June 2019

60 n GAMBICA SUPPLEMENT June 2019 www.drivesncontrols.com PRODUCTIVITY FOR THE FUTURE F orward-thinking machine vendors are constantly seeking ways to improve not just the short-term productivity of their machines, but also their lifetime productivity, which takes into account unplanned downtime and factors such as the production of out-of- specification parts that have to be reworked or scrapped. But how can lifetime productivity be maximised? The design of the machine and the materials used in its manufacture inevitably play an important role in determining its reliability and performance, but improvements in these areas are increasingly difficult to make. There is, however, an innovative and very effective approach to maximising lifetime productivity of machines. This is to adopt industrial analytics. In a nutshell, industrial analytics entails collecting data from a machine, analysing it, and making use of it. These days, collecting data isn’t much of a problem – in fact, there’s often too much of it! But raw data isn’t useful in itself, because the valuable information it contains is concealed by a mass of routine and unremarkable results. This is where analytics comes in – sifting the wheat from the chaff, and alerting machine users to changes that may need their attention. And the last step – making use of the data – means responding to these alerts either automatically or through manual intervention. In short, industrial analytics makes possible the development of innovative data-driven business models. Let’s have a look at what this means in practice. An important element of industrial analytics is the detection of anomalies, which is done by using machine-learning algorithms to look at key parameters relating to the machine and the products it is making. This approach is far more effective than conventional rule-based analysis because the algorithms learn what is “normal” for the machine and the environment in which it is operating, rather than using pre-determined rules that can only reflect what is normal for a typical machine in a typical environment. Advanced machine-learning algorithms also use multi-dimensional data – that is, they take into account data from multiple sensors when making a decision about whether or not a particular data point is anomalous. Overall, the result is that these algorithms reliably detect anomalies that a rule-based approach would miss, and they detect the anomalies sooner. Well-implemented algorithms not only detect anomalies, but also facilitate their classification. This allows the root cause to be determined quickly and easily, as well as allowing informed decisions to be made about the action that needs to be taken, which could range from simply keeping an eye on the situation, through scheduling maintenance during the next shutdown period, to stopping the machine immediately for maintenance to avoid the material and energy wastage associated with producing out-of-specification parts. Industrial analytics also makes it possible to implement predictive maintenance. This differs from reactive maintenance, where machines are maintained only when a problem occurs, and also from preventative maintenance, where machines are maintained on a regular schedule whether they need it or not. Predictive maintenance is data-driven and uses information about the past and current performance of the machine to make accurate predictions about its future condition. This means that reliable assessments can be made about what needs maintaining and how urgently the work is needed, which has three benefits. The first is that the risk of unexpected breakdowns is greatly reduced, the second is that maintenance can often be planned to take place in scheduled machine downtime, and the third is that unnecessary maintenance “just in case” is eliminated. All of these benefits ultimately translate into cost savings, higher productivity and increased profitability. A further important aspect of industrial analytics is that its algorithms can work with machine data to generate invaluable guidance on ways in which energy savings can be achieved. For example, does a particular cooling fan really need to run continuously, or is it only needed at certain stages in the machine’s cycle, or perhaps only when the ambient temperature is higher than usual? With energy costs seemingly set to continue to increase indefinitely, this facet of industrial analytics is a definite benefit for machine end-users. And it is not only the end-users of machines who stand to benefit from industrial analytics – machine vendors also have much to gain. Predictive The business models of larger OEM machine manufacturers are changing. Machine-builders are moving away from the traditional approach of selling machines and moving more toward selling production capacity, says Keith Atkinson of Weidmüller UK. The successful machine vendors of the future will be those that recognise and respond to this trend. Keith Atkinson