Hardware development is a complex task. What applies to established companies such as Apple applies even more to start-ups. When a start-up starts this adventure with little previous experience, a small team and a minimal budget, the team needs good nerves. A hardware development process can lead directly to disaster. Why did we dare to do it anyway?
The oee.ai business model requires data to be collected from manufacturing plants in factories. Obtaining this data involves intensive physical and argumentative hurdles with the responsible interlocutors, which must be overcome anew with every customer. But a honey pot attracts: The market for manufacturing companies using systems is huge. The language of the data is international. We are thus faced with a multi-billion market that remains closed until someone masters this challenge — or finds a path without hurdles.
6 hurdles on the way to collecting OEE data
What hurdles are we talking about? Some of the systems in companies are so old that they do not collect any data themselves. We are in a market for capital goods. 10 years is considered the lower useful life in a majority of industries. It is not uncommon for systems to be used for 30 or more years. And no one disposes of a functional system just to buy a similar system with more sensors and a controller. The topic of “Retrofit” will accompany the 4th Industrial Revolution for a long time to come.
When data is then collected in the system, it is not necessarily at the right point in the process or completely for our use case. For example, we need the interplay of quantities and reasons for faults, which are assigned by the employee via a display from a fault index catalog. No system that has not been specially designed for this purpose has such functionality.
If there is a modern control system, it is very reluctant to change it for two reasons. First, medium-sized companies in particular often lack employees with the appropriate know-how. Programmer for programmable logic controllers (PLCs) are scarce. Even many mechanical engineering OEMs rely on freelancers. Anyone who has employees with this expertise within their own ranks can count themselves lucky — but that is the exception rather than the rule.
The second reason is a risk during the procedure — “never change a running system.” These are production plants that are planned for the long term. No production manager wants to give their COO the explanation “we wanted to try something out” for an unplanned downtime.
However, this is not yet the end of the hurdles: If the systems are modern enough, the plant manufacturer typically has access to the data paid for with a separate license key. For example, five-digit amounts are often required in order to be able to access your own data. In this case, you usually need additional software, such as an OPC UA server, which ultimately outputs the necessary format of the data — combined with additional costs and corresponding IT know-how, of course.
And then there is the diffuse fear of connecting the systems to the Internet. The Stuxnet computer worm from 2010 has anchored itself in the collective memory. The malicious program was developed specifically to attack a monitoring and control system (SCADA system) from the manufacturer Siemens — the Simatic S7. Trust in one's own IT department or (non-IT) management's understanding of how to operate firewalls is limited. As a result, fear prevails and valuable plant data is neither collected nor evaluated.
OEE data collection without system intervention
An IoT gateway, which collects data without system intervention and without a return channel to the system, must therefore be developed for our business model. This makes us basically reluctant hardware developers. But the best artificial intelligence and machine learning algorithms are useless if they are cut off from the data stream. We now have this IoT gateway. Developed in Aachen, produced in Shenzhen. Or as Apple would say: Designed in NRW, assembled in China.
By the way, we are convinced that all of the above arguments will be obsolete in a few years' time. In the future, it will be much easier, if not standard, to be able to access system data at a central location and in a uniform format. But we want to access the data now and until the future becomes reality, we will provide our customers with our own, minimally invasive IoT gateway — as a hardware-as-a-service for rent.