I am Eelco Hoogendoorn, AI specialist, working on machine learning at 3D Hubs. My focus of the past few months has been on developing our instant quoting software for CNC machining and Injection molding.
In this short blog post, I’ll explain in some detail what we did that allowed us to release a whole bunch of new materials with instant quoting and what these developments could mean for the future of digital manufacturing.
On 3D Hubs, our goal is to provide our users with instant quotations for as many manufacturing technologies and materials as possible.
To accomplish this, we've developed an AI system that can analyze CAD files, learn from previously sourced parts (over 1,500,000 of them), rapidly source quotes in our global network of manufacturing service providers, and provide users with an instant quote, all in a few seconds!
It all started about a year ago, when we launched our CNC machining service. At the time, it took us up to 48 hours to provide quotations, as all the calculations were done by hand by our in-house manufacturing experts, who were working closely with our suppliers. After creating a database of fulfilled orders and accompanying quotes, we were able to build a machine learning system. This was the beginning of our smart sourcing engine we use today.
The current generation of our AI system uses deep neural networks and learns from our past orders in order to predict what is the best price possible for our future customers. It continually learns and updates its generalized guidelines on how to price a part, improving its predictions with each new quote.
Deep neural networks are a machine learning technique that has a reputation for being good at ‘generalizing’, hence it is perfect for our application. They have also proven very competent, as they allow us to continuously increase the range of design parameters that we can use to price and source with ever-increasing confidence.
Sourcing parts at the best possible price across our worldwide supplier network is a complex problem, and subtle differences in a design may greatly influence the options and prices available. Not only the process is very multifaceted but, as the data shows, also quite subjective. Suppliers may have different opinions on the difficulty of producing a specific part and none of them will be 100% sure until they actually machined it.
But by analyzing a large number of completed orders, our system can build up confidence that a given design will be manufacturable within a given price range, by some supplier in our network. This means that the system will show to our customers the best price instantly, often even before a supplier has accepted the order. This is a risk 3D Hubs can now confidently take.
Overall, the process saves time to engineers who are looking for CNC machined parts, as they no longer need to shop around, get a machinist up to speed and discuss quality assurance. They just have to upload a CAD file to 3D Hubs and get an instant quote.
The lessons we are learning about the current state of the digital manufacturing market through this process are helping us to build a smarter sourcing engine for engineers to use, now and in the future.
In the short term, we expect to see instant quoting available for injection molding, initially with the most common materials including PP, ABS, and PS. In the long term, 3D Hubs through the smart sourcing engine will be able to continually improve at mitigating the risks every engineer and supply manager faces today when sourcing parts the traditional way.
By combining our global network of manufacturing services with our automated Design for Manufacturing analysis tools, you can be confident that you’re not only getting the best price possible, but also that the part is manufactured the way you need it.
I hope this has given you an insight into what happens behind the scenes after you upload a part and what we do at 3D Hubs. I’d be happy to answer any questions you may have on the smart sourcing engine or machine learning in general, you can contact me via Eelco[@]3dhubs.com.