When the temperature of the screw extruder in the polyester spunbond production line fluctuates by 0.5 ℃, it may cause spinning breakage; When the impurity content of recycled PET slices increases by 0.1%, the final product qualification rate will decrease by 3% – the traditional production mode that relies on manual experience is already unable to meet the extreme requirements for accuracy and stability of high-end polyester spunbond materials (such as new energy vehicle battery cloth and medical sterile protective cloth). With the maturity of big data collection technology and AI algorithms, the polyester spunbond production line is shifting from “passive adjustment” to “active prediction”, upgrading from “manual inspection” to “intelligent diagnosis”. By 2025, leading domestic enterprises such as Hengtian Jiahua and Jiangsu Jinwang’s intelligent production lines have achieved significant results in reducing wire breakage rates by 60%, energy consumption by 18%, and product qualification rates exceeding 99%, confirming the transformative power of “data-driven production”. This intelligent transformation is not only a technological upgrade, but also the core path to reconstruct the “efficiency quality cost” triangle relationship in the polyester spunbond industry.
Full process data collection: intelligent “perception neural”
The entire process of “melting spinning forming reinforcement” in polyester spunbond production involves over 200 key process parameters (such as melt temperature, spinning speed, hot air pressure) and over 100 quality indicators (such as fracture strength, air permeability, defect rate). The “complete, accurate, and fast” data collection is the foundation of AI applications. The current mainstream production line has established a “three-layer perception system” to achieve full chain data coverage from raw materials to finished products.
（1） Raw material side: dynamic traceability and formula optimization
The molecular weight distribution (MWD), impurity content, and moisture content of regenerated PET slices directly affect spinning stability. By using near-infrared spectroscopy (NIRS) sensors and laser particle size analyzers, key parameters of each batch of raw materials can be collected in real time:
Data dimensions: molecular weight (1.8-2.5 × 10 ⁴ g/mol), impurity particle size (0-50 μ m), moisture content (0.005% -0.05%), hue L value (75-88);
Transmission method: Adopting OPC UA protocol, data is uploaded in real-time to the Industrial Internet of Things (IIoT) platform with a delay of less than 100ms
Application value: The big data platform can establish a correlation model between “raw material parameters spinning process finished product quality”. For example, when the moisture content of recycled slices is greater than 0.02%, it automatically pushes process adjustment suggestions of “screw temperature increase of 5 ℃+vacuum degree increase of 0.02MPa” to avoid spinning bubbles.
The practice of Hengtian Jiahua shows that the system improves the efficiency of raw material adaptation by 40%, increases the maximum mixing ratio of recycled materials from 50% to 70%, and does not reduce the strength of the finished product (still maintaining above 28N/5cm).
（2） Production end: real-time monitoring and abnormal warning
Spinning and web forming are the most prone to problems, and high-density sensors need to be deployed to achieve “second level monitoring”:
Spinning process: Install platinum resistance temperature sensors (accuracy ± 0.1 ℃) in each section of the screw extruder, and install melt pressure sensors (range 0-50MPa) at the outlet of the spinning component to capture temperature fluctuations and pressure peaks in real time. When the pressure suddenly rises by 10%, the system determines it as a “component blockage risk” and immediately triggers a warning;
Networking process: Laser thickness gauge (accuracy ± 1 μ m) is used to scan the thickness of the mesh surface, collecting 2000 data points per square meter. Combined with airflow distribution sensors (monitoring hot air speed and temperature uniformity), a three-dimensional model of mesh uniformity is constructed;
Reinforcement process: Install infrared thermometers (resolution 0.5 ℃) on the rollers of the hot rolling mill to monitor the bonding temperature in real time and avoid mesh adhesion caused by local overheating.
These data are preprocessed by 5G edge computing nodes and transmitted to the big data platform to form a real-time mapping of “process parameters – equipment status – product form”, providing dynamic data support for AI regulation.
（3） Finished product end: full inspection traceability and closed-loop feedback
Traditional manual sampling inspection (3 rolls per batch, detection rate<0.5%) is prone to missing small defects (such as holes with a diameter of 0.5mm), while AI vision inspection systems achieve “100% full inspection”:
Hardware configuration: Line array camera (resolution 4096 pixels)+LED strip light source, shooting speed synchronized with the production line (up to 60m/min);
Algorithm capability: Based on a convolutional neural network (CNN) defect recognition model, it can distinguish 12 types of defects including holes, hairs, and uneven thickness, with an accuracy rate of 99.5% and a false detection rate of less than 0.1%;
Closed loop mechanism: The detection data is automatically associated with the preceding process parameters, for example, “hole defects are concentrated in the third spinning position”. The system traces the melt temperature curve of this station and finds “temperature fluctuation ± 1.2 ℃”, immediately pushing parameter calibration instructions.
After Jiangsu Jinwang applied the system, the missed inspection rate of finished products decreased from 5% to 0.3%, and the number of customer complaints decreased by 70%.
Core AI Applications: From “Passive Repair” to “Active Optimization”
Based on the entire process data, AI algorithms achieve three core functions in polyester spunbond production: parameter prediction and regulation, equipment fault diagnosis, and quality closed-loop optimization, directly solving the pain points of “lag” and “experience dependence” in traditional production.
（1） AI prediction of spinning parameters: key to a 60% decrease in yarn breakage rate
Spinning wire breakage is a chronic problem in the industry, which traditionally relies on operators to adjust by “looking at the wire shape and listening to the sound”, resulting in delayed response and low accuracy. The AI prediction model achieves “early intervention” by analyzing historical data (spinning parameters and breakage records of 3 years, 100000+batches):
Model architecture: Using Long Short Term Memory (LSTM) network, input features include 18 dimensions such as melt temperature, pressure, spinning speed, and slice moisture content, and output the “probability of fiber breakage in the next 10 minutes”;
Control logic: When the probability of wire breakage is greater than 15%, the system automatically adjusts parameters – for example, when the melt temperature deviation is 0.8 ℃, the screw temperature compensation is 0.5 ℃; When the spinning speed fluctuates by 5%, the stretching airflow speed is synchronously adjusted by 3%;
Actual effect: After the application of Hengtian Jiahua’s production line, the wire breakage rate decreased from 0.5 times/ton to 0.2 times/ton, and the downtime loss caused by each wire breakage (about 20000 yuan/time) was reduced by 60%, saving more than 1.2 million yuan in annual costs.
More importantly, the AI model can adapt to different types of raw materials – when switching between regenerated PET slices (with an impurity content of 0.8%), the model automatically adjusts the parameter weights, avoiding manual process exploration, and shortening the material replacement and debugging time from 8 hours to 2 hours.
（2） Equipment malfunction AI diagnosis: reduce operation and maintenance costs by 30%
The failure of key equipment in the polyester spunbond production line, such as screw extruders and hot rolling machines, may result in several hours of downtime. The AI diagnostic model achieves “early warning” and “precise positioning” by analyzing equipment vibration, current, temperature and other data:
Diagnosis of screw extruder: Collect motor current (fluctuation range ± 5A) and bearing vibration (frequency 10-1000Hz) data, extract fault features through wavelet transform – when the bearing wears out, the vibration signal peaks at 250Hz, and AI provides a 72 hour early warning to avoid “sudden jamming”;
Hot rolling mill diagnosis: Analyze the temperature distribution of the roller (normal deviation ± 2 ℃). If the local temperature rises abnormally by 5 ℃, AI determines it as “heat transfer oil blockage” and locates the blockage location (accurate to 10cm range). The maintenance time is shortened from 4 hours to 1.5 hours;
Predictive maintenance: Based on equipment operation data, AI generates a “personalized maintenance plan” – for example, if the vibration value of a certain stretching fan increases by 0.2mm/s per month, the model recommends “replacing bearings every 3 months” instead of the traditional “annual replacement” to reduce excessive maintenance costs.
According to industry statistics, after applying AI diagnostic systems, the mean time between failures (MTBF) of equipment has been extended by 40%, operation and maintenance costs have been reduced by 30%, and downtime losses have been reduced by 50%.
（3） Quality AI Optimization: Multi Objective Balanced ‘Intelligent Formula’
Polyester spunbond products need to consider multiple quality indicators (such as strength, breathability, weather resistance), and traditional manual adjustment is difficult to achieve “multi-objective optimization”. The AI optimization model generates the optimal combination of process parameters through multi-objective genetic algorithm:
Case 1: New energy vehicle battery cloth: The requirement is “fracture strength ≥ 32N/5cm+air permeability ≥ 800L/m ² · s+flame retardant grade V-0″. The AI model analyzed 5000+sets of experimental data and output parameters: spinning temperature 275 ℃, stretching speed 1200m/min, hot rolling temperature 155 ℃. The finished product qualification rate increased from 88% to 99%;
Case 2: Recycled PET packaging fabric: The requirement is “minimum cost+strength ≥ 20N/5cm”. The AI model calculates the optimal combination of recycled material blending ratio (70%) and spinning speed (900m/min), reducing the cost by 800 yuan per ton while meeting the strength requirements;
Dynamic adjustment: When customer demand changes (such as air permeability increasing from 800 to 1000L/m ² · s), AI recalculates parameters within 10 minutes without the need for manual trial production, resulting in an 80% increase in response speed.
Core Value: Triple Breakthrough of Efficiency, Quality, and Cost
The application of big data and AI in the polyester spunbond production line is not simply about “showing off technology”, but about realizing the practical value of “efficiency improvement, quality upgrade, and cost reduction” through data-driven approaches, especially suitable for the strict requirements of high-end application scenarios such as medical and new energy vehicles.
（1） Production efficiency: 50% increase in per capita output value
Reduced downtime: downtime caused by wire breakage and equipment failure has decreased from 8 hours/month to 3 hours/month, and production line utilization has increased from 85% to 95%;
Labor cost reduction: AI replaces 60% of manual inspections (such as parameter monitoring and defect detection). For a production line with an annual output of 20000 tons, the number of operators is reduced from 15 to 9, and the per capita output value is increased from 13 million yuan/year to 20 million yuan/year;
Accelerated production change speed: When switching between multiple varieties, the process debugging time has been shortened from 8 hours/time to 2 hours/time, and the response cycle for small batch and customized orders has been shortened from 7 days to 3 days, meeting the supply chain demand for “multiple models and small batches” of new energy vehicles.
（2） Product quality: the key to breakthroughs in high-end development
Stability improvement: The fluctuation range of finished product strength has been reduced from ± 15% to ± 5%, and the fluctuation of air permeability has been reduced from ± 20% to ± 8%, meeting the strict standards of medical protective cloth (GB 19082) and battery cloth (ISO 21469);
Function expansion: AI optimized “flame retardant+weather resistant” modification process has stabilized the limiting oxygen index (LOI) of polyester spunbonded fabric at more than 32%, and extended the outdoor anti-aging life from 3 years to 5 years, successfully entering the field of photovoltaic backplane substrate;
Customization capability: By quickly generating process solutions through AI, full range customization of “weight 10-200g/㎡, thickness 0.1-0.5mm” can be achieved, responding to the material requirements of different parts of new energy vehicle battery packs (such as ultra-thin insulation layers and thick buffer layers).
（3） Cost optimization: Reduce costs by 15% across the entire chain
Raw material cost: Increasing the blending ratio of recycled materials by 20% reduces the cost of raw materials by 1200 yuan per ton; The AI optimized formula reduces the usage of functional additives (such as flame retardants) by 10%, saving 500000 yuan in additive costs annually;
Energy consumption cost: The AI controlled screw temperature and hot air system reduces unit energy consumption from 800 degrees/ton to 680 degrees/ton, saving 480000 yuan in annual electricity bills (0.6 yuan/kWh for industrial electricity);
Waste cost: The qualified rate of finished products has increased by 4%, the loss of each ton of waste (about 5000 yuan/ton) has been reduced by 200 yuan, and the annual cost savings of waste are 400000 yuan (calculated based on a production capacity of 20000 tons).
Challenges and Future Trends
Despite the significant achievements of intelligent applications, the polyester spunbond industry still faces three major challenges: “data silos,” “model generalization,” and “input costs.” At the same time, it will upgrade towards “digital twins,” “edge intelligence,” and “green collaboration” in the future.
（1） Current challenge: the “roadblock” to the implementation of intelligence
Data islanding issue: Some equipment on old production lines (such as old extruders) do not have data interfaces and cannot be connected to the IIoT platform, requiring additional renovation (the cost of renovating a single line is about 2 million yuan); The protocols of different device manufacturers are incompatible (such as Siemens PROFINET and Schneider Modbus), making data integration difficult;
Model generalization ability: The AI model performs well on specific raw materials (such as recycled PET slices from a certain brand), but when switching to a new supplier, the accuracy decreases by 15% -20%, and the model needs to be retrained (taking 1-2 weeks);
Investment return cycle: The initial investment (sensors, AI systems, platform construction) for an intelligent production line is about 8 million yuan, and the return cycle for small and medium-sized manufacturers (annual production capacity<10000 tons) is as long as 3-5 years, which restricts the speed of popularization.
（2） Future trend: from “local intelligence” to “global collaboration”
Digital Twin Factory: Build a virtual image of the production line to achieve a closed loop of “virtual debugging actual production data feedback” – simulate the impact of raw material changes and equipment failures in the virtual environment, optimize processes in advance, and reduce downtime debugging time by another 50%;
Edge intelligent deployment: deploy AI models at edge computing nodes (such as industrial PCs at spinning stations), reduce data processing delay from 100ms to 10ms, achieve “microsecond level regulation”, and adapt to the high-precision requirements of fine denier fiber (below 1dtex) spinning;
Green intelligent collaboration: AI combines carbon footprint data to optimize “low-carbon processes” – for example, when recycled materials are mixed at 70%, carbon emissions are 35% lower than raw materials, and AI automatically prioritizes the formula to help enterprises meet their “dual carbon” goals;
Collaborative intelligence of industrial chain: Upstream PET slice manufacturers share raw material data with downstream spunbond enterprises, and AI models predict raw material characteristics in advance, achieving seamless connection between “raw material production” and approaching zero material change and debugging time.
Conclusion: Intelligent reconstruction of the competitiveness of polyester spunbond industry
The application of big data and AI in polyester spunbond production lines has changed from an “optional” to a “mandatory” option – when new energy vehicles require “zero defects” in battery fabrics, when medical protection requires “full process traceability”, and when the circular economy forces “efficient utilization of recycled materials”, only data-driven intelligent production can achieve a coordinated breakthrough in “precision, efficiency, and environmental protection”.
The practice of leading domestic enterprises has proven that intelligence can not only solve the pain points of traditional production, but also open the door to the high-end market – Hengtian Jiahua’s AI optimized production line, whose products have entered the Tesla battery pack supply chain; Jiangsu Jinwang’s intelligent detection system has enabled medical fabrics to pass the EU CE certification, increasing export volume by 60%. In the future, with the maturity of digital twin and edge computing technologies, and the deepening of industrial chain data collaboration, the polyester spunbond industry will truly realize “intelligent manufacturing 2.0″, shift from “scale dividend” to “technology dividend”, and take the initiative in global competition.
The core of this intelligent transformation is not only the upgrading of technology, but also the reconstruction of production concepts – from “experience driven” to “data-driven”, from “single machine optimization” to “full chain collaboration”, from “product manufacturing” to “value creation”. The intelligence of polyester spunbond production lines will also become a “benchmark sample” for intelligent manufacturing in the polymer materials industry, providing replicable experience for other non-woven fabric categories such as needle punched and hydroentangled.
Dongguan Liansheng Non woven Technology Co., Ltd. was established in May 2020. It is a large-scale non-woven fabric production enterprise integrating research and development, production, and sales. It can produce various colors of PP spunbond non-woven fabrics with a width of less than 3.2 meters from 9 grams to 300 grams.
Post time: Sep-16-2025