Flare Analytics Using Thermal Imaging for Smoke-Steam Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current flare monitoring systems in industrial plants rely on human operators, which are prone to errors due to lapses in judgment and attention, and struggle to accurately distinguish between smoke, steam, and flame, especially in low visibility conditions such as harsh weather or nighttime, necessitating an automated and intelligent monitoring solution.
Innovation Solution
An advanced system employing a deep learning core with optical flow mechanisms and data augmentation, utilizing both vision and thermal cameras to differentiate between smoke, steam, and flame at a pixel level, and providing global connectivity for continuous learning and remote monitoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If human operators are used to monitor flare operations, then continuous monitoring is achieved, but errors occur due to lapses in human judgement and attention
Solution Approach 1:
The patent replaces the mechanical human monitoring system with an automated image processing system that uses optical flow algorithms and deep learning neural networks to detect smoke, steam, and flame. This substitution eliminates human error while maintaining continuous monitoring capability, directly resolving the contradiction between reliability and measurement precision.
Solution Approach 2:
The system performs self-monitoring and self-diagnosis through automated image analysis. The neural network continuously processes camera feeds to detect irregularities without human intervention, enabling the system to serve itself in detecting and reporting flare anomalies, thus improving both reliability and precision simultaneously.
2Measurement precision
If human operators monitor flare operations, then real-time detection is possible, but accuracy decreases in low visibility conditions such as harsh weather or nighttime
Solution Approach 1:
The patent transitions from single-spectrum visual monitoring to multi-spectrum analysis by incorporating thermal imaging alongside visible light cameras. This dimensional expansion allows the system to detect thermal signatures of smoke and flame that are invisible to the human eye, particularly effective in nighttime and adverse weather conditions, thereby improving detection accuracy while enhancing environmental adaptability.
Solution Approach 2:
The system dynamically adjusts monitoring parameters based on environmental conditions. The neural network processes varying image qualities and adjusts detection thresholds, processing speeds, and analysis depth according to lighting conditions, weather patterns, and visibility levels, enabling accurate detection across diverse environmental scenarios without compromising precision or adaptability.
3Productivity
If automated image processing is used to distinguish smoke, steam, and flame, then detection speed increases, but system complexity increases
Solution Approach 1:
The patent segments the complex monitoring task into distinct processing stages: optical flow computation for motion detection, region extraction for candidate identification, and deep learning classification for final categorization. This segmentation allows each module to be optimized independently for speed while managing overall system complexity through modular architecture and parallel processing.
Solution Approach 2:
The system implements a two-stage processing approach where optical flow analysis first identifies candidate regions of interest, and only these selected regions undergo computationally intensive deep learning classification. This partial action strategy processes only necessary portions of the image data at high computational cost, achieving high detection speed while controlling overall system complexity by avoiding full-image intensive processing.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system achieves high accuracy (>90% in all conditions and >95% in ideal conditions) in distinguishing between smoke, steam, and flame, reducing human error and enabling continuous, reliable monitoring and control of flare operations across multiple plants, even in challenging weather or lighting conditions.
Implementation Method 1
utilizing both vision and thermal cameras to differentiate between smoke, steam, and flame
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and system for advanced flare analytics in a flare operation monitoring and control system is disclosed that contains a data acquisition and augmentation mechanism whereby data is aquired through a plant network including images of the flare operations from single or multi-camera hubs. A machine learning-based self-adaptive industrial automation system process the images and data and assigns pixels to the images according to categories selected from smoke, flame and steam. The results of the analysis are displayed and a notice is issued when the percentage of pixies in a specific category falls outside a predeterminmed range.