Case study: Predictive control

A machine manufacturer specialized in the beverage industry was facing a problem of production quality fluctuation due to the different underlying conditions of each location. To get the problem under control, COSMO CONSULT was commissioned to develop a real-time industrial process control system based on a mathematical model. Today the company is able to apply the model quickly and flexibly to different production processes. This made it possible to achieve improved quality levels at significantly lower process costs.

• Analyzing the causes and effects that are affecting the production quality
• Controlling  production processes through an optimization model that minimizes quality degradation and machine downtime
• Transform the expertise of experienced employees into digital knowledge
• Minimizing the cost of complex production processes with system-based optimization

For a machine manufacturer specialized in cleaning and refilling returnable bottles, the quality of production depended heavily on the underlying conditions of the respective production site. In order to ensure the production quality, the machine controllers were repeatedly adjusted to the respective environment. The temperature and air pressure values in the respective production site played an important role in that regard. The basis for this was above all the professional knowledge of individual employees. The company commissioned COSMO CONSULT to develop a real-time process control system that can be applied to different production processes to automate and flexibly adapt the controllers.

In order to guarantee a certain quality level under changing conditions, the project team adapted a mathematical MPC model to the individual situation faced by the machine manufacturer. A major challenge in the implementation was the development of a holistic solution approach (closed loop), which corresponded to the operational requirements.

The model calculates an achievable production quality from the real-time data of the machine controller. If the predicted quality does not meet the specified quality level, the mathematical model enables the employees to change the controller values. The performance targets are automatically adjusted by the system on the basis of an optimized combination of all influencing variables. The performance target is to achieve a fast, cost-minimized correction of all controller values with the purpose of continuous quality optimization.

The holistic solution approach succeeded in achieving an improved quality level at significantly lower process costs. Today, low quality production has dropped and machine downtime is at its minimal level. Above all, the customers of the machine manufacturer are now enjoying higher quality and more benefits from the associated added values. In the future, predictive control is expected to play an important role in the range of services offered by the machine manufacturer.