So you’ve made the decision that you need a tool to help you understand the performance of your machines and get a better idea of when they’re likely to fail. Your ultimate goal is to reduce unplanned downtime and increase those all important KPI’s – OEE, OTIF, etc. So what are the key features to look out for in such tools?
1 - FOCUS
The whole manufacturing analytics market is becoming quite a crowded place. It seems every analytics product has a solution for predictive maintenance. You need to be clear though on exactly what they’re offering. Most only provide the initial environments for you to create a solution so expect to pay for consultants or in-house developers to tailor it to create your custom solution.
This is not an inexpensive task (with significant recurring costs) but may well give you a product that aligns exactly with your own processes and ways of working. It’s a trade-off that needs to be carefully considered.
2 - SCOPE
Predictive Maintenance is a broad area with different solution types. Some products identify trends in machine usage or spares orders and help you anticipate future needs. Others will look for external influencers (such as humidity levels) to correlate against general machine failures to provide you with insights on external factors. A few will forecast failures for individual machines. You may want to have a combination of these or just one. It depends on your particular situation and the cost-benefit return that makes sense to you.
3 - TIME TO FAILURE
You need to be very clear on what machine forecasting you need and what the product can offer. Diagnostic and analytics products will generally only tell you when thresholds (such as temperature or vibration levels) have been exceeded. Of course, once this happens the time to failure is short if not imminent. Such short warning timescales may be insufficient considering your available maintenance slots and spare part lead times.
Machines and their components fail in complex and non-linear ways that are hard to predict. Very few machine health monitoring tools are able to genuinely forecast in this context – if they can, it’s at best a couple of days. Therefore, be sure to get evidence of longer term forecasts before you commit on the investment.
4 - FALSE PROMISES
Be aware that prognostics is not synonymous with predictive maintenance, predictive analytics or machine learning,it can be easy to get the two confused. Essentially condition monitoring evaluates the current state and prognostics determines how long the machine is able to continue working for you. We’ve explained this further here.
Make sure that the tool you’re evaluating isn’t just a condition monitoring tool, so it doesn’t just tell you when your machine is broken but when it’s going to break, so that you can plan and implement your predictive maintenance correctly.
5 - COST
Some prognostics tools rely on a heavily customised approach and extensive historic data. These take time and cost to develop and are likely only suitable for very high-value assets. Moreover, it will be unique per machine type so don’t expect it to work across your manufacturing facility. If you need a broader based solution, be sure to enquire about expanding the scope of the tool.
6 - DOMAIN EXPERIENCE
Operating traditional prognostic tools or the newer breed of self-service analytics products requires an awful lot of domain experience and understanding of the wider manufacturing context. While most diagnostics and prognostics companies can provide this confidence, be aware that many of the analytics solutions come from IT firms and “IT guys”. This has the benefit of getting the latest tech packaged in an attractive and ‘consumer-like’ way, but may lack some of the deeper domain principles.
The overriding consideration is to do your research and get evidence of efficacy. You’ll need to get underneath all the marketing hype and really understand the true capabilities. If you can, demand a free pilot period to evaluate the product.
7 - INVESTMENT
Some solutions are only compatible with specific hardware and may not work correctly if you already have some condition monitoring in place. Refitting hardware is a costly and unnecessary expense to be avoided. The prognostics tool you choose should be capable of interfacing with whatever hardware and factory historians you already use.
8 - COMPATIBILITY
Some application developers sell a whole ‘solution stack’ which means that you’re locked in to their technology and can’t change solutions should you find something that better fits your business as it grows. It’s important for the solution you adopt to be compatible with software systemsnotdeveloped by the developer of the prognostics tool that you are evaluating and so uses standard and easy to use technologies like RESTful web services.
You’ll also want to ensure that the developer has good working relationships with other industrial automation software providers to make sure that you don’t have to take on the time-consuming role of middle-man.
9 - EASE OF USE
The most advanced applications can be useless if those meant to benefit from it can’t access the power hidden away. Is the application you’re looking at going to be able to provide you the information you need to know within seconds of accessing it or are you going to have to navigate complex functions and have to do any manual interpretation of the condition monitoring or prognostics information it is giving you?
10 - IS IT CLOUD-BASED
Beyond meaning that there is nothing to install if you want to use the application and that it should be compatible with mobiles and tablets as well as standard desktop PCs, cloud computing offers the ability for an application to learn from thousands of machines at the same time, giving benefits to all users of that platform.
Many condition monitoring installations require manual installation and configuration which means you can forget about using them when you’re out of the office to keep an eye on things.