Use of Deep Learning to Analyze Product Lifecycle Data and Identify Errors

Penn State Applied Research Lab leads this project that aims to create a way to use deep learning, a subset of machine learning, to perform automated analyses of product lifecycle management (PLM) system components and detect data anomalies.


PLM systems manage a lot of data about a product’s lifecycle, including computer-aided design (CAD) models, drawings, and specifications. Depending on the software from which it originated, that data has many different formats. As a result, using a PLM system as a data repository can mean hundreds of thousands of files in heterogeneous formats. The volume of data coupled with the lack of standardization requires automated pruning. That pruning will decrease duplication that makes it difficult to find previously designed components and to cut the costs of the information technology needed to manage the infrastructure that supports the data and software.

Proposed Solution

Step one is development of a topography of PLM features and corresponding pain points that prevent standardization. Step two is using the findings to create a deep learning model that can detect anomalies. Step three is coming up with a transfer model and implementation pipeline so that anomaly detection and data analysis can be shared among manufacturers without compromising any single company’s data privacy.


All major manufacturers utilize at least one PLM or product data management (PDM) system so streamlining and better management of these systems can benefit a wide array of users in digital engineering environments. Plus with this model, many manufacturers can use the same workflow to improve identification of PLM errors and anomalies.