Author: Eva Černčič
The innovation company IRNAS, providing the smart IoT solutions for industrial applications and the cutting-edge technology manufacturer of overvoltage protective and EMI suppression components, Bourns, did join forces on the NexGenHVEC project to accelerate development and advance the production line by implementing data collection and machine learning technologies.
WEBINAR: Hands-on experience with LR1110 LoRaWAN transceiver with GNSS and WiFi functionality
The main aim of the NexGenHVEC project is to advance the industry sector by utilizing cutting edge technology and modern research capabilities. The research goal is to improve thermal stability of hybrid electronic systems and components (varistors and capacitors) through the development of advanced materials and improved manufacturing practices. To maximize and accelerate improvement, implementation of smart sensor system, data collection and analysis, together with prediction models and machine learning algorithms for forecasting varistor performance, was set to be a vital component of the project.
Knowledge and usage of production parameters and environmental data can be of crucial importance for every production line or industrial plant, as it can help maximize production output and efficiency, without additional costs or resources. Moreover, when used and analysed correctly, collected data can unravel hidden patterns and reasons for lower quality, faulty produce and sub-optimal performance that would otherwise remain undetected due to complex interaction of many factors. Still, the main obstacle for successful implementation of data collection and machine learning in the industry sector is the large amount of resulting data, which without popper analysis and understanding, does not yield desired results.
As an important part of the NexGenHVEC project is research and utilization of the machine learning algorithms and collected data analysis for improved performance of produced electronic components, IRNAS and Bourns Ltd. collaboratively approached the task step by step, to avoid the trap of collecting too much unnecessary data that could blur the important patterns and correlations. The first step was to identify all parameters that could affect the final product and then decide on critical variables that will be observed and included in the learning system. The production line was closely inspected, and a list of possible observables was established. Due to the advanced, multi-step production procedure, a number of parameters proved to be too extensive for simultaneous implementation and hence bottom-up design approach of data collection and analysis was chosen, where the focus was first set on individual groups of variables that could later give rise to a more complex system.
Initial examination indicated that environmental parameters play an important part in the final output, thus it was identified that the environmental monitoring system could significantly benefit the analysis and its results, as every produced component passes through multiple microenvironments, with vastly different environmental conditions.
To verify the assumption, a system of LoRaWAN Senstick smart sensors, developed in partnership with Sensedge, supporting temperature, relative humidity, air pressure and other environmental parameters, was strategically placed through the production plant, to gain an insight. Senstick sensor units are based on LoRaWAN technology, which enables autonomous data stream on long distances (up to 10 km) and low power consumption. They are applicable for indoor and outdoor usage and are as such very robust and low maintenance, an important but often neglected factor in large-scale implementations. Sensors can be integrated in any cloud platform and a data visualization system for a user-friendly experience and direct insight in real-time data. Thus, along with data collection for research analysis, a visualization platform was set up, to offer direct insight and real-time data usage, to be used even during the development stage.
At the same time, a cloud data collection system was established to support initial data analysis, to guide the prediction model development and testing. Data on selected production line variables and final components performance results were added to the system, allowing for exploratory analysis and model selection. Correlation analysis suggested a relationship between various variables and helped in formation of learning models that were tested on the test dataset. The research and development procedure indicated areas where a more efficient data collection system needs to be implemented to achieve optimal performance.
Based on the collected information and gained knowledge the next step in the project is an on-site implementation of an upgraded sensor system, together with real-time implementation of prediction models and smart analysis algorithms that will, alongside real-time data insight, offer valuable understanding of the effect the production methodologies, environmental conditions and material properties have on the final product. To achieve maximal, on-site impact, a direct implementation of machine learning on the embedded system will be explored, utilising Edge Impulse platform for efficiency and straightforward implementation.
While the main aim is to accelerate the development of the new generation of electronic components, with superb thermal performance, the effects of the project are expected to be even more far-reaching and to have a beneficial impact on the whole production line, improving quality and efficiency. Implemented advancements in data monitoring and analysis are in-line with Industry 4.0 agenda, to digitalize the production line, create added value, and move one step closer to a full smart factory implementation.
The investment is co-funded by the Republic of Slovenia and European Union, European Regional Development Fund in the total amount of 1.682.494,00 €.
Advantage area and sub-area: Mobility – Systems and components for safety and comfort: Active-passive structural components.