Five Steps From Raw Data to Industry 4 2018-07-02
Brainboxes pictures the journey from the machine level to Industry 4 insight and beyond
The aim of Industry 4 is to create a smart factory, combining the intelligence offered by IT systems with automation.
In the smart factory the physical machinery of the factory floor is connected to IT systems via the internet or cloud. Intelligence comes from insight and automation contributing to machine learning.
Whilst industry has always been a source of great innovation, there is also caution about introducing disruption to otherwise sound and profitable systems. There needs to be a strong, provable incentive to change and the relatively rapid drive to embrace Industry 4 is perhaps in no small part down to the tangible and often substantial financial benefits the smart factory presents.
Reports agree that industry will account for a large proportion of the things connected in the IoT. Most of these things already exist and there will be a significant transition period where industry looks to retrofit equipment of all ages. As production machinery often represents one of the largest capital investments for businesses, updating production lines is a substantial strategic decision. Changing a piece of equipment just to get an ethernet port isn’t likely to happen, so businesses must look for other ways to connect to the wealth of information trapped in their devices.
Remote I/O devices provide a way to make that IoT connection and take a step towards Industry 4. Brainboxes make a number of devices that can connect to just about anything and get information out to the network or cloud. Understanding how simple data capture can take us from machine to insight can demystify Industry 4 and provide ways to start tackling it.
Here is the way Brainboxes pictures the journey from the machine level to Industry 4 insight and beyond:
Machines produce signals just by being on. A basic example is a light that indicates when a machine has power. Inside a machine there will be signals that turn on fans in response to temperature, or change the colour of the Andon light if the conveyor stops. Outside the machine there are also signals produced by common sensors. Lots can be measured by a simple two wire sensor and this provides a very affordable way to add monitoring to a machine.
At this level these signals can be pretty rough, noise really, that doesn’t mean much by itself, such as high/low, on/off, 1/0. But, this is all the input needed for a remote I/O device. These signals can be taken from the machine, from a PLC, HMI or from stand-alone/add on sensors.
When we assign a value to signals they start to become useful to us. For example, if the signal to the fan is high (it is on), coupled with the conveyor running then the machine is ready to build. The machine is in a good state so the Andon light is green. If the light is green and a sensor detects an object then we can say a product has been built.
By recording the machine state over time we can start to determine how much of a production shift a machine was up and running for, or how long it took complete a cycle and make a product. This is the basic data needed to complete the OEE equation. This requires data to be stored in a database, which can either be on a sever, in the cloud or by the device itself in what is known as edge processing.
Essentially machine statistics over time; insight can be achieved by looking at patterns in shift data. Like correlating machine temperature against breakdowns or optimising the maintenance schedule.
This is where we reach a level of abstraction, a step away from the factory floor and the mechanical processes, where we add in the factors and nuances that are unique to our application. The insights and resulting changes can be surprising and extend beyond the production line.
The last level is to use intelligent software to learn machine patterns to implement predictive maintenance. This part requires some specialist skills, but luckily there are platforms out there already that aim to help with this; IBM Watson, Google’s Cloud Prediction API and TensorFlow, Microsoft Azure and Siemens’ Mindsphere. These all offer cloud based ways to tackle data analytics, whether its big data or just big to you.
It’s still early days for this type of software and it remains to be seen which platforms will become adopted and which will fall by the way. For industry there may well be greater adoption of software from trusted familiar brands, such as Siemens, and less take up of open source offerings or those associated with IT, like Google.
Whilst companies which have a history of understanding the demands and concerns of industry may have an initial advantage, they must also innovate to bring the best user experience. Really great software blends into the background of our lives, providing benefit, expediting, easing and integrating effortlessly. It should be no different with the systems we develop for Industry 4.