Textile/fabric manufacturing has historically been a labour-intensive industry which is characterised by low-fixed capital investment; a wide range of product designs and, hence, input materials; variable production volumes; high competitiveness; and often high demand on product quality. Thanks to advancements in Artificial Intelligence, textile manufacturers can now improve efficiency and augment capabilities of employees, drawing significant business insights out of historical and real-time operational data. Though adoption of Artificial Intelligence in textile industry is still in early stages, we have listed down major use cases where such adoption has taken place:
Defects in fabric reduce the value of textile products. To counter this problem, Artificial Intelligence techniques such as Artificial Neural Network (ANN) are applied for defect identification in fabric inspection of textile industry. The images to be analysed are obtained from image acquisition system and saved in relevant standard format (JPEG, PNG etc.). Features are extracted from the acquired image and feature selection method is used to reduce the dimensionality of feature set by creating new feature set of smaller size that are a combination of old features. Multi Layer Back Propagation algorithm is used to train and test the ANN.
For a leading Indian foam manufacturer, Valiance developed a similar “Intelligent Defect Identification” platform where images of normal and abnormal foams were submitted to localized “learning service” to discern OK vs Not OK characteristics of parts or components of foam that meet quality specifications and those that don’t.
The importance of inspection to product quality is great, since defects can significantly reduce process (for instance, in fabrics by as much as 60%). In textile production, in-line inspection is a slow process owing to the slow roll of the fabric out of the weaving machine, rendering human inspector not cost effective.
Since fabric pattern can have multiple aspects such as weaving, knitting, braiding, finishing, and printing, replacing visual inspection with vision-based inspection might help Textile manufacturers avoid human fatigue and errors in the detection of novelties and defects. Eventually, it leads to save on costs and time taken for inspecting the quality of the final fabric end-product. Typically the manufacturer might install the camera-based inspection system in their factories and input a few hundred images of “good” final samples, and “bad” samples.
The platform learns the weaving pattern, yarn properties, colours and tolerable imperfections from these images. This training period could be of a couple of weeks post which the platform might potentially start detecting defects (like wrong knitting patterns) in the textile end-product, saving human effort of assessing hundreds of yards of material. Several challenges are inherent in inspecting fabric patterns, namely their complexity, variability and the sheer numbers of fabric types.
Color is very important aspect of products for customers. The appearance of a product is perceived to be related to its quality. Like other industries this is also important in Textile industry. The color of a product may be judged generally to be “acceptable” or “unsatisfactory”, or it may be judged in more detail to be “too light”, “too red” or “too blue”. Such judgements can be made visually or instrumentally based on a perceived difference between an ideal product standard and a sample. When this difference is quantified, tolerances can be established.
While traditional color tolerancing was done based on numeric descriptions of color through ”instrumental tolerancing systems”, that method generally had a lot of false positives compared with visual inspections, causing delays in the approval process because of the need for careful human intervention. To counter this problem, an AI enabled platform, similar to Defect Identification, can be developed that has Pass/Fail (P/F) feature to help improve the accuracy and efficiency of instrumental tolerance.
This platform can take into account historical data of visual inspection results from human operators while creating the tolerances. The system can then be tested for new batches to automatically set AI tolerances, training the system to determine which samples pass and fail.
Businesses looking to leverage AI to enhance quality, production and lower costs would require a large trove of existing data for AI applications to learn from and significant amount of time, costs and domain expertise for successful integration.
It seems evident that real-world AI applications in Textile industry are still at a nascent stage and as we move ahead 5 years, we expect that a strong ROI from current machine vision applications might encourage more enthusiasm and adoption for AI in general in Textile industry.