GRINDING 4.0

How Machine Learning can boost your predictive analytics

FOOD & BEVERAGE

The history of the Customer is that of a business which, for over 120 years, has pursued a company vision based on passion for work, for the product and the land in which it operates.
Hermes Reply supported the customer in its "Digitalization path" towards Industry 4.0 paradigm, introducing a Quality System based on Machine Learning algorithms with the purpose of helping operators during grinding operations.

MANUFACTURING QUALITY SYSTEM

Hermes Reply developed a Cloud based web application powered by input data from grinder and tests results in order to identify potential anomalies in advance and keep high quality standards.

Specific Machine Learning algorithms provide an estimation of future grinding test results and suggest new setting parameters for the grinder.
Results are shown to operators via real time dashboards, comparing real data VS forecasts.
The Operators can directly send the setpoints to the grinder through the application.
A mobile application alerts operators in the event of problems relating to specific parameters and events.

TECHNOLOGIES: Machine Learning | Python | .Net Core | Angular | OPC UA

CHALLENGES

Grinding is one of the most delicate stages of production, often left to the experience and sensitivity of the operators: the application of Machine Learning algorithms in this area represents a real challenge due to the complex relations among environmental variables and machine parameters.

The solution implemented takes as input data the operating parameters of the grinder and the quality data resulting from the particle size analyzes.

Using the current data together with the collected history, Machine Learning and statistical algorithms (i.e.: Random Forest) are applied in order to make a forecast on the next granulometric quality results and on the possible values of the parameters of the grinder in order to maintain the quality values within the minimum and maximum thresholds foreseen for the mixture being produced.

The algorithms do a continuous retrain based on the new data received.

The operator has a web dashboard at his disposal in which he can view both historical and current data and forecasts for the main particle size variables in the form of graphical trends.

In this way it is possible to see in advance if the values are deviating outside the expected quality thresholds.
In this case he can access the grinder parameter setting page, which shows both the current real time values and the forecast of the possible settings, being able to decide to send these values to the PLC.

  • strip-0

    Hermes Reply is the Reply group company specialized in architectural and technological solutions, Application Maintenance services for Automotive & Manufacturing industry and Management Consulting services. Hermes Reply accompany the customers throughout the digital transformation process in order to support the adoption of digital enablers and achieve short-term and long term results, combining in-depth knowledge of production processes, industry 4.0 technologies, delivery skills and strategic vision.