ML Peak Detection for Non-Routine Flaring Events
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting anomalous flaring events in industrial facilities are labor-intensive, time-consuming, and prone to errors, making it difficult to distinguish between planned and unplanned flaring, assess flaring trajectories, and identify underlying issues like leaking valves, which can lead to inefficiencies and safety risks.
Innovation Solution
A machine learning (ML) model is trained using historical flare data to automatically detect peaks in flaring, identify parameters, and manage flaring events by filtering and clustering data points to form clusters, enabling accurate detection of non-routine and emergency flaring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual plotting and visual scouting for peaks is performed by engineers, then detection capability is maintained, but time consumption and labor intensity increase significantly
Solution Approach 1:
The patent replaces the manual mechanical process of plotting data and visually scouting for peaks with an automated machine learning-based detection system. The system uses historical flare data, predefined definitions of flaring, and labeled flare data to train an AI/ML model that automatically identifies peaks, eliminating the need for engineers to manually plot and visually inspect data while maintaining detection accuracy.
Solution Approach 2:
The system enables self-service detection where the AI/ML model autonomously performs peak detection without requiring continuous human intervention. The model is trained on historical data and predefined criteria, allowing it to independently analyze flaring events, identify peaks, and provide detections that would otherwise require manual engineering review.
2Measurement precision
If engineers manually detect and investigate each flaring event, then detailed analysis is possible, but labor resources and errors increase
Solution Approach 1:
The patent replaces manual detection and investigation processes with an automated AI/ML-based system that processes flaring data algorithmically. The system uses trained models to identify peaks and parameters automatically, reducing human error and increasing consistency while maintaining the ability to provide detailed analysis through structured parameter identification and clustering.
Solution Approach 2:
The system incorporates feedback mechanisms where the AI/ML model continuously learns from historical flare data and predefined definitions to improve its detection accuracy. The model processes labeled flare data and adjusts its parameters to better identify peak characteristics, providing feedback loops that reduce errors and improve reliability over time.
3Extent of automation
If semi-auto-machine learning tools are used, then automation is introduced, but customization requirements and time investment increase
Solution Approach 1:
The patent applies preliminary action by pre-training the AI/ML model on historical flare data and predefined definitions of flaring before deployment. The model is pre-configured with labeled flare data and peak detection criteria, allowing it to perform automated detection without requiring extensive customization or time investment from users during actual operation.
Solution Approach 2:
The system provides universality through a multi-functional AI/ML platform that can handle various flaring scenarios, peak detection types, and analysis requirements through a single integrated system. The model is designed to process different data formats, apply multiple detection criteria, and provide comprehensive analysis without requiring separate customized tools for each function.
4Quantity of substance
If noisy data with erratic readings is processed, then complete data coverage is achieved, but baseline extraction becomes labor-intensive
Solution Approach 1:
The patent replaces manual baseline extraction processes with automated AI/ML algorithms that process noisy data and identify patterns algorithmically. The system uses the trained model to filter out erratic readings, identify meaningful baseline characteristics, and extract relevant information from noisy flaring data without requiring manual analysis while maintaining complete data coverage.
Solution Approach 2:
The system applies the extraction principle by automatically isolating and extracting meaningful baseline information from noisy flaring data. The AI/ML model identifies and separates signal patterns from noise, extracting relevant baseline characteristics and filtering out erratic readings to provide clean, actionable data while maintaining comprehensive data coverage.
Data Source
AI summary
A method for training a machine learning (ML) model for peak detection in flaring is disclosed. The method comprises receiving, via least one processor, historical flare data associated with one or more flare stacks over a predefined time period; training, via least one processor, an artificial intelligence (AI)/machine learning (ML) model, based at least on the historical data, predefined definitions of flaring, and labeled flare data; determining, via least one processor, one or more peaks in the flaring using the trained AI/ML model; identifying, via least one processor, one or more parameters associated with each of the one or more peaks; determining, via least one processor, whether the one or more parameters satisfy predefined parameters; and deploying, via least one processor, the trained AI/ML model for managing the flaring upon determining the one or more parameters associated with each of the one or more peaks satisfy the predefined parameters.