Boiler Pressure Trend Prediction Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing heating systems face issues with excessive pressure drops in heat transfer fluids, leading to blocking conditions and maintenance challenges, even without leaks in the hydraulic circuit.
Innovation Solution
A method using a machine learning classification model, trained on pressure signal data, predicts the likelihood of boiler blocking due to low pressure within a future time window, allowing for proactive maintenance interventions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the boiler operates with pressure values lower than the standard range, then the system can continue operating without blocking, but the pressure drop becomes excessive leading to maintenance issues
Solution Approach 1:
The system performs preliminary actions by analyzing historical pressure data and predicting future pressure drops before they cause blocking conditions. The machine learning model forecasts pressure trends and triggers maintenance notifications in advance, allowing proactive intervention rather than reactive blocking.
Solution Approach 2:
The system implements continuous feedback by monitoring pressure values in real-time, comparing them against predicted trends, and generating notifications when maintenance is needed. This closed-loop feedback enables the system to adapt to actual pressure behavior and adjust maintenance timing accordingly.
2Reliability
If the boiler enters blocking condition due to low pressure, then component damage is prevented, but operational disruptions occur requiring user intervention
Solution Approach 1:
The system performs self-service by automatically monitoring pressure trends, predicting blocking conditions, and notifying users or technicians in advance. This eliminates the need for user intervention to check pressure manually or respond to blocking conditions, as the system autonomously manages pressure surveillance and maintenance scheduling.
Solution Approach 2:
The system takes preliminary action by predicting pressure drops and issuing maintenance notifications before blocking conditions occur. This allows scheduled maintenance during convenient times rather than disruptive interventions when blocking occurs, improving ease of operation while maintaining component protection.
3Difficulty of detecting and measuring
If traditional pressure monitoring is used, then blocking conditions are detected, but progressive pressure reduction without leaks is not distinguished from actual leaks
Solution Approach 1:
The system performs preliminary analysis by collecting and analyzing historical pressure data to establish normal pressure patterns and trends for each system. This contextual understanding allows the model to distinguish between progressive pressure reduction (normal aging) and sudden pressure drops (leaks) by comparing actual readings against predicted trends based on historical behavior.
Solution Approach 2:
The system uses feedback from historical pressure data to continuously refine its understanding of normal pressure patterns. By comparing real-time pressure readings against learned historical trends, the system can detect anomalies that deviate from normal progressive pressure reduction, thereby distinguishing leaks from normal aging without losing contextual information.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
The object of the present invention is a method for controlling the pressure of the heat transfer fluid circulating in a heating system comprising a thermo-sanitary appliance (1), in particular a gas boiler, such method comprising in sequence at least the following steps: - Step 1: extrapolating the minimum values (P.MIN) of the pressure of the fluid from the original minimum values (P.RAW), within a time observation window (T.analysis) corresponding to a given number of previous days; - Step 2: identifying one or more ramps of the minimum values (P.MIN) of the pressure of the fluid, within said time observation window (T.analysis); - Step 3: calculating a plurality of characteristics relating to the trend of the pressure of the fluid on the basis of said one or more ramps of the minimum values (P.MIN) identified in Step 2; - Step 4: providing said characteristics calculated in Step 3 to a machine learning classification model, adapted to produce as output a three-class classification, each of which is representative of the more or less high probability that in a future time window (T.prox) said thermo-sanitary appliance (1) will block due to an excessively low value of the pressure of said fluid.