Industrial Plant Virtual Assistant for Automatic ML Model Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Industrial plant operators face challenges in obtaining data analytics and predictions during operations, as these needs emerge unpredictably and require advanced analytical expertise, often relying on experience and gut feeling due to the complexity of matching analytical models with current situations.

Innovation Solution

A computer-implemented method for controlling a virtual assistant that receives information requests, determines a machine learning model specification, and provides responses using a machine learning model, incorporating user action and context information to offer data analytics and predictions, enabling real-time information delivery and automatic generation of models.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If operators rely on experience and gut feeling to solve information needs, then the system is easier to operate, but the measurement precision and reliability of data analytics are insufficient

Engineering Contradiction:
Improveprecision of data analyticsVSAvoidease of using analytical models
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system performs automatic model selection and configuration without requiring operator expertise. The virtual assistant autonomously selects appropriate analytical models, configures parameters, and generates predictions based on the information request, eliminating the need for operators to manually match models with situations.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The virtual assistant acts as an intermediary between operators and complex analytical models. It translates natural language information requests into model specifications, automatically selects and configures appropriate models, and presents results to operators, bridging the gap between user simplicity and analytical complexity.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Adaptability or versatility

If the system provides comprehensive data analytics and predictions, then the information needs are better matched, but the device complexity increases

Engineering Contradiction:
Improveadaptability to information needsVSAvoidcomplexity of model management system
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The virtual assistant provides a universal interface that handles diverse information needs through a single system. It can process various types of requests (predictions, anomaly detection, pattern matching) using multiple analytical models, making the system adaptable to different information needs without requiring separate specialized systems.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system uses model specifications as abstract representations that can be stored, reused, and modified. Once a model is selected and configured for a particular information need, the specification can be copied and applied to similar future requests, reducing the complexity of managing diverse analytical solutions.

Inventive Principle:
Principle #26Copying

3Productivity

If the system automatically generates analytical models, then the productivity is improved, but the difficulty of detecting and measuring appropriate models increases

Engineering Contradiction:
Improvespeed of generating analyticsVSAvoiddifficulty of selecting appropriate models
Core Design Contradiction:
ProductivityVSDifficulty of detecting and measuring

Solution Approach 1:

The system pre-configures model specifications and maintains a library of available analytical models with their parameters and requirements defined in advance. When an information request is received, the system quickly matches the request against pre-defined model specifications rather than creating models from scratch, improving productivity while managing complexity through pre-prepared configurations.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3979017B1Method for controlling a virtual assistant for an industrial plant
Publication Date: 2024.08.28 ABB (SCHWEIZ) AG
  • EP3979017B1 patent drawingFigure 1~2
  • EP3979017B1 patent drawingFigure 3
  • EP3979017B1 patent drawingFigure 4

AI summary

The invention relates to a method for controlling a virtual assistant (10) for an industrial plant, comprising: receiving (S10), by an input interface (20), an information request (Ir), wherein the information request (Ir) comprises at least one request for receiving information about at least part of the industrial plant; determining (S20), by a control unit (30), a model specification (Ms) using the received information request (Ir); determining (S30), by a model manager (40), a machine learning model (M) using the model specification (Ms); providing (S40), by the control unit (30), a response (R) to the information request (Ir) using the determined machine learning model (M).