Efficient machine maintenance using artificial intelligence in Beneš a Lát, s.r.o.

Initial situation – the company needs to minimize machine downtime due to maintenance and the occurrence of unexpected failures due to the failure of a machine component in order to maximize the use of technology.

The standard approach in most companies today is to deal with preventive or reactive maintenance, i.e. repair occurs at the moment the manufacturer prescribes and worse only at the moment the equipment breaks down. A better solution in this situation is to perform maintenance preventively. In comparison to reactive maintenance, this minimizes machine downtime to only the duration of the repair. There is, however, a potential disadvantage – if the service intervals are set incorrectly (too short), parts are replaced with a high residual life, which increases service costs and causes more frequent machine downtime.


The solution is to fit the monitored equipment with sensor systems (one or more sensors – temperature, pressure, vibration, EE consumption, speed, flow rate…etc. depending on the specific equipment and system complexity), which will enable data collection and communication about the status of the equipment and current production with a link to ERP and MES systems, if the company is equipped with them, or at least to technological parameters and system parameter settings with a tolerance of “OK” status.

We then monitor the state of the machine (its parameters and production) over time and compare the collected data with the calibration curves of individual parameters – the curve is captured at the moment when the machine is in perfect condition (ideally after inspection, overhaul, medium or overhaul) and at the moment when the product produced on it is OK – i.e. it corresponds to the requirements of the customer or drawing documentation. When using AI solutions, the reference comparison method can also be used.

At the same time as the parameters, we monitor the errors which occur during the operation of the machine – service failures, the need for retooling, etc. We then monitor the correlations of the individual parameters in the period before the failure and look for a dependency that will help us next time to predict the deterioration of the system before the failure occurs… this way the remedy is faster, cheaper and can be planned in time for the machine downtime or perhaps the time when it is adjusted to another production.

If the company uses APS, the information can be used to optimize the production plan with links to upstream and downstream machines.If we already have information about the need for repair, we can work with it to assign work to a specific maintenance component, to escalate using mobile clients, and much more.

Watch the case study video: https://www.youtube.com/watch?v=OsNaH9gUQ5E

What Beneš a Lát gained by working with the Brain4Industry consortium<

The outcome for the company is a working model using AI solutions that allows them to process huge amounts of data, find correlations and move from a reactive and preventive maintenance format to a predictive one. Once a company has a model, it is very easy to gradually add similar devices. If other parametric sets need to be monitored, existing parts and logical units can be used and the methodology for processing and using data already exists. Thus, each additional machine is already much easier and less time consuming to implement.

And if a company plans to use Digital Twins in its practice, this data is the basic input for monitoring system behaviour and addressing feedback loops.

  • Measurable outputs include a 5% increase in the lifetime of replaced parts.
  • Minimized occurrence of unplanned downtime with a 15% downtime savings.
  • When machine anomalies are detected, we also expect to increase product quality, thus reducing scrap and remanufacturing costs – here we are currently collecting data to confirm it.
  • In BaL, also an impact on the quality of the entire hydraulics distribution and system status has been noted, which affects 30 connected machines, so the savings are multiple.

The estimated return on investment of CZK 1.2m is 3 years.

AI allows for non-correlations between complex data that are beyond the power of humans to find. By using AI solutions and advanced sensors, operations that were previously the prerogative of humans alone, such as visual inspection operations, can be automated with consistent results, moving from standard 95% inspection success rates to 99% or more.

What companies can gain from such cooperation:

  • Tracking machines and equipment with measurable variables will enable unambiguous monitoring of the performance of individual machines and entire systems…., i.e. OEE evaluation.
  • Introduction of predictive maintenance to save costs in spare parts, downtime during breakdowns and “early” service.
  • Information for energy optimization solutions…, i.e. which machines I can controllably switch off or on according to their performance.
  • MES systems and work schedules amendment and prioritizing maintenance.
  • Creating a digital twin of the operation with feedback control.

The case study was prepared by:

Milan Chlada

AI Specialist

View profile