Abnormal Temperature Detection Using Relative Change Rate Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormal temperature detection technologies face challenges in accurately identifying temperature changes due to daily and seasonal variations, leading to potential erroneous detections and difficulties in setting appropriate thresholds, especially in varying ambient environments.
Innovation Solution
An abnormal temperature detection device that includes a temperature acquirer, a preprocessor for calculating difference data and generating substituted data, and a learner using machine learning to determine if temperature data is normal or abnormal by comparing it to reference data, thereby improving accuracy and reducing environmental influence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional threshold-based detection is used to identify abnormal temperatures, then the detection method is simple, but the detection accuracy deteriorates due to daily and seasonal temperature variations causing erroneous detections
Solution Approach 1:
The patent transforms the temperature detection approach by changing from absolute temperature threshold comparison to relative temperature change rate analysis. Instead of using fixed thresholds that fail under varying environmental conditions, the system calculates the rate of temperature change over time periods, adapting the detection criterion to dynamic conditions. This resolves the contradiction by improving detection accuracy through parameter transformation while maintaining reasonable system complexity.
Solution Approach 2:
The patent introduces dynamic analysis by examining temperature changes over multiple time periods rather than using static threshold values. The system compares temperature changes across different time windows (e.g., 1-hour, 6-hour, 12-hour periods) to identify abnormal patterns. This dynamic approach adapts to daily and seasonal variations, improving detection accuracy without requiring complex environmental modeling.
2Loss of information
If multiple deviated values are introduced to detect temperature change direction, then the temporal change direction can be identified, but the probability of erroneous detection increases when there is contradiction between detection results
Solution Approach 1:
The patent segments the temperature detection analysis into multiple time-period intervals (e.g., 1-hour, 6-hour, 12-hour periods) and evaluates temperature changes in each segment independently. By dividing the detection process into hierarchical time segments, the system can identify abnormal temperature patterns at appropriate temporal scales while reducing false alarms from normal fluctuations. This segmentation approach preserves temperature change direction information while improving detection reliability through multi-scale validation.
3Adaptability or versatility
If fixed thresholds are used for abnormal temperature determination, then the detection algorithm is simple, but it is difficult to set appropriate thresholds in varying ambient environments
Solution Approach 1:
The patent implements a self-adaptive detection system that automatically adjusts to environmental conditions without requiring manual threshold configuration. By using relative temperature change rates compared to historical patterns and multiple time-period analyses, the system self-calibrates to daily and seasonal variations. This eliminates the need for environment-specific threshold tuning while maintaining high adaptability, resolving the contradiction between environmental versatility and system complexity.
Data Source
AI summary
An abnormal temperature detection device includes a temperature acquirer configured to acquire temperature data, a preprocessor configured to perform preprocessing for calculating difference data indicating a difference sequence in the temperature data, generating substituted data in which either of a piece of difference data which is a positive number or a piece of difference data which is a negative number is substituted with zero, and calculating a similarity between the substituted data and reference data obtained from training data in which normality or abnormality has been identified, a learner configured to perform machine learning on the similarity calculated by the preprocessor and output an identification function, and a determiner configured to determine whether temperature data for detection acquired by the temperature acquirer and subjected to the preprocessing by the preprocessor is normal or abnormal by using the identification function.