Odor detection method for refrigeration appliance and refrigeration appliance

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing technologies fail to accurately detect and eliminate peculiar smells in refrigeration appliances, affecting user experience.

Innovation Solution

An odor detection method using multiple machine learning models, where gas data is processed by first models trained on single types of food for high accuracy and second models trained on mixed types of food for adaptability, determining odor attributes based on prediction confidence and deviations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a single type of food data model is used for odor detection, then the prediction accuracy for specific food types is high, but the adaptability to mixed food types is poor

Engineering Contradiction:
Improveodor detection accuracyVSAvoidadaptability to mixed food types
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent segments the odor detection task by training separate machine learning models for different food types (meat, seafood, vegetables, fruits) and then combining their predictions. Each model specializes in detecting odors from its specific food type, achieving high accuracy for individual foods while the ensemble approach provides adaptability to mixed food scenarios.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent merges multiple single-food-type models into a comprehensive detection system. By integrating the predictions from several specialized models through a voting mechanism, the system achieves both the high accuracy of specialized models and the adaptability needed for mixed food type detection.

Inventive Principle:
Principle #5Merging (Combining)

2Measurement precision

If multiple machine learning models are used for odor detection, then the detection accuracy is improved, but the model complexity increases

Engineering Contradiction:
Improveodor detection accuracyVSAvoidmodel complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The complex detection task is segmented into multiple simpler sub-tasks, each handled by a dedicated machine learning model trained on specific food type data. This segmentation allows each model to remain relatively simple while the collective system achieves high accuracy through the combination of specialized models.

Inventive Principle:
Principle #1Segmentation

3Reliability

If prediction confidence is used to determine odor attributes, then the reliability of odor detection is improved, but the detection time increases

Engineering Contradiction:
Improvereliability of odor detectionVSAvoiddetection time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system performs partial evaluation by first assessing prediction confidence levels before committing to a final odor attribute determination. When confidence is sufficient, the detection process can be concluded quickly; when confidence is low, additional processing is performed. This partial action approach maintains reliability while minimizing unnecessary time consumption.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4174487B1Odor detection method for refrigeration appliance and refrigeration appliance
Publication Date: 2024.07.24 BSH HAUSGERATE GMBH
  • EP4174487B1 patent drawingFigure 1
  • EP4174487B1 patent drawingFigure 2~4
  • EP4174487B1 patent drawingFigure 5

AI summary

An odor detection method for a refrigeration appliance and a refrigeration appliance are disclosed. The method includes: obtaining gas data in the refrigeration appliance; inputting the gas data into a plurality of first preset machine learning models respectively, and obtaining first prediction results of the first preset machine learning models, where the first prediction result includes an odor attribute in the refrigeration appliance and a corresponding probability, and the first preset machine learning model is trained based on gas data of a single type of food under a plurality of odor attributes; determining prediction confidences of the first preset machine learning models according to differences between a plurality of first prediction results; and determining, if the prediction confidences are high, an odor attribute corresponding to a probability maximum in the plurality of first prediction results as a current odor attribute in the refrigeration appliance. The solutions of the present invention can improve the accuracy of odor detection, which is conducive to timely detection and elimination of a peculiar smell, thereby better maintaining air in the refrigeration appliance fresh.