Grinding process optimization with machine sound monitoring solution
Neuron soundware technology, based on evaluating the health of the machine according to its sounds with AI and machine learning algorithms, is applicable not only to the prevention of production downtime but also in process control within production processes such as verification of the quality of the material in the grinding process. Let us dive into a recent use case.
The challenge: optimize the grinding process and increase its efficiency
While condition monitoring is a natural area for Neuron soundware technology leverage, other areas may easily benefit as well. In this case, the grinding of bauxite in an industrial mill has been the subject of interest.
The customer, the major chemical company, needs to deliver identical consistency of grinded bauxite from each manufactured batch. To achieve this, the customer has deployed an 18 hours grinding cycle. Why 18 hours? The customer’s experience has been that this is needed time, regardless of the quality and incoming material consistency, to achieve the same manufacturing output quality. Therefore, the length of the grinding cycle would secure that the outcome is always fulfilling the minimum requirements. While the mill is closed during the cycle, the customer has had no practical chance to verify on the spot if the material is already ready in a shorter period, let’s say 16 hours as a real-life example.
The situation: the predetermined grinding cycle length may not be optimal
By definition, the fixed grinding cycle in this case has meant playing safe. However, has it been playing effectively as well? Not really. The real-life tests have proved and indicated that while 18 hours had always meant achieving the desired quality, very often the material had been ready in 14-17 hours, which had resulted in process inefficiency of 5% to 22%.
The solution: IoT process monitoring solution
Neuron soundware deployed an IoT sensor-equipped nGuard solution to listen to the mill and process the data. The non-invasive sensors are installed on mill components (bearings, drum). The gathered data has been evaluated by a customized artificial intelligence model to deliver a strong 95%+ accuracy of defining each part of the process cycle including the final product readiness.
While in the noisy production environment there are many variables complicating the sound measurement and evaluation, the Neuron soundware team, repeatedly in the global top 5 AI anomaly detection competition – DCASE, has achieved together with the customer an outstanding result.
Conclusion: Neuron soundware process monitoring can increase production efficiency by 11%
Switching from a defined cycle to a ready-product-driven cycle will result in an average saving of 1- 2 hours per production batch and can increase production efficiency by 11% by better organizing the product flow. Besides that, the customer will save energy costs, reduce service intervals, and extend the machine lifetime.
Neuron soundware solution is the only solution on the market that can detect the quality of the material during the grinding process which is beyond the reach of standard methods. Material sound interpretation using AI algorithm proved itself as the right tool for process monitoring where the end product quality is the main criteria.