Neural Network Laundry Classification from Drum Acceleration Current

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional laundry treatment apparatuses face limitations in accurately sensing laundry weight and quality, requiring extensive time and expert settings, which increases energy consumption and total washing time due to imprecise measurement methods.

Innovation Solution

A laundry treatment apparatus utilizing machine learning and an artificial neural network to classify laundry based on current values obtained during motor rotation, allowing for rapid and accurate determination of laundry weight and quality, and adjusting operations accordingly.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional algorithms with experimental constants are used to sense laundry weight, then the sensing method can be implemented, but the accuracy in sensing laundry weight is limited and requires lots of time to find accurate setting values

Engineering Contradiction:
Improvelaundry weight sensing accuracyVSAvoidtime to find setting values
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces conventional mechanical/mathematical algorithms with neural network-based artificial intelligence. The neural network learns optimal sensing parameters automatically from training data, eliminating the need for manual experimental constant determination. This substitution enables accurate laundry weight sensing without requiring experts to spend extensive time finding optimal setting values, as the system self-optimizes through machine learning.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The neural network performs self-learning and self-optimization by automatically determining optimal sensing parameters from training data. The system improves its own performance without external intervention, continuously refining its weight sensing accuracy through accumulated experience. This self-service capability eliminates the need for manual adjustment of experimental constants and enables the system to adapt to different laundry conditions autonomously.

Inventive Principle:
Principle #25Self-service

2Measurement precision

If accurate laundry weight sensing is not achieved, then the spin-drying operation must be performed for lots of time to ensure proper drying, but this increases total washing time and energy consumption

Engineering Contradiction:
Improvelaundry weight sensing accuracyVSAvoidenergy consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent employs neural network-based AI to achieve precise laundry weight sensing, enabling the system to accurately determine the optimal spin-drying parameters. This precise measurement capability allows the washing machine to perform spin-drying operations efficiently without excessive time or energy consumption, as the neural network predicts the exact weight and recommends appropriate drying settings from the first attempt.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system uses the neural network's weight sensing output as feedback to optimize spin-drying operation parameters. By continuously monitoring the sensed weight and adjusting drying settings based on this feedback, the system achieves energy-efficient operation. The feedback mechanism ensures that the spin-drying process is precisely tailored to the actual laundry weight, preventing both under-drying and unnecessary energy consumption from over-drying.

Inventive Principle:
Principle #23Feedback

3Adaptability or versatility

If conventional comparison methods are used to sense laundry weight, then the system can distinguish between large and small laundry weight, but it cannot accurately sense various kinds of laundry weight

Engineering Contradiction:
Improvelaundry weight classification capabilityVSAvoidlaundry weight sensing accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent replaces simple comparison-based classification with neural network-based pattern recognition. The neural network analyzes multiple input parameters simultaneously and identifies complex patterns in laundry weight characteristics, enabling accurate classification across diverse laundry types. This AI approach provides both high adaptability to different laundry kinds and precise weight sensing, overcoming the limitations of conventional binary comparison methods.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system transitions from one-dimensional binary comparison (large vs. small) to multi-dimensional analysis by incorporating multiple input parameters and neural network features. This dimensional expansion enables the system to differentiate among various laundry types and weights with high precision, creating a comprehensive classification system that handles diverse washing scenarios effectively.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP3623513B1Artificial intelligence laundry treatment apparatus and method of controlling the same
Publication Date: 2022.03.30 LG ELECTRONICS INC
  • EP3623513B1 patent drawingFigure 1
  • EP3623513B1 patent drawingFigure 2
  • EP3623513B1 patent drawingFigure 3~4

AI summary

Disclosed is an artificial intelligence laundry treatment apparatus including a washing tub configured to receive laundry, the washing tub being configured to be rotatable, a motor configured to rotate the washing tub, a controller configured to control the motor such that the washing tub is rotated while being accelerated, and a current sensing unit configured to sense current of the motor, wherein the controller is configured to obtain laundry weight and laundry quality from output of an output layer of an artificial neural network pre-trained based on machine learning using a current value sensed by the current sensing unit during accelerated rotation of the washing tub as input of an input layer of the artificial neural network within a range within which the laundry moves in the washing tub.