean that the OEMs supply expensive assets ‘by the hour’ and so take on the associated maintenance risk. Many hundreds of assets can be managed manually if you throw enough people at the task – for manufacturing and other industries that simply isn’t economically feasible.
Breaking from tradition
Advances in in Machine Learning and cloud computing together with the emergence of masses of cheap sensor data from Industry 4.0 / the IoT, allow us to look at condition monitoring as a data rather than a resourcing issue – whilst also opening up opportunities to automate the pain away and bring in new technologies like prognostics.
We refer to ‘what is happening to the machine right now’ as automated condition monitoring & diagnostics. Taking an opposite approach to most condition monitoring packages, we don’t want the user to be sitting down interacting with the data (raw or features) – that job is better suited to a machine so why get a human in the way? At Senseye, we only want to bring the user in to look at data when there’s something of interest to look at and perform an action on (schedule maintenance of an asset). In a simplified way, the workflow for something like vibration data then looks a little like this:
This lightens the workload on the engineering team – allowing them to focus on improving availability instead of analysing data.
What this really means
The less you need to rely on manual expert input and analysis, the quicker, more accurate and more efficient your predictive maintenance program can be, crucially this translates to reduced downtime (in manufacturing up to a 50% improvement has been seen) with dramatically reduced operating costs and a longer asset lifespan.
Automatic condition monitoring & prognostics - PROGNOSYS
We’ve developed an easy to use, cloud-based condition monitoring, diagnostics and prognostics software tool to automatically forecast machine failure which works with any machine and any IoT solution. We’d love to help with your predictive maintenance plans and are offering a free Industry 4.0 & predictive maintenance consultation, get in touch and let's start improving things!