A Search Engine for Manufacturers Using Product Manufacturing Information (PMI) Enhanced 3D Model Search

The type and complexity of products designed by small and large teams have drastically changed due to advancements in digital design tools, manufacturing process technology and connectivity afforded by the internet. Finding available and reliable manufacturers is time consuming and is often tasked to sourcing agents who scour through existing supplier databases or found using keywords through web search. In this paper, we investigate 3D shapes augmented with product manufacturing information (PMI) as the key source to match user request to appropriate manufacturers who are capable of completing a job. The key challenges are gathering data about manufacturers’ capabilities, obtaining data with regards to the models that they have produced in the past and then finally matching algorithms to work with product engineering models. We present a solution approach that addresses portions of this challenge by building a search engine system that supports user queries based on 3D models and/or manufacturing process related text keywords. To enable text based queries, we built a web-crawler that searched through thousands of websites owned by job shop service manufacturers, extracted discriminating text from their public webpages and indexed them to create a capability profile about a job shop service provider. For 3D model based queries, we extended the original spherical harmonics based 3D shape search to match parts that also share tolerance information tagged to the user provided 3D model. The approach provides us with discriminative power in identifying objects that also share similar GD&T call outs on the model as they can often entail various additional process plans. We have tested our approach against a repository of 200 models and text data collected from websites of manufacturers. Our results show that that we are able to identify parts fairly quickly given a multi-modal user input. Future work must involve testing it against much larger data sets for robustness and wide scale applicability.