The above edge-processed thermal data represents the spatial cooling effectiveness of an ink delivery system critical to process control. It is the result of building, testing, refining, and applying layers of algorithms in a manner to provide a digital representation of the physical heat transfer measurements needed.
The below edge-processed data shows in a human readable form the spatial heating effectiveness of the fusing power delivery system critical to process control. This is also formulated using layered algorithms. The processed numerical results are transfered back to a centralized database for operations to monitor the fleet of customer printers.
Building digital transformation of physical systems takes a specialized way of thinking about problem solving, namely how to utilize computational tools in a way that is efficient for the system. However, there is another opportunity for improvement, which is applying knowledge of physical system characteristics in a manner that can further simplify digital algorithms.
In my work creating algorithms I work from both sides of the problem to speed up delivery of a robust and efficient solution, whether that be 2D image/Thermal processing or one-dimensional datasets. Below is an example using a conditional series of image processing algorithms used to quantify the effectiveness of powder dose for system monitoring
Below is an example using an interval application of linear regression to detect when a powder containing vessel runs out of powder. The algorithm detects the presence of this behavior by finding an intersection point of two lines that has a slope change exceeding a threshold. The intersection pointer at this conditions signifies the point in time (layer) in which the powder containing vessel was emptied. When this occurs it is reported as a red flag to operations because it represents an underlying problem with that printer.