But how are these decisions made by the system? The basis for this is machine learning. With the collected historical data, a model is optimized by a training algorithm. The model learns from the data the connections between the processes and their sequence, the configuration of the system and the recorded KPIs. The model must also define which KPIs are positive, neutral or negative. On the basis of this information, the model can then decide which system configuration delivers positive key figures – good results – and which does not.
The configuration can then be adapted accordingly. A prerequisite for the direct adaptation of the configuration is that the system can provide live data to the model or training algorithm. Only with live data can current events in the processes be analyzed and the configuration directly adapted. Subsequent adaptation of the system configuration is also possible if no live data is available.
Smart Process Control – An example from manufacturing
But how can all this be implemented in practice?
Smart Process Control can be used in manufacturing, for example. There, the analysis model is trained with historical data from the involved IT systems. For example, with the confirmed process data from SAP Production Planning or an MES system. Sensor data can also be used. With the help of process mining, the system reconstructs the process flows from the data provided and calculates various key figures. The key figures can now be used to check whether the products currently being produced meet the quality requirements or whether the processes meet the specified criteria.
It could also be deduced whether the machines are overloaded. Machine learning algorithms are used to find patterns and correlations in the sensor data. Based on these findings, decisions can be made directly and automatically by the system. Such decisions can be, for example, changing the production sequence or adapting the machine configuration. The press pressure or the furnace temperature can, therefore, be adjusted if the material has a slightly different composition as usual.
What possibilities does Smart Process Control offer?
The benefit is important for new technologies in practice. If a technology does not benefit the user, it will not be able to make an impact.
A major advantage of Smart Process Control is that the processes and parameters can be adapted directly during execution. This means that fewer resources are wasted than in the case of errors that are not detected until after the fact. Furthermore, the continuous monitoring and adaptation of the processes can guarantee an optimal result.
If Smart Process Control is combined with Predictive Maintenance, an optimal maintenance window for the machines can be found. This window is derived from the recorded and evaluated utilization of the machine and the forecast values determined by prediction algorithms. Smart Process Control also increases transparency throughout the whole process. Products can be tracked during production. This makes it possible to forecast a delivery date so that service requests can be reduced.
Are there obstacles to the introduction of Smart Process Control?
In order to apply Smart Process Control, basic historical data must be available. If this data is not available, the model cannot be trained and the implementation of Smart Process Control will be very difficult. However, basic data is present in every manufacturing company. Such data can be manufacturing feedback or reporting.
Other possible risks may arise if the analysis model has been trained with unchecked or incorrect historical data. If this is the case, decisions made by the model or algorithm can be wrong and jeopardize the entire process. Incorrect system decisions can be bypassed by the system initially suggesting only possible courses of action. In this case, we speak of a recommender system. The final decision is made by a domain expert. If the system has developed sufficient stability, the recommender system can switch to automated decisions.
The live data must be prepared to the required extent so that it can be used by the Smart Process Control System. If a sensor fails, important data required for decision-making is missing. Therefore, it is important that the system informs the user about sensor failures and missing data flows and that the automated control adjustment is restricted for the time period. The fact that a responsible person is questioned beforehand about the planned decision in the event of uncertainties or failures can be such a restriction.
Since the demand for (live) data is high in order to make precise decisions, the implementation and interface effort can also be high. This is the case, for example, when many legacy systems are in use whose data is not easily accessible. To keep this effort low, technologies such as Enterprise Architecture Integration (EAI) or platforms such as IoT platforms can be used. Here, only one interface needs to be addressed by the Smart Process Control System.
Smart Process Control gives the user more control over the processes in their company. However, some prerequisites, such as historical data and live data, must be available in order to use Smart Process Control effectively.