Pipe Condition Modeling for Time-Based Failure Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting pipe failures in fluid networks are inaccurate due to limited historical data and the need for extensive direct inspections, failing to provide reliable estimates of both current and future pipe conditions.
Innovation Solution
A method involving clustering pipe sections into classes based on structural and environmental parameters, conducting condition assessments on a representative sample, and using supervised machine learning to estimate pipe conditions and predict failures with a predictive model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If direct inspection methods (visual, electromagnetic, acoustical) are used to evaluate pipe condition, then measurement precision is improved, but productivity deteriorates due to the need to inspect each pipe individually and the inability to perform future failure predictions
Solution Approach 1:
The patent segments the pipe network into groups based on shared characteristics (material, diameter, age, environment) and performs inspections on representative samples from each group rather than every individual pipe. This segmentation allows the system to maintain measurement precision for condition assessment while dramatically improving productivity by reducing the total number of inspections required.
Solution Approach 2:
The patent creates statistical models that copy the behavior patterns observed in inspected pipes to predict conditions of uninspected pipes. By using historical failure data and condition assessment results from sample pipes, the system generates predictive models that estimate failure probabilities for the entire network, eliminating the need to physically inspect every pipe while maintaining assessment accuracy.
2Productivity
If statistical models based on historical pipe break data are used, then productivity is improved by reducing inspection needs, but measurement precision deteriorates due to lack of current pipe state information
Solution Approach 1:
The patent implements a feedback mechanism where actual inspection results from representative sample pipes are continuously fed back into the statistical models. This feedback loop allows the models to be calibrated and updated with real-world condition data, ensuring that predictions for the broader network remain accurate while maintaining high productivity through reduced inspection requirements.
Solution Approach 2:
The patent transforms the statistical models by incorporating multiple parameters including pipe material, diameter, age, environmental conditions, and observed failure patterns. By changing and expanding the parameter set used in statistical analysis, the system achieves both high productivity through efficient data utilization and high measurement precision through multi-factor predictive accuracy.
3Measurement precision
If comprehensive pipe network inspections are performed to gather sufficient data for accurate predictions, then measurement precision is improved, but loss of time increases due to the extensive inspection requirements
Solution Approach 1:
The patent performs preliminary actions by strategically selecting and inspecting representative sample pipes from each network segment before conducting full-scale predictions. This preliminary inspection of key representative pipes provides sufficient data to calibrate models and generate accurate predictions for the entire network, significantly reducing the total time required compared to comprehensive inspections of all pipes.
Solution Approach 2:
The patent applies partial action by inspecting only the necessary portion of the pipe network (representative samples) rather than the entire network. This partial inspection approach provides sufficient data for accurate statistical modeling and failure prediction, achieving the required measurement precision while minimizing time loss by avoiding redundant inspections of all pipes.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a computer-implemented method, computer program, and device for evaluating timed-based probabilities of failure of sections of a pipe network. To do so, the pipe sections are clustered into classes based on structural and environmental parameters; within each class a sample of pipe sections are selected to be inspected. The scores that are obtained through the inspection are used to train a model of pipe conditions of pipes in a class, in order to estimate the pipe conditions of pipes that have not been inspected. The pipe conditions are used to parameterize a predictive model of pipe failures.