Monitoring software uses a range of methods to process production line data to provide users with the means to monitor and address process deviations. These methods are usually grounded in single-point measurements, where alarms are set off when a measurement falls below or goes beyond a setpoint value.
Some software also includes a predictive layer that often utilizes machine learning to seek the signals of an upcoming potential deviation - this is known as pattern recognition and is often applied e.g. in predictive maintenance. While this rightly so provides operators with the means to act on an occurring issue, the method ignores the intricacies of a production line and the impact that one measurement can have on multiple others. Countering this ignorance is the essence of harmony modelling.
Unpacking the Concept of Harmony Modelling
Harmony modelling is a unique and advanced monitoring technique that explains the current harmony in a plant by estimating thousands of tag values based on millions of interdependencies. You can imagine harmony as being the balanced state of your production.
At its core, harmony modelling involves constructing various regression models in a meticulous series of events using automated tools. A harmony model essentially functions as a regression model, estimating the value of a dependent variable (a tag) based on explanatory variables (multiple other tags). When the error in the model remains small, it signifies that the tags are in harmony with each other, indicating a smooth operation. However, an increase in error suggests issues at the plant level, signalling potential disruptions.
Practical Applications of Harmony Modelling
To illustrate harmony modelling's practical utility in monitoring software, consider a simplified scenario involving the outflow through a filter. Let's say we have dirty water flowing through the filter. Water flows in and water flows out. The measurements available to us are the inflow, the pressure, and the outflow of the filter. As inflow increases so does the pressure and outflow.
In a scenario where the filter becomes clogged, software using regular modelling methods would be able to detect the problem at a point where water can no longer flow into the filter or when pressure is higher than the pre-defined maximum value, sounding the alarm that it is time for a clean-up.
In contrast, a harmony model is able to identify the problem at a stage when the clog is building up. This is because although the pressure in the filter in this scenario is not beyond its limits (as this normally just means that we are pumping in more water), harmony modelling takes into consideration the relationship between the increased pressure and other measurement values such as the inflow and outflow - highlighting that something is off in the relationship between the measurements.
While the harmony model itself does not offer a direct solution, it effectively highlights the issue with the filter, enabling operators to take corrective action promptly. At times, we might be talking about differences in minutes or small percentage points, but at the scale of a production line, this can lead to substantial savings.
By leveraging harmony models, operators gain the ability to detect sudden changes in process behaviour and understand which group of measurements is directly impacted by the identified disharmony. Quite often, the problems are more complex than a clogged water filter. This deeper insight is invaluable, allowing users to comprehensively understand phenomena and associated risks, thereby facilitating more effective corrective and preventive actions.
Conclusion
Harmony modelling represents a transformative approach to process optimization in today's data-intensive industrial landscape. By leveraging AI and advanced regression modelling techniques, plant operators can proactively address emerging issues, improve process efficiency, and reduce downtime.
The power of harmony modelling lies not only in its ability to detect deviations but also in its capacity to provide invaluable insights for data-driven decision-making, ultimately resulting in improved operational standards and resource optimization.
As industries continue to embrace this paradigm shift, harmony modelling emerges as a pivotal tool in the toolbox of forward-thinking plant managers and higher management seeking to navigate the complex challenges of modern manufacturing. For a cherry on top, harmony modelling makes operators’ lives less stressful – and who could deny that that’s something we would all need in life.
Harmony modelling the core behind Harmony - one of our software modules for improved production efficiency.
