Industrial Virtual Assistant With AutoML for Plant Predictions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Operators in industrial plants face challenges in obtaining necessary data analytics and predictions for maintenance and anomaly detection, which are not foreseeable during system design, requiring advanced analytical expertise and effort, often relying on experience and gut feeling due to the complexity of matching topology information with sensor data.
Innovation Solution
A method for controlling a virtual assistant that uses an input interface to receive information requests, determines a model specification, and employs a machine learning model to provide responses, including process variable values and alarms, with an autoML pipeline for generating suitable models in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If virtual assistants are used to aid operators in industrial plants, then information needs can be addressed, but the complexity of matching topology information with sensor data and generating predictions remains high, requiring advanced analytical expertise
Solution Approach 1:
The patent introduces an intermediary system comprising a control unit and model manager that mediates between the virtual assistant's information requests and the industrial plant's sensor data. This intermediary automatically performs the complex task of matching topology information with sensor data using machine learning models, shielding operators from the underlying complexity while providing easy access to predictions and analytics through natural language interfaces
2Measurement precision
If advanced data analytics and predictions are generated manually, then accurate information can be obtained, but it requires significant analytical expertise and time effort
Solution Approach 1:
The system implements self-service through automated machine learning model selection and execution. The model manager automatically selects appropriate machine learning models based on the information request and executes them without human intervention. This allows the system to generate accurate predictions and analytics autonomously, eliminating the need for operators to manually select models or have advanced analytical expertise while maintaining high accuracy
Solution Approach 2:
The patent employs preliminary action by pre-configuring multiple machine learning models and making them available in a model library. When an information request is received, the system can quickly retrieve and execute pre-prepared models rather than creating them from scratch. This preliminary preparation significantly reduces the time required to generate predictions while maintaining accuracy
3Reliability
If machine learning models are selected and configured manually, then accurate predictions can be made, but the process becomes difficult and time-consuming during operations
Solution Approach 1:
The patent implements universality through a multi-functional model manager that can handle various types of information requests using different machine learning models. The system maintains a library of diverse models that can address different prediction needs (anomaly detection, time series forecasting, pattern recognition, etc.). This universal approach allows the system to reliably handle multiple types of predictions through a single automated interface, maintaining both reliability and high productivity during operations
Data Source
AI summary
A method for controlling a virtual assistant for an industrial plant includes receiving by an input interface an information request, wherein the information request comprises at least one request for receiving information about at least part of the industrial plant; determining by a control unit a model specification using the received information request; determining by a model manager a machine learning model using the model specification; and providing by the control unit a response to the information request using the determined machine learning model.