Introduction
When a mill has to switch quickly between smaller lots and more fabric variations, every minute saved in setup and every defect avoided matters. Woven fabric production has already moved well beyond purely mechanical looms. Warping, sizing, weaving, and even fault detection now use automation in different forms. The next step is a weaving system that can measure its own behavior, adjust settings, and keep production stable with less operator intervention.
Automation in Woven Fabric Production
Weaving remains the most widely used method for fabric production, and it has also become one of the most automated areas in textile manufacturing. The industry has adopted automatic shuttle and shuttleless looms such as rapier, air-jet, and projectile looms. Computer technology has also made automatic fault detection, design support, and artificial intelligence part of the weaving floor.
In practice, automation in woven fabric production can be linked either to machine operations or to material handling. That division is useful because it shows how wide the subject really is. Some automation steps deal with how the machine runs, while others deal with how yarn, beams, and fabric are moved and monitored.
Automation in Warping and Sizing
Warping has seen automation in automatic section positioning, pneumatic stop brakes, automatic warp beam loading, doffing and chucking, sensors for machine stop, and automatic creel movement. These functions reduce the amount of manual intervention needed when warp preparation is changed or interrupted.
Sizing has also been automated in several ways. Automatic yarn tension control, automatic humidity and temperature control during sizing, and automatic machine stop in case of fault detection all help keep the process more consistent. This is one of those areas where textile automation is often underestimated in practice, because a steady sizing process has a direct effect on later weaving performance.
Automation in Weaving
Weaving itself has also moved toward more automated control. Automatic detection of warp breakage, automatic pick repair, automatic color selection, and automatic fabric fault detection are all part of modern weaving automation. Even with these improvements, warp beam change and mending of warp threads are still done manually in many cases. That remaining manual work is important to remember, because it shows that the shift toward full automation is still in progress.
Cyber-Physical Production Systems in Smart Weaving
A more advanced direction in textile automation comes from cyber-physical production systems, or CPPS. In these systems, physical objects and information-processing systems work together. When intelligent, self-optimizing CPPS are built into process chains, productivity and product quality can increase.
This matters especially for textile producers in high-wage countries, where smaller lot sizes and growing product variation create constant pressure on efficiency. One way to respond is through CPPS and cognitive machinery. Current approaches focus on intelligent algorithms inside machine control systems so that the production process can optimize itself instead of relying only on operator experience.
Multiobjective Self-Optimization of the Weaving Process
Why Self-Optimization Matters
Weaving is one of the oldest and most common fabric manufacturing processes. It has been used since about 4000 BC, by crossing warp and weft threads at right angles. Today, woven fabrics are used in apparel such as jeans and lining fabrics, in geotextiles for erosion protection and soil reinforcement, and in technical textiles such as fireproof fabrics, airbag fabrics, and reinforcements for fiber composites.
Because production costs are lower in some regions, much textile manufacturing has shifted to Asian countries. At the same time, high-quality and technical textile production has increasingly moved to Europe and other developed countries, including developing countries. High-wage countries such as Germany face a different challenge. Small lot sizes mean shorter cycle times, more frequent fabric changes, and more pressure to produce economically.
A fabric change often means that the weaving machine has to be reconfigured. A machine with about 200 parameters may need a new setup after every fabric change. To find the best configuration, operators often carry out weaving trials. Depending on the operator’s experience, these trials can require weaving up to 120 meters of fabric before the best settings are found.
Self-optimization systems are designed to adapt their inner state or structure when input conditions or disturbances change. In this case, target values can include capacity, lot size, quality, or energy consumption. Such systems usually follow three continuous steps: analyzing the actual situation, determining the targets, and adapting system behavior to reach those targets.
Measurement and Control Setup
The self-optimizing weaving concept was applied to the weaving process by using a programmable logic controller, or PLC, as the main control platform. For signal processing and the execution of the self-optimization routine, the ibaPADU-S module system was used. It included the ibaMS16xAI-20 mA analog input module, the ibaMS16xAl-10 V analog input module, and the ibaPADU-S-IT-16 central processing unit.
The analog input modules collect and process signals from the sensor system and are connected to the base unit through a back panel bus. The central unit is then connected to a computer through a TCP/IP interface. On the computer, the software ibaLogicV4 is installed. Together with the modular system, it forms a software-based PLC, or soft-PLC. The program for self-optimization is created on the computer and transmitted to the central unit through TCP/IP, where the runtime platform executes it.
Warp tension was measured with the yarn tension sensor TS44/A250. This sensor generates a voltage signal in the range of 0 to 10 V, proportional to the current yarn tension. It is placed in the middle of the weaving machine, between the back rest and the warp stop motion. Its data connector is linked to the soft-PLC through an analog/digital converter.
Air consumption was measured with the flow sensor SD8000. It generates a signal proportional to compressed air consumption in the range of 4 to 20 mA, and the output data are transferred wirelessly to the soft-PLC. The power measurement module collects characteristic values of the three-phase supply and stores them in the process image. To access those values, the module is connected to a fieldbus controller through a terminal bus. The fieldbus controller is then linked to the soft-PLC through TCP/IP, and communication takes place in Modbus format.
In that setup, the soft-PLC acts as the Modbus master and sends a request to the fieldbus controller, which acts as the Modbus slave. The controller identifies the requested value in the process image of the power measurement module and returns it in the response. Once the response reaches the soft-PLC, the data are available for signal processing. The active power is then requested from the process image in the same way.
Online Fabric Fault Detection
At ITA in Aachen, a measuring system for online error detection during fabric production was developed. A camera takes pictures of the fabric and the images are checked for defects through digital image processing. The software runs on a separate computer, and the camera is installed above the part of the weaving machine where the fabric is being produced.
The system can detect defects immediately after the fabric is produced. It is calibrated using a flawless piece of fabric, and any deviation from that calibrated condition is classified as a defect. Depending on the share of incorrect pixels in the image, the fabric is assigned to a quality category.
Defect detection is carried out in real time during the weaving process. The computer running the image processing software is connected to the soft-PLC through TCP/IP, and the quality category is transmitted continuously. A category of 0 means the fabric quality is accurate, while a category of 4 means the fabric is destroyed.
How the Optimization Routine Works
The self-optimization program for the weaving process works in a clear sequence. First, an experimental design is calculated automatically. In this design, the three setting parameters, static warp tension, vertical position of the warp stop motion, and revolutions per minute, are varied. The user sets the parameter spaces so the algorithm stays within a feasible range.
Next comes the test procedure. The weaving machine sets up every test point, and sensor data that describe the objective functions are recorded for each parameter setting. In the third step, the collected data are used to calculate three regression models, one for each objective function, and each model describes the objective function in relation to the setting parameters.
In the final step, an optimized setup of the weaving machine is calculated using desirability functions and a numerical optimization algorithm. Before the optimization is executed, user-defined preferences for warp tension, energy consumption, and fabric quality can be included through target weights. The preference scale for each objective function is divided into three sections, low, middle, and high.
The communication between the weaving machine and the soft-PLC is continuous, so the machine can run the entire self-optimization program autonomously.
Desirability Functions and the Nelder-Mead Algorithm
Desirability functions are used to combine several objective functions into one common function. The idea is simple enough, even if the implementation is technical. For each objective function, a desirability value is defined on a scale from 0 to 1. If the desirability is zero, the result is invalid for the optimization routine. If it is one, the result is optimal.
The total desirability is calculated as the geometric mean of the individual desirabilities. This total value also falls between 0 and 1. A total desirability of 1 means that all target functions are in the optimal range. If even one target function has an invalid value, the total desirability becomes 0.
The process parameters that produce the highest total desirability define the optimal operating point for the weaving process. Numerical algorithms are well suited for finding that point, and they are more efficient than grid search methods. The Nelder-Mead algorithm is often used for this purpose. It searches for a combination of the three parameters, basic warp tension, revolutions per minute, and vertical warp stop motion position, that maximizes the total desirability.
The algorithm starts with initial values, which create the starting point. From there, the parameter combinations move toward the optimal values during the iterations. In a minimization problem with m parameters, the algorithm considers m + 1 parameter combinations. Their function values are calculated, sorted, and the worst point is replaced in the next iteration through the standard operations of the algorithm.
Validation in Long-Term Testing
The self-optimization program was validated in a long-term test in the laboratory of ITA under industrial conditions. The test lasted 8 hours, matching a normal shift duration. It was carried out with an air-jet weaving machine, the OmniPlus 800 by Picanol, using polyester filament yarn of 330 dtex for both warp and weft in a twill 3/1 pattern.
After execution of the self-optimization program, the algorithm calculated the following optimal settings: basic warp tension of 3.71 kN, revolutions per minute of 522, and vertical warp stop motion position of 20 mm. These settings were compared with reference settings from an industrial weaving mill that processed the same material: basic warp tension of 4 kN, revolutions per minute of 900, and vertical warp stop motion position of 0 mm.
During the long-term test, the following process data were recorded: efficiency of the weaving machine, meaning the amount of fabric or weft insertions, and warp or weft defects and breakages. Data from the objective functions were also recorded with the ibaPDA software. The measured values were plotted over the main shaft position of the weaving machine, which is the rotating angle of the machine’s main drive.
The self-optimization program enabled the weaving machine to find an operating point that improved all objective functions compared with the conventional reference settings. Efficiency was higher with the optimal setting, mainly because machine downtime was reduced. With the optimal machine settings, two weft defects occurred because the weft threads collided with sagging warp threads. With the suboptimal settings, six weft defects occurred.
The machine also ran more stably at the lower speed. The higher number of weft defects at 900 rpm can be explained by the less favorable machine speed. These defects result from faulty transport of the weft threads across the width of the weaving machine.
The setup advantage was equally clear. Without self-optimization, a machine operator needed about 120 minutes to configure the weaving machine and find suitable process settings. The self-optimization program completed that task in 30 minutes, reducing setup time by 75%.
Conclusion
Automation in woven fabric production is no longer limited to faster looms. It now covers warping, sizing, weaving, fault detection, and process control through data-driven systems. The self-optimizing weaving example shows that a machine can measure its own behavior, adjust key parameters, and improve efficiency, quality, and energy use at the same time. Just as important, it can cut setup time sharply when lot sizes are small and fabric changes are frequent. As product variation keeps growing, smart weaving systems are likely to become even more practical on the production floor.
