Predictive Maintenance by Intelligent Mining of Manufacturing Incidents Using Spark NLP

This session presents lessons learned from a real-world manufacturing company that saved over 3,600 hours of manual labor in one year.

This was done by applying Spark NLP to automate root cause analysis of service trip reports – using a combination of automated summarization, ngrams, named entity recognition, and sentiment analysis in one scalable NLP pipeline.

We’ll cover the end-to-end process from raw documents to communicating the root cause of failures by machine type, and how this can be extended into a predictive maintenance model that predicts incident types given a work order problem statement.

The initial load of all history included 2,000+ documents that were between 5 and 12 pages long each with multiple tables. Whenever a new service trip report is entered into a SharePoint site, within 10 minutes, the document is processed and available in the Power BI dashboard.

Previously, this manufacturer had human beings in a conference room for 3 weeks reading groups of reports. Now reports are processed timely with no human effort and processed accurately.

The automation of the service trip reports is in production. Currently, this work is continuing with developing a predictive maintenance model that includes Spark NLP’s word embeddings, sentence embeddings, and deep learning text classification models.

About the speaker

Angelina Maria Leigh 

Data Scientist at Hitachi Solutions America

Angelina is a Data Scientist at Hitachi Solutions America where she specializes in applying Natural Language Processing (NLP) solutions to clients across industries. Angelina is also an Ambassador of Women in Data Science (WiDS) at Stanford University where she is building a Chicago Education Outreach team to inspire and educate high school students in data science, artificial intelligence (AI), and related fields.

With five years of academic research experience in neuroendocrinology and mindfulness meditation, Angelina has a strong track record of systematically investigating a problem and assisting in solving the problem.


Fred Heller 

Data Scientist at Hitachi Solutions America

Fred is academically trained as a mathematician and computer scientist and also trained as a chef in French and Italian traditions with a culinary arts degree.

This combination of mathematics and food production led to a career in supply chain management where he developed a passion for optimization tools while looking to reduce costs and increase customer satisfaction by locating warehouses and scheduling fleet movements optimally.

He also became interested in the Lean Six-Sigma methods for reducing waste and earned a Lean Six Sigma Green-Belt while working on an accounts payable data management problem.  Fred now works across a variety of Data Science specializations including NLP, IIOT, and Forecasting.



Sessions: October 5 – 7
Trainings: October 4, 12 – 15


Presented by