I. Core Breakthroughs

Breakthrough 1 : An innovative approach quantifies the contribution weights of four major sources of variation to gap overlap, thereby identifying the primary root cause of defects.

Previous studies have analyzed only a single type of error in isolation, making it impossible to compare the relative magnitudes of various sources of variability. In this paper, under standardized experimental conditions, we simultaneously measured four types of errors and, using measures such as standard deviation and fluctuation ranges, … 99% Three-dimensional quantitative comparison of quantile bias reveals that the lateral slippage of the tow on the compaction roller is due to clearance. /  The primary cause of overlap is a factor whose weight far exceeds that of robot positioning and the inherent width deviation of the fiber tow, thereby correcting the industry’s widespread misconception that robot accuracy is the core bottleneck.

Breakthrough 2 : Establishing a standardized in-situ testing platform for multi-sensor synchronous measurements

Innovatively integrating three types of high-precision sensing devices into the placement head, the system simultaneously acquires the robot’s end-effector trajectory, the lateral position of the fiber tow between rollers, and the cross-sectional profile of the tow before and after compaction. This enables synchronous recording of positional and geometric errors within a single placement cycle, thereby mitigating data distortion caused by inconsistent test conditions in time‑division measurements and establishing a reproducible… AFP The quantitative testing paradigm for fluctuations.

Breakthrough 3 : Distinguishing the two fundamental underlying mechanisms of “positional fluctuations” and “geometric fluctuations”

Classification based on the essence of forming:

• Positional deviation: the robot’s trajectory is inaccurate, and the roller causes lateral displacement of the tow, thereby altering the transverse coordinate of the tow placement.
• Geometric deviations: uneven width of incoming material and deformation caused by compaction and extrusion, which alter the cross-sectional dimensions of the tow itself.

The coupling of two types of waves jointly generates a gap. /  Overlap—this classification system can be directly applied to modeling input errors in lay-up simulations.

Breakthrough 4 : Output a statistical benchmark database that can be directly used for process optimization.

For the four categories of web‑handling fluctuations—mean, standard deviation, and extreme‑value distribution ranges—the manufacturer can use these metrics to pinpoint equipment bottlenecks: excessive slippage should be addressed by optimizing tension and roller surface friction; width variation can be controlled by managing the incoming slit material; and robot drift can be mitigated through an enhanced spatial compensation algorithm, enabling targeted equipment upgrades.

II. Technological Innovation

Innovation 1  Multi-sensor Collaborative In-situ Measurement Hardware Solution

• Laser tracker ( Leica AT960 ): Millisecond-level data collection for the placement head TCP Three-dimensional coordinates, calculation And Deviation from the lateral trajectory, capturing the robot’s dynamic positioning error;
• 500  Megapixel industrial camera: Positioned directly facing the compaction roller zone, it uses grayscale image analysis to detect the lateral offset of the tow edge relative to the roller’s centerline, thereby quantifying the tow’s lateral slip.
• Double 2D Laser line scanning: LLS Pre-measure the original width of the incoming filament bundle, LLS Post‑compaction measurement of the cross‑section after molding, accurately capturing the width shrinkage caused by hot pressing and extrusion.
• 3D Print a lightweight sensor bracket, with modular disassembly and compatibility for mainstream platforms. AFP-XS Class-based placement head, with low hardware costs and easy retrofitting.

Innovation 2  Hierarchical Data Parsing Algorithm Framework

• Robot Error: Programming Theory And Statistical analysis of point-by-point differences between the calculated coordinates and the measured coordinates, with positioning dispersion differentiated across acceleration, constant-speed, and deceleration phases.
• Inter-roller slip: checkerboard‑calibrated pixels -  Millimeter conversion factor, gray-level thresholding to segment the tow boundaries, and output of the relative offset distribution with respect to the roll center;
• Tow width: laser profiling Z Highly accurate edge detection on both sides of the filament, automatic calculation of the cross-sectional total width, and filtering of dust and reflective noise.
• Statistical modeling: Fit frequency distributions to four types of bias, output the characteristic parameters of a normal distribution, and provide… 99% Engineering limit deviations are used in defect-tolerance design.

Innovation 3  Quantitative Analysis Model of Defect Coupling Mechanism

A single fluctuation is insufficient to create an out-of-tolerance gap; only the cumulative effect of multiple errors can result in a scrap‑level defect. The study proposes a logic for estimating such cumulative effects: lateral slip. +  Robot‑induced bias in the same direction, when superimposed, will significantly widen the gap; incoming wide‑strip material. +  Compaction and swelling can easily lead to overlapping, providing data support for the ply‑placement path‑offset compensation algorithm, thereby reducing the reserved offset in ply placement and improving the uniformity of fiber volume fraction.

Innovation 4  Facing AFP Optimization Logic for Production Line Implementation

According to the volatility-weighted tiered optimization strategy:

• Level‑1 optimization (highest priority): Increase tow tension, optimize the surface roughness of the press rollers, and suppress lateral slippage.
• Secondary optimization: full‑area error compensation in the robot’s workspace and dynamic velocity regulation, reducing trajectory drift during acceleration and deceleration.
• Three-tier control: At the upstream slitting process, tightly manage the tow width tolerance and maintain stable compaction temperature and pressure to minimize deformation and variability.

III. Application Scenarios

• Aviation AFP Large fuselage / Optimization of the Wing Box Mass-Production Line (Core Scenario)

Wide-body airliner fuselage skin and wing panels AFP On the production line, by benchmarking against the fluctuation data presented in this paper to pinpoint the root causes of line defects, targeted adjustments were made to tow braking tension and laying speed, thereby reducing rework hours associated with gap‑overlap issues and improving equipment utilization.

• AFP Development and Iteration of Head-Placement Equipment

Equipment manufacturers are developing the next generation. AFP The actuator can optimize the feed rollers and guide‑slot geometry based on slip‑wave metrics, thereby reducing lateral yarn‑bundle movement and minimizing molding defects at the hardware level.

• Routing Path Simulation and Defect Tolerance Design

Composite material simulation and process planning engineers input four types of wave‑like statistical distributions into the numerical model, enabling precise prediction of the probability of gap overlap along different ply‑layup paths, optimizing inter‑tow offset compensation values, and balancing fiber uniformity with defect risk.

• AFP Algorithm Development for Online Monitoring Systems

Laser Online Inspection AI The model training benchmark dataset distinguishes between robot drift, silk‑like displacement, and variations in material width, enabling automated classification and alerting of defect root causes rather than merely detecting gap sizes.

• Universities and colleges / Standard Test Protocol for the Enterprise Composite Materials Process Laboratory

Build AFP The performance calibration platform directly reuses the multi-sensor testing procedure described in this paper, enabling standardized comparisons of fluctuation levels under different placement heads and process parameters, thereby establishing a unified cross‑platform benchmark.

IV. Summary of the Article

(1) Core Research Findings

• generate AFP The four major sources of manufacturing variability—gap, overlap, robot trajectory error, lateral slippage of the tow on the pressure rollers, variations in incoming material width, and compaction-induced deformation—account for significant defects. Among these, lateral slippage of the tow between the rollers is the leading cause of defects. 99% The limit deviation reaches 0.534mm , accounting for the standard tow width 16.4% ；
• Geometric-type fluctuations (incoming material /  The compaction width exhibits smaller variability, resulting in a weaker impact on molding consistency compared to position‑related fluctuations.
• Under the condition of straight, uniform-speed laying, the robot’s positioning error exhibits a bimodal distribution and shows a systematic unidirectional bias, which is correlated with the fixture plane and the robot’s pose.
• The four types of fluctuations exhibit a superimposed amplification effect; when deviations in the same direction accumulate, they can give rise to severe gaps that exceed aviation acceptance criteria. /  Overlapping defect.

(II) Engineering Application Value

For a long time, aviation AFP There is no quantitative traceability mechanism for addressing defects on the production line, making it impossible to determine whether the issue stems from equipment, incoming materials, or the process itself. This paper presents a comprehensive multi-sensor quantitative testing methodology and benchmark data, enabling precise defect localization. By prioritizing optimization of tow tension and roller‑system design, the defect rate can be significantly reduced, leading to substantial cost savings. 35% rework downtime, thereby enhancing the mass-production efficiency of large composite components. Meanwhile, for path‑compensation algorithms and online monitoring… AI The model provides empirically measured statistical inputs, supporting… AFP Implementation of digital closed-loop quality control.

(3) Current Research Limitations

Only the straight‑line unidirectional lay‑up test has been completed; curved‑surface redirection and large‑angle steering conditions have not been verified (steering exacerbates tow slippage); only a single ply specification was tested. 6.35mm Thermoset prepreg filament, unexpanded 12.7/25.4mm Wide‑tape and thermoplastic tape‑layup processes; a quantitative relationship describing the degradation of mechanical properties under multi‑field coupling has not yet been established.

(4) Directions for Future Expansion

• Conduct angle‑varying ply placement tests on hyperbolic molds to quantify the amplification of tow slippage under steering conditions.
• Comparative tests of multi-specification tow and thermoplastic in‑situ laying were conducted to establish a universal waviness database.
• Coupling mechanics test, constructing “wave amplitude” -  Clearance dimension -  “Component strength reduction” quantification curve;
• Develop a real-time slip-feedback closed-loop control system for dynamic correction during the lay-up process. And Laying-out coordinates.

Image and Text Guide

Figure 1： AFP Schematic cross-section of a tape laminate structure —— Ideal AFP The layered structure of the composite material comprises adjacent tapes (left), a gap (middle), and an overlapping region (right).

Figure 2 : Gaps and overlapping areas caused by movement along the traction path

Figure 3： AFP Schematic cross-section of a tape laminate structure —— Ideal tape placement (left), tape layout under different positioning conditions (middle), tape layout for various geometric shapes (right)

Figure 4 : Causing gap and overlap defects AFP Manufacturing Process Variation Sub-Analysis

Figure 5 : Display AFP With multiple sensor configurations on the head CAD Model. The stacking direction is positive. X Axial direction.

Figure 6： AFP Plan Details

Figure 7 : Example frame images captured by the camera. The annotated portion shows the roller’s central axis and the measurement window (left image); the reference image includes a checkerboard scale (right image).

Figure 8 : Example profile obtained by two laser line scanners

Figure 9： A 、 B 、 C 、 D (From top to bottom) shows the manufacturing variations in measurements that cause gaps and overlaps. Please note. X Axis and Y The axes use the same scale; the mean and ± 1 The σ value is marked in orange; the nth 99 Percentile deviation scores are indicated by gray lines; the fitted curve is marked in yellow.

Table 1 Summary of Manufacturing Variations in Measurements Causing Gaps and Overlaps