As a leading supplier of high-quality and customized heat exchangers, FUNKE has built up expertise over four decades, meeting customer requirements with competence and precision. Our close partnership with FUNKE enables us to accompany the entire process from calculation to completion of customized equipment. Through interdisciplinary cooperation, we ensure holistic optimization of heat exchangers, which is characterized by innovation, individuality and quality.
Quality is a fundamental principle for FUNKE and this is consistently realized through various quality management measures within the production line. A key element in this process is the control of the number of heat plates (11) in a heat exchanger. This traditionally manual process is both time consuming and error prone. In collaboration with FUNKE, we at digatus.ai have worked to optimize quality assurance while accelerating it through an AI-based solution.
We developed different solutions using both instance segmentation and object detection to select the best performing approach. Combined with state-of-the-art computer vision algorithms we are able to count the number of heat exchanger plates achieving an accuracy of 99%.
Since there was no dataset available, we developed a pipeline for synthetic data generation.
The utilization of synthetic data generation presents a valuable opportunity for expediting the development of AI systems without the need to collect data over extended periods of time. By operating within a virtual environment, we gain full control over crucial factors such as lighting conditions, material properties, object shapes, and their respective positions and rotations. This level of control enables us to fine-tune and optimize these factors to create high-quality synthetic datasets that accurately represent real-world scenarios.
