Process Plant Variable Monitoring Using Rate-of-Change Clusters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for monitoring measured variables in process plants require detailed knowledge of the process and are complex, involving separate models for each process step and frequent data acquisition, making them inefficient and costly.
Innovation Solution
A computer-implemented method that continuously acquires measurement data to determine vectors representing rates of change of measured variables, which are then compared to a reference cluster to detect anomalies, allowing for real-time monitoring without complex model creation or detailed process knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detailed process knowledge and separate models for each process step are used for monitoring, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent combines multiple process step models into a single integrated monitoring system that processes measurements from all process steps uniformly. Instead of creating and maintaining separate models for each process step, the system uses a unified approach where all measuring instruments are monitored through common evaluation criteria, significantly reducing device complexity while maintaining anomaly detection capability
Solution Approach 2:
The monitoring system is designed to be universal and applicable to all process steps and measuring instruments without requiring specific process knowledge. The same monitoring methodology and evaluation functions are used across different process steps, making the system multi-functional and eliminating the need for separate specialized models for each step
2Reliability
If frequent data acquisition is performed for continuous monitoring, then reliability is improved, but loss of time increases
Solution Approach 1:
The system performs preliminary actions by pre-defining evaluation functions and reference ranges during system setup. Measurement data is continuously acquired and immediately evaluated against these pre-established criteria, eliminating the need for time-consuming model creation and analysis during operation. This allows frequent data acquisition with minimal processing delay
3Measurement precision
If separate models for each process step are created, then measurement precision is improved, but ease of manufacture worsens
Solution Approach 1:
The system segments the monitoring task by measuring instrument rather than by process step. Each measuring instrument is evaluated independently using the same universal methodology, which simplifies system implementation. The segmentation approach allows straightforward integration of new instruments without requiring complex inter-step model relationships
4Measurement precision
If detailed process knowledge is required for model creation, then measurement precision is improved, but ease of operation worsens
Solution Approach 1:
The monitoring system operates autonomously without requiring detailed process knowledge from operators. The system automatically evaluates measurement data against pre-defined reference ranges and evaluation functions, performing self-service anomaly detection. Operators simply need to interpret the system's output without needing to understand complex process relationships or create custom models
Data Source
AI summary
A monitoring method includes monitoring variables that change over time and are continuously measured during repeated executions of a dynamic process. The measurement data include training data in a training period and monitoring data measured subsequently to the training period. Vectors are continuously determined based on the measurement data, the vector components of which comprise respective rates of change of the measured values of the individual measured variables determined for a series of successive points in time on the basis of the measurement data, wherein the points in time of each series cover a time window of a duration and the time windows of successive vectors are shifted. The vectors are stored as a reference cluster. The vectors are compared with the reference cluster. An anomaly is determined when a vector is outside the reference cluster.