OptiStrick: Detect Errors in Flat Knitting Earlier with AI-Powered Quality Assurance

October 20, 2026

Flat knitting machines can efficiently produce a wide variety of textiles. However, even minor defects on a knitting needle or contamination from fiber dust can cause missed or dropped stitches, thereby compromising the overall quality of the textile. If such defects are not detected until a later quality inspection, scrap may already have been produced.

Monitoring the knitting process directly on the machine

The OptiStrick research project investigated how the knitting process can be monitored directly at the machine. The goal was to develop an AI-supported system that detects defects in real time—that is, as soon as they occur. Unlike inspections of the finished textile, this approach focuses on the ongoing production process.

Camera-based systems, among other technologies, are used for quality assurance in the textile industry. While they can detect defects in textile surfaces, the project description notes that they often involve high investment costs and significant computational effort. OptiStrick therefore opted for a modular system that can be retrofitted and combines various sensor technologies with machine learning.

Sensors in Action: What the Tests Show

The Institute for Textile Machinery and High-Performance Textile Materials Technology (ITM) and the Fraunhofer Institute for Digital Media Technology (IDMT) investigated acoustic, vibroacoustic, strain gauge, and optical sensors. Acoustic sensors detect noise, while vibroacoustic sensors also detect vibrations. The measurement data was analyzed using neural networks.

The tests conducted in the laboratory and in industrial production revealed particular potential for acoustic and vibroacoustic sensors: They achieved accuracies of over 90 percent in detecting process states and needle defects. Strain gauge sensors also delivered good results but were more sensitive to changes in their position.

The developed demonstrator achieved an average accuracy of around 80 percent in the classification of process parameters, thereby confirming the fundamental practical feasibility of the approach. It is particularly noteworthy that the system has a modular design and is therefore relatively easy to retrofit. This allows even smaller knitting mills to benefit from AI-supported quality assurance without having to invest in costly camera technology.

Prospects for Maintenance and Other Applications

In addition to defect detection, the system should also identify wear at an early stage and provide indications of potential machine failures. This would allow for more proactive maintenance planning.

OptiStrick demonstrates that acoustic and vibroacoustic sensors hold great promise for quality assurance on flat knitting machines and can serve as a cost-effective alternative to traditional image processing systems. The developed concept helps reduce scrap, makes maintenance measures easier to plan, and strengthens the competitiveness of companies in the textile industry. Furthermore, the approach offers potential for use in other textile production processes.