The way that this so called “diet” works is that it makes the incomplete data sets, or the data sets with errors, to be available for setting up the calibrations of regressions that are needed to teach our measuring system how to translate the measured data into the proper moisture and density values that are, of course, in relation with the values previously delivered from the customers reference values or laboratory values.
Sometimes the data sets, besides being incomplete, as mentioned previously can also contain errors; and it is very important for these things not to happen in order to make a stable calibration. These errors can easily come from false results, mixed up samples, problems in the reference values, or even form the production anomalies, also known as outliers. When it comes to these errors, the most difficult part is not to fix them but to recognizes which were the samples that gave you the errors.
Usually, where you find the most amount of problems are in products that are similar but still different from each other, for an easier example let’s say we have a product like cornflakes and sugared coated cornflakes, as you know they are both made from corn but with different final ingredients.
Let’s make an easy example of the type of problems you can encounter in these products, you have a bag of puzzle pieces, some belong to the same puzzle, others to a similar one but in a different color and some to a completely different puzzle and to make things worse, a lot of parts are missing. Pretty difficult to solve, no?
TEWS researchers now found a way how to analyze these big data sets and make it right.
To stay with the example of the puzzles, our tool will sort the different puzzles, will look to complete incomplete pictures with pieces of the same puzzle which only have a different color and reconstruct and add necessary pieces to get a full picture. This results in much less samples to be provided by the lab, and a smaller number of data points needed.
So, from thousands of different data sets you will get neat and clean data sorted out for different products and groups, even when the sets contain wrong data. As a result, you will have a stable group of calibrations/products settings.
For a skilled scientist this would take from days to weeks, and of course our system is also the result of 30 years of experience, but now available for all our customers as a quick, reliable, and affordable way of setting up your production. This will increase your calibration process up to 400%.
This is a game changer in the industry, it has never existed before. An algorithm that makes use of the similarities that different products have, this was created with a sophisticated mathematical method – which was never discovered. We are proud to say that we are the first ones to have it in the market.
Data science plays a very important role for us at TEWS, as our goal is to always provide robust and accurate microwave moisture measurement devices, and we keep striving to improve them as much as possible.
Let’s stop talking about our product and start improving your product set-up! Get in touch with our CSM.