Table Encoder Error Detection for Industrial Plant Configuration Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Manual entry of configuration data in collaborative spreadsheets for industrial plants often leads to human errors and error propagation, especially when new equipment is introduced, due to incorrect values and auto-correction issues.

Innovation Solution

A method and system using a trained table encoder to automatically detect errors in configuration tables by estimating error probability values for each cell and comparing them with a threshold, with a comparator to identify erroneous cells, and an active learning strategy to correct or predict configuration properties.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If manual entry of configuration data is performed in collaborative spreadsheets, then flexibility and ease of operation are improved, but human errors and data quality deteriorate

Engineering Contradiction:
Improveease of configuration data entryVSAvoiddata quality
Core Design Contradiction:
Ease of operationVSReliability

Solution Approach 1:

The system enables self-service error detection by automatically analyzing configuration tables using trained machine learning models. The table encoder and error detector components autonomously identify potential errors without requiring manual verification, allowing the system to self-correct configuration issues while maintaining manual entry flexibility.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system implements feedback mechanisms by continuously monitoring configuration data quality through error probability estimation. When errors are detected, the system provides feedback to users through highlighted warnings and suggestions, enabling real-time correction while preserving the collaborative spreadsheet workflow.

Inventive Principle:
Principle #23Feedback

2Productivity

If configuration values are copy-pasted from existing entries for new equipment, then productivity is improved, but error propagation worsens

Engineering Contradiction:
Improveconfiguration speedVSAvoiderror propagation
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system acknowledges the necessity of copying configuration values for new equipment but introduces intelligent validation. The error detector compares copied values against expected patterns and flags anomalies, allowing rapid configuration while preventing error propagation through automated verification of copied data.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The system applies preliminary anti-action by proactively detecting and preventing error propagation before it affects new equipment configurations. The trained models identify suspicious patterns in copied data and raise warnings, allowing users to correct issues before finalizing the configuration.

Inventive Principle:
Principle #9Preliminary anti-action

3Reliability

If automated error detection is implemented using trained models, then data quality is improved, but system complexity worsens

Engineering Contradiction:
Improvedata qualityVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system segments the error detection task into specialized components: a table encoder for processing configuration data, an error detector for identifying issues, and an active learning manager for model improvement. This modular architecture manages complexity by dividing the automated detection system into independent, manageable modules with specific functions.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system introduces an intermediary layer between manual data entry and final configuration validation. The trained machine learning models act as intermediaries that automatically analyze configuration tables, estimate error probabilities, and provide suggestions, reducing the complexity burden on end users while maintaining high data quality.

Inventive Principle:
Principle #24Intermediary (Mediator)

4Measurement precision

If error probability estimation is performed for each cell, then measurement precision is improved, but computational energy consumption worsens

Engineering Contradiction:
Improveerror detection accuracyVSAvoidcomputational energy
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system applies partial action by focusing error probability estimation on specific high-risk cells or tokens within configuration tables rather than uniformly processing every cell. The active learning manager prioritizes cells with higher uncertainty or historical error patterns, reducing overall computational energy while maintaining detection precision for critical areas.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4425282A1Method and system for automatically detecting errors in industrial plant configuration tables
Publication Date: 2024.09.04 SIEMENS AG
  • EP4425282A1 patent drawingFigure 1~2
  • EP4425282A1 patent drawingFigure 3~5
  • EP4425282A1 patent drawing

AI summary

For automatically detecting errors in industrial plant configuration tables (CT), a table encoder (TE) is trained to estimate an estimated probability that the cells in the tables contain an error. According to an embodiment, the table encoder is initially trained with a set of configuration tables. An Active Learning Query Strategy (ALQS) selects candidate cells (CC) from the set of configuration tables with a mixed strategy combining uncertainty sampling with a penalty for picking multiple candidate cells from the same configuration table. A user interface receives labeled cells (LC), wherein the labeled cells contain the candidate cells as well as labels indicating whether the candidate cells are erroneous. A training component (TC) performs gradient updates (GU) on the table encoder, using the labeled cells. As a result, the table encoder is re-trained after each user interaction to become a table token classification model. Using active learning, which is novel with regard to token classification in tabular data, allows to reduce the required amount of labeled data efficiently. The Active Learning Query Strategy balances uncertainty versus diversity and ensures that the candidate cells are selected from a diverse range of configuration tables. This embodiment provides a novel active learning workflow for tabular data and deep machine learning models. Furthermore, this embodiment saves experts time in annotating cells in configuration tables.