When unsatisfied search queries define new passives

November 15, 2016 // By Julien Happich
Just before electronica, passive components manufacturer KEMET was promoting an advanced passive components search engine, offering engineers to search for part numbers way beyond the company's offering. ComponentEdge lets users search over 6.4 million part numbers with descriptive text or through interactive filtering, and the tool can cross reference from 145 different manufacturers.

At electronica, EETimes Europe met with KEMET's Vice President of Marketing Johnny Boan to understand what sort of benefits the passives company would get by listing more components than those on its portfolio. We suspected that data mining would be a reason, Boan confirmed.

An interesting feature of ComponentEdge is that when exact matches are not available, it displays the differences between the nearest matching components and gives the user the option to change filter criteria. Once part numbers are found, each result provides real-time distributor pricing, availability, links to 3D CAD models, RoHS information and part number specific datasheets.

"To design ComponentEdge, we hired three design engineers from distinct industries, Cable, Phone and T&M. Build us the website that you'd like as your engineering centre, we asked them", recalls Boan, making a point to provide as much information about the parts as possible to help engineers make their decisions.

"We use our in-house parameter simulation tool K-SIM to create characterization plots automatically based on the components' data sheets, hence we are able to give precise performance comparisons between our parts and even competition's offerings", Boan explained.

When asked about the sort of insight KEMET can get from data mining the search queries of such a broad engine, Boan explained: "We are constantly looking at the top 10 and top 1000 plots generated and that gives us a huge insight on what parts the designers are after. But what is of particular interest is the white space of unsatisfied queries, when designers have entered specific parameters and yet no part corresponds exactly to what they searched for. I go to our R&D department and ask them, can we build it?"

Now, since you've gone that far in data mining, how about using deep learning and taking all of KEMET's raw materials, manufacturing processes and known working designs to even automate the design path towards these new parts? We asked.