Richard Schouten - 30 March 2020
There is a lot you can find on the topic Industry 4.0, digital twins and predictive analytics and maintenance. I have been involved for almost a decade in the development of a predictive analytics platform based on process and transactional data.
However, I still have not been able to do all the math. The reason why?
First, there is not sufficient data to predict specific failure mechanisms.
Second, data in storage is not designed for failure analysis.
The absence of sufficient data is party/mainly caused by a lack of sensors.
Let’s take an example from our Asset Health Center at Sitech Services in Geleen, the Netherlands. Let’s say we want to be able to predict future failure of a pump (see photo). As can be seen, there is not one sensor in sight. And based on experience I can assure you that this is common in the industry. Brownfield plants have sufficient sensors to operate the plant safely and meet product quality, and typically only sufficient sensors on expensive equipment to avoid a major breakdown. However, to build a real predictive model, both process and condition data are required.
A significant reason for the absence of sensor equipment is cost. A sensor in combination with a cabled connection, power supply, engineering, programming of PLC and/or DCS and connection to the historian, including an update of all drawings and diagrams costs around US$10k. Yet most of the equipment in a manufacturing plant does not even cost US$10k or at failure lead to a loss of US$10k. And for a proper digital twin, you need a lot of sensors, costing a lot of investment, so something else is required for Industry 4.0. There are options such as wireless Hart, which is still expensive and does not allow for many connections.
Another option is the use of low-cost sensors which use either LoRaWan or NB-IoT. Still, these have either too low connectivity speed or are not suitable to install in an explosive atmosphere as they are not ATEX-certified.
Another hurdle is getting the acceptance and approval of the model and prediction by the engineers and the technicians. But that is an entirely different story.
It is fair to state that developing a digital twin in a brown-field manufacturing plant is at this moment not economically feasible. And I do not wish to stop any of the pilots or developments everyone is working on, I only want to get the expectations right and everyone to understand it takes time and money to develop a proper solution for the industry.
This is why Sitech is working together with different companies to develop a solution, and we know we cannot do it all alone; more partners are required to make it a success. I also believe that all other start-ups and scale-ups have their limitations and at some point have to seek partners as well. Together they can disrupt the industry developing low-cost ATEX certified sensors, a highly secure, reliable and low-cost industrial ATEX certified network, and an open platform to store and forward the data, the PI in the cloud.