Air Conditioner Data Thinning for AI Learning Traffic Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The large amount of operation history data transmitted from air conditioners to server devices for AI learning increases communication traffic, potentially lowering learning accuracy if data is simply deleted to reduce traffic.

Innovation Solution

An air conditioner system that includes an adapter which acquires operation history data, determines if the data change is within a predetermined range, and thins the data by deleting unnecessary entries before transmitting it to a server, maintaining learning accuracy while reducing traffic.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If all operation history data are transmitted to the server device, then learning accuracy of AI is maintained, but communication traffic becomes large

Engineering Contradiction:
Improvelearning accuracyVSAvoidcommunication traffic
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential operation history data that is necessary for AI learning by identifying and removing redundant data. The system determines which data items contribute to learning accuracy and transmits only those, rather than transmitting all collected data. This extraction principle resolves the contradiction by separating useful data from unnecessary data, maintaining learning accuracy while reducing communication traffic.

Inventive Principle:
Principle #2Taking out (Extraction)

2Quantity of substance

If operation history data are deleted to reduce communication traffic, then communication traffic is reduced, but learning accuracy may be lowered

Engineering Contradiction:
Improvecommunication trafficVSAvoidlearning accuracy
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent applies local quality by treating different operation history data items differently based on their importance for learning. Instead of uniformly deleting or transmitting all data, the system selectively processes each data item according to its specific contribution to AI learning. This allows the system to maintain high learning accuracy for critical data while reducing traffic for redundant data, resolving the contradiction through differentiated data handling.

Inventive Principle:
Principle #3Local quality

3Quantity of substance

If operation history data are simply deleted to reduce traffic, then communication traffic is reduced, but data needed for generating learning model become insufficient

Engineering Contradiction:
Improvecommunication trafficVSAvoiddata sufficiency
Core Design Contradiction:
Quantity of substanceVSLoss of information

Solution Approach 1:

The patent performs preliminary action by pre-processing operation history data before transmission to identify and retain only the essential information needed for learning model generation. The system analyzes data characteristics in advance, determines which data items are necessary for maintaining data sufficiency, and prepares an optimized data set for transmission. This preliminary filtering ensures that communication traffic is reduced while the remaining data remains sufficient for generating accurate learning models.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3913450B1Air conditioner
Publication Date: 2024.09.18 FUJITSU GENERAL LTD
  • EP3913450B1 patent drawingFigure 1
  • EP3913450B1 patent drawingFigure 2
  • EP3913450B1 patent drawingFigure 3

AI summary

An air conditioner (2) comprises a control unit (2B,14) that controls the air conditioner using a learning model; and an adapter (3) including a communication unit that communicates with a server device (5) that generates the learning model on the basis of operation history data of the air conditioner. The adapter (3) includes an acquisition unit (14A), a determination unit (14B), an erasing unit (14C), and a transmission unit (14D). The acquisition unit (14A) acquires the operation history data every predetermined cycle from the air conditioner. The determination unit (14B) determines whether an amount of change between pieces of temporally continuous data out of a plurality of pieces of the operation history data acquired by the acquisition unit, the pieces of the operation history data having different contents from each other, is within a predetermined range that is defined corresponding to a content of a piece of the operation history data. The erasing unit (14C) leaves at least one of the pieces of continuous data and deletes the other pieces of continuous data in a case where the amount of change between the pieces of the continuous data is within the predetermined range. The transmission unit (14D) transmits the operation history data after the other pieces of continuous data is deleted by the erasing unit to the server device.