Data-based color defect cause analysis and quality assurance AI solution

This is Impix's AI solution that predictS color quality and reduceS defects in the textile manufacturing process.

1. Pain Point

Industry Characteristics

- Need to improve quality and yield to strengthen global competitiveness in the textile and artificial leather manufacturing industry

- Cost delays due to waste and re-production when color deviations and defects occur in the dyeing process

- Inefficiency in process operations due to reliance on the experience and subjective judgment of skilled workers

Data and System Aspects

- Although data such as MES and CCM exists, there is no system in place to utilize it for AI-based analysis and quality prediction.

- There is no system in place for analyzing and predicting the causal relationships between quality-affecting variables and color defects in each process.

Equipment and Infrastructure Aspects

- Insufficient data standardization due to manual recording and collection of on-site data.

- Need to ensure consistency in color judgment and build an AI-based prediction system.

2. AI Solution

Application of A2LAB-based cause analysis and quality prediction AI solution

3. Construction Goals

Detailed Goals

- Improve yield and quality by minimizing color defects

- Establish a system for quality prediction and defect cause analysis through AI-based data analysis

- Transition to a data-based decision-making system and establishment of an intelligent smart factory

- Standardization of processes and ensuring consistent quality through on-site application

4. Construction Details

Application of data integration management technology

- Integration of heterogeneous data such as MES, CCM, and manual data, and construction of a dataset for AI analysis

- Verification of data consistency and improvement of data quality through preprocessing

Application of AI algorithm technology

- Comparison of XGBoost, LightGBM, and RandomForest algorithms, selection and application of optimal algorithms

- Construction of color judgment prediction models and defect detection models

- Analysis of the correlation between time-series variables and color quality, and analysis of the influence of key variables

Application of process quality prediction and simulation

- Visualization of prediction results and key variables on-site using BI-Tool and A2LAB visualization functions

- Color quality prediction simulation function based on virtual data

- Proactive quality management environment enabled through quality prediction prior to dyeing machine setting changes

5. Construction Effects

Reduction in dyeing correction rate

- 9.3% before introduction

- 6.1% after introduction

Reduction in annual dyeing quality loss costs