Enterprise optimisation based on data and artificial intelligence

This topic was brought up in a presentation by technology manager Peter Bosmans and Pedro Durlinger from Sitech Services, who has ten years of experience in maintaining 30+ factories and over 5 years of experience with Industry 4.0.

Peter Bossmans introduces the main daily tasks of their clients (see Fig. 1) and at the same time talks about the benefits that come from artificial intelligence, such as:

Reduction of maintenance costs by (15-50)%;

Decrease in operating costs by (10-20)%;

Increase productivity and OEE up to 15%.

Figure 1. Daily Tasks for Sitech Services Customers

It also describes in some detail the methodology (see Fig. 2), which consists of:

Communication of information, events and telemetry;

Data cleaning;

Event cleaning;

Dispersion-based cleaning;

Features and interactions;

Simulation, larger grid search;

Checking the model;

Model selection;

Deploy Models: Azure Function Containers.

Figure 2. UltraGrid methodology

As examples, Peter Bossmans shows two maintenance cases:

Predictive maintenance of an electric motor, where a problem is detected in one of the motor windings before the equipment fails.

Figure 3. Preventive maintenance based on the electric motor

Prescriptive maintenance of the heat exchanger (TO), in which, after cleaning, the heat exchanger shows a change in its behavior. It is by the change in the ratio between pressure and flow that it can be determined that the TO is faulty.

Figure 4. Preventive maintenance based on ТО maintenance

Peter Bossmans noticed that before the traditional cleaning with the TO, the flow decreased and the pressure increased. At the same time, after cleaning, we got a larger flow and a smaller pressure difference for TO.

Figure 5. Array of data on flow and pressure difference in the ТО

If the pressure / flow is increased over the HT for a certain period (even above the projected limits), then the operation of the HT becomes more efficient, since with a smaller pressure difference, a larger flow is possible.

Figure 6. Array of data on flow and pressure difference in ТО

After analyzing the data, Peter Bossmans hypothesized that if you increase the pressure and flow through the TO, will it perform better? After that, an experiment was carried out, where once the maintenance was completely purged (at maximum pressure, maximum flow rate within the design limits). What happened next was that this "in-process" cleaning had the same effect as traditional cleaning: more flow and less pressure difference.

Figure 7. Array of data on maintenance in the period from 06.2019 to 06.2021

Thus, it can be concluded that the hypothesis based precisely on the analysis of the data has been confirmed, and the traditional cleaning of the TO is no longer required. The result is an annual savings of € 10,000 in heat transfer costs.

In confirmation, Peter Bossmans cites a business case where he shows the following values:

Average costs for maintenance restoration, including man-hour (for maintenance, every 3 years): 20,000 euros;

Production losses for materials (for maintenance over 3 years): 18,000 euros;

Production losses per man-hours (for maintenance over 3 years): 3,500 euros.

The total average cost of cleaning is around € 41,500 for a 3-year maintenance.

Commentary by Maxim Tyutyunnik

Peter Bosmans from Sitech Services raises current customer problems that are typical for domestic enterprises - downtime, raw data, the loss of qualified personnel from the technical service due to a large number of non-creative work. In the AI use cases described, analyzing the motor current and extending the wash interval of the heat exchanger, the amount of processed data is impressive. Increasing flow in any plate heat exchanger does reduce deposits, but raises questions about the correct heat exchanger for a particular application.

In any case, the application of AI in industry, for processing data sets, has already reached the level of a proven and affordable technology.