Information extraction from scanned manufacturing diagrams using Artificial Intelligence (AI) and Machine Learning (ML)

Business objectives

Engineering schematics are represented in several ways, including via piping and instrumentation diagrams (P&IDs). P&ID diagrams will illustrate schematics of a process flow through pipes, tanks, actuators, and control valves as well as complex engineering workflows. Piping and instrumentation diagrams (P&ID) design and engineering is a time-consuming and labour-intensive process. P&ID diagrams have been manually generated, scanned and stored as image files. Coforge offers a service to automate the extraction of components shown in drawings and manuals pertaining to components, in order to provide business insights and benefits.

Challenges in current process

• Due to noisy textual information and inaccurate identification of symbols, which usually have minute visual changes, mapping drawing with the plant is a difficult procedure to implement
• Lack of ability to mark missed components with the bill of materials and estimate the production cost
• Identification of pipelines that may converge and diverge at different points in the image

Solution

• Coforge Intelligent Document Processing solution provides end-to-end data extraction from P&ID diagrams using Machine Learning models. State-of-the-art deep learning models identify and isolate valves, pipeline codes, pipelines, inlets and outlets for detecting symbols
• Text extraction utilising OCR engines, a key component of P&IDs that identifies and describes diagram elements
  

• Once the component has been identified the solution will locate the component's operating, threshold, performance, and troubleshooting parameters from its datasheets, installation instructions, and warranty documents. The parameters will be stored in the scalable database
• The extracted fields used to justify the design decisions and outline the specifications for the processes
• Real time data from the Plant is compared against the threshold parameters stored in Database for providing Alerts, Predict Alarms and perform predictive maintenance

Benefits

• Version management of Drawings
• Flagging missing components in the Drawing
• Clustering/Search Components based on projects/plant
• Real-time Alerts
• Alarm prediction
• Predictive maintenance
• Design justification