Back to results
US 2020/0005094 A1Published application
Machine learning analysis of piping and instrumentation diagrams
AssigneeSchneider Electric Systems USA, Inc.
InventorsBhaskar Sinha, Ashish Patil, Amitabha Bhattacharyya, Venkatesh Jagannath, Sameer Kondejkar
Life of this patent
Estimating the term…
Abstract
Automated evaluation and extraction of information from piping and instrumentation diagrams (P&IDs). Aspects of the systems and methods utilize machine learning and image processing techniques to extract relevant information, such as tag names, tag numbers, and symbols, and their positions, from P&IDs. Further aspects feed errors back to a machine learning system to update its learning and improve operation of the systems and methods.
Claims20
A method, comprising:
- accessing, by a machine learning system executing on a computing device, an image of a piping and instrumentation diagram (P&ID);
- determining, using a symbol classification engine of the machine learning system executing on the computing device, whether the accessed image includes a symbol indicative of a process control element, the symbol classification engine being trained, using supervised learning, to classify images into a plurality of classes based on whether the accessed image includes the symbol indicative of the process control element and a relative position of the symbol in the P&ID;
- storing, in a computer-readable memory storage device, a symbol output indicating the symbol indicative of the process control element upon determining the accessed image includes the symbol;
- determining, using a tag classification engine of the machine learning system executing on the computing device, whether the accessed image includes a tag corresponding to the symbol, the tag classification engine being trained, using supervised learning, to classify images into a plurality of classes based on whether the accessed image includes the tag corresponding to the symbol;
- and
- storing, in the computer-readable memory storage device, a tag output indicating the tag corresponding to the symbol upon determining the accessed image includes the tag.
3 more elements
A system, comprising:
- a piping and instrumentation diagram (P&ID) analysis processor;
- and
- a computer-readable storage device storing processor-executable instructions, wherein the processor-executable instructions, when executed by the processor, comprise:
- an image converter configured to receive a P&ID in an electronic format and convert the P&ID into a compute image;
- one or more filtering algorithms configured to process the compute image to obtain an approximate size of a symbol therein;
- a symbol extractor configured to extract the symbol from the compute image and store the extracted symbol on the computer-readable storage device;
- one or more tag pre-processing algorithms configured to remove symbol lines from the symbol in the compute image;
- a tag extractor configured to extract a tag component of the symbol in the compute image and store the extracted tag on the computer-readable storage device;
- and
- an intelligent system configured to:
- receive symbol feedback from the one or more filtering algorithms,
- receive tag feedback from the one or more tag pre-processing algorithms,
- retrain the symbol extractor based on the received symbol feedback, and
- retrain the tag extractor based on the received tag feedback.
11 more elements
A method, comprising:
- receiving a digitized piping and instrumentation diagram (P&ID);
- analyzing, using a symbol engine, the digitized P&ID to identify and extract a symbol from the digitized P&ID, wherein said analyzing results in at least one symbol error;
- analyzing, using a tag engine, the digitized P&ID to identify and extract a tag from the digitized P&ID, wherein said analyzing results in at least one tag error;
- storing the extracted symbol and tag on a computer-readable storage medium;
- retraining the symbol engine based on the at least one symbol error;
- and
- retraining the tag engine based on the at least one tag error.
4 more elements
Dependencies are inferred from the claim text.