Abnormal Temperature Detection Using Machine Learning Feature Synthesis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormal temperature detection technologies struggle to accurately distinguish between changes in ambient temperature and temperature changes caused by fires or other anomalies, especially when there is no significant difference in the amount of change, leading to difficulties in setting appropriate threshold values and potential delays or false alarms.
Innovation Solution
The proposed abnormal temperature detection device employs a machine learning-based approach that involves preprocessing temperature data to calculate feature amounts, synthesizing these features with past determination information, and generating an identification model to accurately determine whether acquired temperature data is normal or abnormal, even in the presence of ambient temperature changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a threshold value method is used to detect abnormal temperatures, then the detection method is simple and easy to implement, but it cannot accurately distinguish between ambient temperature changes and fire-induced temperature changes when the amounts of change are similar
Solution Approach 1:
The patent transforms the temperature detection problem from using absolute temperature values to using temperature change rates (derivatives). By calculating the rate of temperature change and comparing it against dynamic thresholds that account for ambient temperature trends, the system can distinguish between gradual ambient changes and rapid fire-induced changes, resolving the contradiction between simple implementation and accurate detection.
Solution Approach 2:
The patent introduces dynamic threshold adjustment based on the current ambient temperature and its change rate. Instead of using a fixed threshold, the detection threshold adapts to the current thermal environment, allowing the system to maintain high detection accuracy across varying ambient conditions while keeping the implementation relatively simple through automated adaptation.
2Device complexity
If a fixed threshold value is used for abnormal temperature detection, then the detection logic is straightforward, but false alarms occur when ambient temperature changes are significant
Solution Approach 1:
The patent implements feedback by continuously monitoring the ambient temperature and its change rate, then using this information to dynamically adjust the detection threshold. The system feeds back the current thermal state to modify the decision criteria, enabling it to distinguish between normal ambient fluctuations and abnormal fire conditions, thereby reducing false alarms while maintaining reasonable detection logic complexity.
3Measurement precision
If machine learning methods are used to improve detection accuracy, then the ability to distinguish abnormal temperatures improves, but the device complexity and computational requirements increase
Solution Approach 1:
The patent simplifies the machine learning approach by transforming the raw temperature data into derived parameters such as temperature change rates and second derivatives. This parameter transformation reduces the complexity of the learning model while preserving the essential patterns needed for accurate detection, allowing the system to achieve high detection accuracy with relatively simple computational models.
Data Source
AI summary
An abnormal temperature detection device includes a temperature acquirer configured to acquire temperature data; a preprocessor configured to perform preprocessing of calculating a feature amount of the temperature data; a synthesizer configured to perform synthesis processing of synthesizing the feature amount and past determination information indicating whether the temperature data acquired in the past is normal or abnormal; a learner configured to generate an identification model by performing machine learning on a result of synthesis by the synthesizer; and a determiner configured to determine whether temperature data for detection which is acquired by the temperature acquirer, and on which the preprocessing is performed by the preprocessor and the synthesis processing is performed by the synthesizer, is normal or abnormal by using the identification model.