Micro-Seismic Event Classification Using Dual ML Pipelines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current microseismic monitoring systems generate vast amounts of data that require extensive manual review to identify significant incidents affecting oil and gas production operations, leading to inefficiencies and delays in corrective actions.
Innovation Solution
A method utilizing two fault-tolerant machine learning pipelines, an acoustic and a visual pipeline, to classify microseismic events by converting data into power spectrum and spectrogram representations, applying deep learning models for accurate prediction and filtering out noise, thereby reducing the need for manual review and enhancing data processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual review is used to classify microseismic events, then classification accuracy can be maintained through expert judgment, but processing time and operational efficiency deteriorate due to extensive manual analysis requirements
Solution Approach 1:
The patent replaces manual mechanical review processes with automated machine learning systems. Two distinct ML pipelines (acoustic and visual) are implemented to automatically classify microseismic events, substituting human expert analysis with computational algorithms that process data faster while maintaining or improving classification accuracy through consistent application of trained models.
Solution Approach 2:
The patent introduces machine learning models as intermediary systems between raw microseismic data and final classification results. These ML pipelines act as mediators that transform raw sensor data through multiple processing stages (feature extraction, classification, validation) to produce accurate event classifications without requiring direct manual intervention for each event.
2Reliability
If extensive manual review is conducted to identify significant incidents, then classification thoroughness is improved, but productivity and response time to operational incidents deteriorate
Solution Approach 1:
The patent segments the classification task into two independent machine learning pipelines: an acoustic pipeline that analyzes audio characteristics of microseismic events and a visual pipeline that examines waveform patterns. This segmentation allows parallel processing of different event features, improving both thoroughness through multiple analysis angles and productivity through concurrent execution, thereby reducing overall response time.
Solution Approach 2:
The patent implements preliminary filtering and preprocessing steps within the ML pipelines that automatically identify and flag significant incidents before full classification. This preliminary action enables the system to prioritize processing of critical events, improving response time for urgent operational incidents while maintaining thorough classification through subsequent detailed analysis.
3Device complexity
If traditional data processing methods are used, then system complexity remains manageable, but noise filtering capability and false positive reduction deteriorate
Solution Approach 1:
The patent creates copied and transformed representations of the original microseismic data through spectrograms and other feature extractions. These copied representations (visual and acoustic features) allow the ML pipelines to analyze events from multiple perspectives simultaneously, improving noise filtering capability and reducing false positives while the modular pipeline architecture keeps overall system complexity manageable through reusable processing components.
Data Source
AI summary
A method for classifying a microseismic event, including: analyzing microseismic event files through a combination of two fault tolerant machine learning pipelines, an acoustic machine learning pipeline and a visual machine learning pipeline; and generating a classification prediction for the microseismic event files by combining predictions from the acoustic machine learning pipeline and the visual machine learning pipeline.