Comprehensive assessment of fragmented fiber shedding from recycled cotton textiles. Part III integrated qualitative and quantitative assessment of released fibers and particles from woven textiles Authors: Gabriela Krupincová, Mercedes Pereira Published: 2026 ABSTRACT Fragmented fiber release from textiles represents a growing environmental concern, particularly for systems containing recycled cotton. This study presents the third integrative stage of a systematic investigation linking fiber quality, yarn parameters, and woven fabric construction to the emission of fiber fragments and particulate matter during repeated laundering. Three mechanically recycled cotton blends (B, C, D) were processed into 29.5 tex rotor-spun weft yarns and woven with high-quality Egyptian combed cotton warps under industrial conditions. For all yarn variants, two dye shades (light and dark) were applied to introduce controlled color patterning in warp and weft directions of the woven structure. Fabrics underwent five simulated laundering cycles using the AHIBA apparatus under intensified mechanical agitation conditions, and released contaminants were quantified using image analysis.The number and size of released impurities were evaluated. Feret-based descriptors were used to characterize particle types, classified into two subgroups (dust and trash). Fiber-type impurities were further categorized into individual fiber fragments and bundles. Fiber fragment length was evaluated, as well as the maximum representative lengths of fiber structures using skeleton-based analysis. Results indicate that weft yarn quality and yarn state (gray vs. dyed) affect the number and morphology of released impurities. Fabrics containing highly contaminated yarns (type B) showed the highest initial shedding, while subsequent cycles showed a decrease in readily detachable material and progressive fiber fatigue and fibrillation. Individual fiber fragments were categorized into ultrashort, short, and long subgroups. Statistically significant differences among experimental groups were confirmed by Kruskal–Wallis tests (p < 0.001).The estimated representative maximum lengths of fiber fragments (3,044 µm for l max branch and 261.75 µm for l fibre bundles) complete the morphological characterization of fiber shedding.A systems-level approach linking fiber quality, yarn parameters, and fabric construction enables a hierarchical assessment of fragmented fibers and particle release across fiber, yarn, and fabric levels within recycled cotton textile systems. Introduction Fragmented fiber pollution generated during textile production, use, and wet processing is increasingly recognized as a significant source of environmental contamination in aquatic and atmospheric ecosystems. Numerous studies have confirmed that fiber release occurs across a broad range of textile materials and processing conditions, leading to the development of various quantitative and morphological assessment approaches. Particular attention has recently been devoted to recycled cotton systems, where fiber fragmentation behavior is strongly influenced by the quality of recycled input material, spinning technology, and the resulting yarn structure. Previous studies demonstrated that the incorporation of pre-consumer cotton waste significantly affects the quality parameters of rotor-spun yarns as well as the structural and mechanical properties of woven fabrics produced from these yarns1,2. In addition, fabrics produced from recycled rotor-spun yarn systems exhibit increased surface contamination and higher release of particulate impurities during use and laundering2. The influence of spinning technology and yarn structure on fiber release behavior has also been reported in studies focused on both natural and synthetic fibers. Ishtiaque3 showed that fiber properties and waste generation vary during different stages of rotor spinning and differ between gray and dyed cotton systems. Similarly, Jabbar and Tausif4 demonstrated that yarn structure and spinning technology significantly affect the release of fragmented fibers during laundering. Comparable observations were reported for polyester yarn production, where fiber release and particulate contamination were linked to specific processing stages and manufacturing parameters5. Conventional assessment of released material is predominantly based on gravimetric analysis, which evaluates total mass loss after laundering or abrasion4,5. Although robust and widely used, this approach does not provide information on the morphology or composition of the released material. In cotton-based systems, its applicability is further limited by cellulosic fiber hygroscopicity and measurement uncertainty in small-scale experiments. To address these limitations, microscopy- and image-based approaches have been introduced for morphological characterization of released fibers and contaminants. However, most existing studies evaluate fiber release at the fabric or effluent level without considering the structural heterogeneity of textile systems. Consequently, the measured response reflects the combined effect of multiple overlapping release mechanisms originating from different yarn systems and structural levels of the textile. This limitation becomes particularly significant in recycled cotton materials, where fragmentation behavior is influenced not only by fabric construction, but also by yarn structure, spinning technology, fiber quality, inter-fiber friction behavior, and the previous mechanical processing history of recycled fibers3,4,6. In woven structures with asymmetric constructions or hybrid warp/weft configurations, the resulting fiber release, therefore, cannot be interpreted as a homogeneous material response, but rather as the combined behavior of structurally different yarn systems. Consequently, conventional fabric-level assessment alone does not enable clear identification of the mechanisms governing fiber release within complex recycled textile structures. To overcome the above-mentioned limitations, the present research was structured into three interconnected parts, each targeting a distinct hierarchical level of the textile system and fiber release mechanisms. Part 1 focused on verifying the processability of rotor-spun staple yarns containing high proportions of recycled cotton fibers into woven fabric structures under industrial conditions. Due to the reduced quality of recycled fibers processed by rotor spinning technology, a specific hybrid woven structure was designed combining different warp and weft yarn characteristics and varying sett densities in the main directions. The study investigated structural changes in woven fabrics, mass loss, and the intensity of fiber release during repeated laundering and drying cycles. The results demonstrated that the laundering process leads to significant structural stabilization of the woven plain weave textile, particularly during the initial washing cycles, while the amount of released material gradually reaches a stabilized state associated with fabric shrinkage stabilization. At the same time, it was confirmed that recycled cotton materials release not only fiber fragments, but also fibrillated fiber parts, residual plant impurities and dust particles originating either from the recycled fibrous raw material or generated during mechanical stress induced by repeated use, laundering, and drying. Although Part 1 provided quantitative information regarding the total amount of released material and the structural changes occurring in woven fabrics, gravimetric assessment at the fabric level does not allow unambiguous differentiation of the contribution of individual yarn systems nor detailed characterization of the morphology of released fragments. The measured values inherently represent the integral response of the combined behavior of warp and weft yarn systems within a complex woven structure. For this reason, Part 2 focused on the yarn level and introduced an adapted image-analysis-based methodology enabling detailed characterization of fiber fragments and particulate contaminants released from rotor-spun recycled cotton yarns under controlled conditions of a single washing cycle. The methodology enabled differentiation of individual types of released material, including single fibers, fiber bundles, trash, and dust particles, while providing information on length distribution and morphology of fiber fragments. The results of Part 2 demonstrated that fiber release behavior is significantly influenced by the quality of the recycled input material, the yarn state (gray/dyed), and the structure of the rotor-spun yarn itself. At the same time, it was confirmed that the yarns exhibit an inherent tendency toward fragmentation even before their integration into woven fabric structures. However, separate evaluation of woven fabrics (Part 1) and yarns (Part 2) alone does not sufficiently explain how the inherent fragmentation behavior of yarns is reflected in the resulting fiber release from real woven structures during repeated washing and drying. It therefore remains unclear to what extent the integration of yarns into woven fabrics modifies fragmentation mechanisms, alters the redistribution of released particles, or affects the morphology of the released material as a consequence of weave structure, sett density, and the mutual interaction of warp and weft systems. For this reason, Part 3 represents the integrative stage of the study, combining the quantitative assessment of structural changes and mass loss obtained in Part 1 with the detailed image-based characterization of released fibers and particulate contaminants adapted and set in Part 2. This part is organized to capture the complex interactions governing fiber release from textile structures and to contribute to the knowledge base by integrating results across different hierarchical levels of the textile system. It demonstrates that the quality of recycled cotton fibers, together with appropriately selected yarn parameters, including optimized spinning conditions at suitable twist levels, and fabric construction parameters, collectively determine both textile performance and the propensity for fiber release during repeated washing and drying cycles. Importantly, even recycled fibers can be processed into textiles of optimal functional quality while enabling systematic quantification of emitted fibers and impurities. Establishing clear links between fiber release behavior and material properties is therefore essential for a meaningful evaluation of the environmental footprint of textile products. Aim The third part of the interconnected research represents the final integrative stage of the study and provides a comprehensive, methodologically consistent evaluation of impurity release from woven fabrics produced from rotor-spun staple yarns incorporating repeatedly used cotton in the weft system. This part builds directly upon the outcomes of the preceding stages and systematically links their results into a unified analytical framework. Specifically, it integrates experimentally determined data on fabric structural changes and quantitative measurements of mass loss caused by repeated laundering obtained in Part 1. These results are combined with the advanced image-analysis methodology adapted at both warp- and weft-yarn levels during a single washing cycle in Part 2. This approach enables detailed characterization of released particles and fibres with respect to their distribution and morphology. Within this integrated methodological approach, the released impurities (fibre and particle types) analysed in Part 1 from the perspective of overall weight are subjected to detailed examination using the analytical procedures designed and verified in Part 2. The objective is to determine whether variations in weft yarns, mainly arising from differences in the quality parameters of recycled cotton fibres (fibre materials B, C, and D) and the state of the weft yarns (gray or dyed), significantly influence the composition and morphology of material released from the textile structure during repeated simulated laundering using the AHIBA instrument and drying. Particular emphasis is placed on evaluating these effects across five consecutive wash cycles to capture trends in contaminant release. Materials and methods The specification of samples, the procedure of simulated washing Only a concise overview of the samples and their processing route, from fibers to fabrics, is provided here, since detailed information and sample coding are given in Part 1 and summarized in Table 1. Three blends of mechanically recycled cotton waste (B, C, D) were processed to produce 29.5 tex (20 Ne) open-end rotor yarns for woven fabrics. Material B contained high levels of dust, trash, and short fibers, while C and D included 20% recovered fibers from gray or bleached fabrics, affecting dyeing behavior and yarn quality. The fibers were pre-processed, carded, and spun on a rotor Rieter R37 machine under the same conditions, yielding yarns suitable for home textiles and denim. Dyeing in a bobbin form with vat dyes produced generally acceptable color uniformity, despite variations due to fiber contamination, demonstrating the potential for sustainable yarn production from recycled cotton. For all yarn variants, two dye shades (light L, and dark D) were selected to evaluate the influence of yarn composition and impurity content on dyeing behavior and shade uniformity7. The selected shade combinations additionally enabled the development of various visual effects within the woven structure, including warp-dominant, weft-dominant, and combined color patterning effects.Table 1The description of samples.Material typeABCDType of yarnwarpweftFiber material blend used for weft yarn100% virgin cotton80% blowroomwaste 2/10% noils/ 10% virgin cotton/80% bowroom waste 3/ 20% post-production gray cotton waste80% bowroom waste 3/ 20% post-production bleached, optically brightened cotton wasteYarn count [tex]2 × 629.5State of the weft yarnlight colorgraylight colordark colorgraylight colordark colorgraylight colordark colorLegend of yarnAGBLBDBGCLCDCGDLDDD Table 1The description of samples.Material typeABCDType of yarnwarpweftFiber material blend used for weft yarn100% virgin cotton80% blowroomwaste 2/10% noils/ 10% virgin cotton/80% bowroom waste 3/ 20% post-production gray cotton waste80% bowroom waste 3/ 20% post-production bleached, optically brightened cotton wasteYarn count [tex]2 × 629.5State of the weft yarnlight colorgraylight colordark colorgraylight colordark colorgraylight colordark colorLegend of yarnAGBLBDBGCLCDCGDLDDD The description of samples. 80% blowroom waste 2/10% noils/ 10% virgin cotton/ Woven fabric samples were produced on an industrial SOMET rapier loom with a STÄUBLI jacquard (speed: 435 picks/min, reed width: 148 cm, hooks 1344) to evaluate recycled cotton yarns under realistic production conditions. The woven fabrics were designed with a heterogeneous (hybrid) structure combining high-quality warp yarns with recycled weft yarns in order to ensure sufficient processability and weaving stability of weft yarns containing a high proportion of mechanically recycled cotton fibers with reduced fiber quality. The warp consisted of high-quality Egyptian combed cotton (2 × 6 tex, S-twist), and the weft of recycled yarns (gray or dyed, 29.5 tex) in a plain weave, chosen for structural cohesion and to minimize stress on weaker weft threads. Nominal densities were 29 ends/cm and 20 picks/cm, ensuring sufficient coverage of recycled weft yarns across all samples in non-square construction. Part 1: Woven fabrics underwent five controlled washing and drying cycles. Structural changes were assessed through warp and weft densities, fabric mass, thickness, shrinkage, porosity, and cross-sectional analysis. This approach enabled a comprehensive evaluation of fabric stabilization and microstructural alterations induced by repeated laundering. The contaminants released during simulated washing were assessed on a small-scale laboratory setup using the AHIBA apparatus under conditions conforming to EN ISO 105-C06 and ISO 4484–1. The washing procedure was conducted at 60 °C for 30 min, with a heating rate of 4 °C min-1 and a rotational speed of 40 min⁻1, followed by cooling to room temperature. Each bath contained 100 mL of tap water and 25 steel balls to enhance the mechanical action of the washing process. After washing, all fabric samples were dried under controlled conditions, including air 24 h pre-drying between baking paper layers followed by oven drying at 105 °C to obtain a stable dry mass while minimizing environmental contamination. Released fibers, fragments, and other contaminants were collected by filtration using a UHELON 130 T EXTRA polyamide monofilament sieve (diameter of monofilament used in warp and weft direction 35 μm, density of ends 145 per cm/ density of picks 130 per cm, size of pores 25 μm × 35 μm) and quantified gravimetrically relative to the initial fabric mass. Washing water was dyed to enhance the visibility of released contaminants, and procedural controls allowed checking contamination. Redeposited particles were further examined by SEM, and their elemental composition was characterized using EDS, allowing both qualitative and quantitative assessment of the released impurities. Part 2: A methodology for qualitative and quantitative assessment of contaminants released from woven fabrics during simulated washing was adapted and set. The study addresses the limitations of mass-based measurements and visual inspection by incorporating precise procedural controls. The experiment was performed on 100-m yarn hanks using the AHIBA apparatus under the same controlled conditions selected for the previous experiment. This setup enables the detection of released fibers and impurities while distinguishing warp and weft contributions. Procedural blanks were included to monitor potential contamination from air or handling. The input fibre materials (B, C, D) differed in quality, including fibre length, short-fibre content, and contamination levels. Yarn properties such as hairiness and strength were characterized to support the interpretation of shedding behavior. Released fibers and particles were collected on polyamide monofilament filters and weighed using a high-precision balance (0.01 mg readability), providing the initial experimental data required to develop the methodology for subsequent image-based analysis. Part 3: Filter residues containing deposited fiber fragments, fiber bundles, and other impurities obtained from the repeated washing and drying experiments described in Part 1 were used as the analytical input dataset. The image-analysis workflow adapted and set in Part 2 was directly applied to these samples, ensuring consistent morphological and statistical evaluation across yarn- and fabric-level measurements within a unified methodological framework. The procedural blank control (BC) established in Part 2 was consistently applied for all analysed sample types and used to correct for background contamination during image analysis and data evaluation. To maintain clarity and avoid excessive duplication of presented data groups, the blank control datasets are not explicitly shown in the graphical outputs. The quantitative and qualitative assessment of fibre shedding Scanning of filters with redeposited contaminants from controlled washing of woven fabrics was performed under standardized conditions designed to capture objects across the full expected size range while minimizing optical artefacts. The scan area, resolution, calibration, and illumination settings were optimized based on prior validation to ensure reliable detection of both the smallest particles and the longest fibres, while maintaining feasible data size and processing time. Images were acquired using an automated macroscope system with controlled stage movement, tiled acquisition, and subsequent stitching into a single calibrated image of the central filter region. Representative filters were selected from replicate experiments based on mass proximity to the sample mean and uniform spatial distribution of contaminants. Detailed acquisition parameters and analytical steps are summarized in the accompanying graphical workflow scheme in Fig. 1, and full details of the image processing pipeline are provided in Part 2. All size- and length-related descriptors, including fibre fragment lengths as well as Feret-based dimensions of dust and trash particles, should be interpreted as morphological representations of fragmented and fibrillated structures obtained via image-based reconstruction, rather than absolute physical dimensions of intact fibres or particulate contaminants.Fig. 1Flowchart of image processing. Flowchart of image processing. Scanning Electron Microscopy (SEM) was conducted for all yarn samples to examine surface morphology before and after laundering using a VEGA 3 TESCAN microscope. Representative images included in the manuscript are limited to selected samples (materials B and D, gray and dyed variants) to ensure clarity and readability of the presented results. Materials B and D represent the most contaminated fiber input and the shortest staple length with the highest short fiber content, respectively. Statistical data processing All experimental data were statistically processed as part of the methodological adaptation introduced in Part 2, including data pre-processing and statistical evaluation. The procedures comprised tests for homogeneity and normality, following standard approaches in exploratory data analysis8. Previous studies indicated that particle size descriptors, such as Feret diameters of impurities (feretmax dust, feretmax trash), may approximate a normal distribution. However, in the present dataset, an exponential transformation was required due to pronounced skewness. Conventional descriptive statistics, including mean, standard deviation, and 95% confidence intervals, were calculated. Fiber fragments-related parameters, such as lengths of individual fiber fragments (l individual fibers), estimation of maximum length of the branch (l max branch), and the estimation of the longest shortest path for a fiber bundle (l fibre bundles), also exhibited skewed distributions. To address this, statistical modelling (l individual fibers, l max branch, l fibre bundles) was performed using two complementary approaches: A) Parametric approach using Gamma distributions: Gamma distributions were applied to model positively skewed continuous variables, particularly fiber length-related descriptors. The Gamma distribution is suitable for non-negative, right-skewed data and allows robust estimation of distribution parameters (shape and scale), from which mean values, standard deviations, and confidence intervals can be derived. This approach enables a physically meaningful representation of fiber fragmentation processes, where short fragments dominate but long-tail distributions are present. Rare long fibers with negligible statistical influence on the dataset were treated as outliers and excluded using a median absolute deviation (MAD) criterion (3 × threshold), after verification of their limited occurrence. B) Non-parametric bootstrap approach: Bootstrap resampling was performed using 1,000 iterations, where each iteration involved random sampling with replacement from the original dataset to generate synthetic resampled datasets. For each iteration, statistical descriptors (mean and interquartile range) were calculated, and the final confidence intervals were derived from the empirical distribution of these 1,000 resampled statistics. This approach does not assume any specific distribution and is therefore robust for highly skewed data. When the assumptions required for parametric ANOVA, namely normality and homogeneity of variances, were not met, group differences were evaluated using the Kruskal–Wallis test as a robust nonparametric alternative. In this study, the term “groups” refers to predefined experimental categories of yarns and fabrics, in particular differences in recycled cotton blend (materials B, C, and D), yarn state (gray vs. dyed variants), and structural configuration of the samples as defined in Table 1. This method determines whether the medians of independent groups differ significantly by ranking all observations and comparing the summed ranks across groups. A statistically significant Kruskal–Wallis result indicates that at least one group differs from the others. Due to the large number of experimental groups, repeated washing cycles, and multiple morphological descriptors evaluated in this integrative study, the Kruskal–Wallis test was primarily used as a robust screening tool for identifying overall statistically significant differences among experimental groups. This approach is particularly suitable for skewed distributions and enables reliable comparison without relying on parametric assumptions. Results and discussion Expectations are formulated for the selected qualitative parameters of contaminants, and results are interpreted using comparative plots. These plots group fabric samples by fiber type (B, C, and D) and account for differences in weft yarn finishing, whether gray or dyed in light or dark colors. The discussion is organized in a logical sequence that emphasizes variations in trends both within individual samples and across the complete set of samples. Assumptions and expected outcomes It is hypothesized that the laundering process will induce the release of both particulate matter (dust) and fibrous fragments. While the quantity and morphological characteristics of these contaminants are expected to evolve across sequential washing cycles, a central question remains as to whether the intensification of mechanical stress, facilitated by the inclusion of steel balls, will allow the fabric to reach a steady-state shedding plateau or if cumulative fatigue will result in sustained contamination4 and9. Higher shedding rates are anticipated in fabrics containing gray, undyed weft yarns compared to those with dyed weft yarns. This difference is attributed to the stabilizing effect of the dyeing process, which acts as a pre-wash by removing loosely bound surface fibers and residual manufacturing debris. In particular, fabrics made with weft yarn type B, manufactured from highly contaminated regenerated cotton, are expected to release significantly higher dust particle counts, especially during the first washing cycle3,5 and6. Findings from Part 2 indicated that even high-quality combed compact warp yarns release an amount of impurities comparable to that of rotor-spun counterparts. Therefore, it is assumed that the total number of contaminants released from the woven fabrics may not differ substantially across experimental variants. Moreover, due to the intrinsic heterogeneity of recycled materials and the limits of instrumental precision, many qualitative differences in particle size and fibre fragments length distribution are likely to remain statistically indistinguishable, falling within overlapping 95% confidence intervals. Repeated exposure to cyclic mechanical stress is expected to induce progressive fiber damage. While the initial wash is dominated by the release of production-inherited fragments exhibiting transverse damage, subsequent cycles are predicted to reveal fibrillation and surface ruptures as cotton fibers approach their mechanical fatigue limits10–12. The evaluation of fiber shedding from the textile structure The absolute number of impurities detected in the central regions of the filters selected for analysis, with respect to their proximity to the median mass of released contaminants, is presented in Figs. 2, 3, 4, 5. The results indicate substantial cycle-to-cycle variability in the number of dust particles ndust, ranging from 948 to 4605 particles, suggesting that the release process is not yet fully stabilized under the applied washing conditions. Samples containing gray weft yarn within the respective fabric groups consistently exhibited elevated ndust values, particularly for GC and DC fabric types, indicating a higher susceptibility of these material configurations to particle release. This behavior may be associated with the combined effect of recycled fibre content and yarn structure, which can promote increased fibre fragmentation during repeated washing cycles. Importantly, no clear stabilization trend was observed over the five washing cycles, which further supports the conclusion that the system remains in a transient release regime rather than reaching equilibrium shedding behavior.Fig. 2The number of dust particles ndust in the filters after each washing step for the fabric groups B, C, and D.Fig. 3Examples of the trash particles detected on the filter.Fig. 4The number of fibre fragments nfibres and bundles nbundles in the filters after each washing step for the fabric groups B, C, and D.Fig. 5The binary portion of all contaminants in the filters after each washing step for the fabric groups B, C, and D. The number of dust particles ndust in the filters after each washing step for the fabric groups B, C, and D. Examples of the trash particles detected on the filter. The number of fibre fragments nfibres and bundles nbundles in the filters after each washing step for the fabric groups B, C, and D. The binary portion of all contaminants in the filters after each washing step for the fabric groups B, C, and D. Trash-type particles with feretmax > 0.5 mm were observed only sporadically and in very low numbers, indicating that large fragment release is not a dominant mechanism under the tested conditions. Specifically, they were detected as follows: LC—2 particles in the 4th and 5th wash; DC—2 particles in the 2nd wash, 1 particle in the 3rd wash, and 3 particles in the 4th wash; GD—3 particles in the 3rd wash and 2 particles in the 4th wash; and LD—1 particle in the 2nd wash and 1 particle in the 4th wash. These rare occurrences suggest stochastic release events rather than systematic structural degradation. For reference, representative examples of cellulose-based plant impurities (trash particles) detected on the filter after laundering are illustrated in Fig. 3. Partial or complete coloration of these particles is visible due to their cellulose nature and similar dye affinity to cotton fibers, which additionally facilitated their separation from the background during image analysis. The number of released fiber fragments nfibers, expressed as individual fibers, ranged from 276 to 1774, while fiber bundles nbundles ranged from 26 to 4821. In most cases, the highest release of both individual fibers and fiber bundles occurred during the first washing cycle, followed by a progressive decrease in subsequent cycles, indicating an initial wash-off effect of loosely attached or weakly bound material. Within individual sample groups, differences were relatively small, suggesting a comparable release behavior under identical processing conditions. Moreover, only minor differences were observed between fabrics containing gray versus dyed weft yarns, indicating a limited influence of coloration on fiber release mechanisms. Importantly, the evaluation of dust particle counts included correction based on the procedural blank control (BC), which was particularly relevant for the first washing cycle. In samples with higher background contamination levels, especially within the yarn type B group, a larger proportion of non-representative particles and artefacts was identified and removed during image preprocessing. This ensured that only particles attributable to actual fibre release from the textile structure were included in the final dataset. As a consequence, the reported values for the first cycle represent net fibre release after background subtraction and provide a more reliable basis for inter-sample comparison. Across all sample groups, fabrics containing weft yarn type B consistently exhibited slightly higher levels of both individual fiber fragments nfibers and fiber bundles nbundles compared to fabrics containing weft yarn types C and D. This behavior is consistent across all measured impurity-related metrics and reflects the higher proportion of mechanically weakened and short fiber components present in type B yarn. These structural characteristics increase the probability of fiber detachment during repeated mechanical loading, leading to higher fragment counts across washing cycles. This trend may be associated with differences in fiber composition and structural cohesion of the respective yarn systems. A similar tendency was also observed in Part 1 of this study based on gravimetric analysis, where fabrics with gray recycled cotton weft yarns exhibited higher overall contaminant mass, although partial overlap between groups was attributed to the limited sample size. Consistent trends were also confirmed in Part 2 at the yarn level, where the evaluation of individual yarns showed higher release of particles and fibre fragments for type B yarn compared to types C and D. The observed fibre release behavior is also influenced by the fabric structural configuration. The weave structure, yarn type and yarn quality, in combination with fabric density parameters, define the number of fibre–fibre contact points and frictional constraints within the textile system. These interactions, determined by yarn density (expressed as ends and picks per cm) and the degree of interlacement, may contribute to restricted fibre mobility and reduced release of loosely bound fibre fragments under mechanical stress. This observation can be interpreted in the context of previously reported findings on same rotor-spun yarns, which demonstrate that fibre length distribution, short-fibre content, and contamination level strongly influence yarn cohesion, uniformity, and hairiness1,2. In particular, studies on rotor-spun yarns produced from pre-consumer cotton waste further show that fibre contamination level, residual impurities, and cleaning efficiency significantly affect yarn structural integrity and stability1. These yarn-level characteristics were also confirmed in Part 2 of this study, where type B yarn exhibited higher levels of dry lint generation and impurity content compared to types C and D, indicating reduced structural quality. Within this framework, a higher proportion of mechanically weakened and short fibres, as observed in type B yarn, is expected to reduce structural cohesion and increase susceptibility to structural instability. Consequently, this leads to an enhanced probability of fibre release under cyclic mechanical loading. To provide a clearer perspective on the amount of released impurities, the binary fraction of all contaminants is also presented. The data of binary portionall show that this value ranges from 0.0052 to 0.014 across the samples. The highest value was observed for the fabric with gray weft yarn during the first washing cycle. Emissions are generally highest in the initial cycle when the weft yarn is in the gray state. When dyed weft yarn is used in the woven structure, where impurities may have been partly removed during the dyeing process, short fiber ends must be mobilized from the structure to enable their release through mechanical friction between the fabric and the steel balls. This may explain the observed increase in the binary portion in subsequent cycles, suggesting that the release of impurities follows a cyclic behavior related to fiber alignment, fixation, and progressive structural fatigue within the fabric hierarchy. Subsequent cycles show lower or stabilized values. The typical dust particle sizes ferretmax dust for the observed fabric types and individual washing cycles are broadly comparable across samples, ranging from 23.75 to 56.67 µm. However, despite this apparent similarity, the confidence intervals do not fully overlap, see Fig. 6, indicating subtle but statistically distinguishable differences between certain sample groups. The size of trash particles is significantly larger, ranging from 421 to 658 µm, clearly separating this fraction from the dust population and confirming the presence of two distinct impurity size regimes. Representative examples of trash particles are shown in Fig. 3.Fig. 6The size of dust in terms of ferretmax dust in the filters after each washing step for the fabric groups B, C, and D. The size of dust in terms of ferretmax dust in the filters after each washing step for the fabric groups B, C, and D. The visual comparison of selected estimators of fibre-type impurity length (l individual fibers, l max branch, l fiber bundles) is presented for both the original dataset and the dataset with extreme values removed, see Figs. 7, 8, 9. Box plots are shown for all grouped samples categorized according to the type of recycled cotton fibres used in the production of the weft yarn (B, C and D). These plots are complemented by mean estimates and corresponding confidence intervals calculated using two approaches: variant A – Gamma distribution (black) and variant B – bootstrap method (red).Fig. 7Fiber fragment length lindividual fibers in filters after each washing step for the fabric groups B, C, and D.Fig. 8Fiber fragment length of lmax branch in filters after each washing step for the fabric groups B, C, and D.Fig. 9Fiber bundle length lfibre bundles in filters after each washing step for the fabric groups B, C, and D. Fiber fragment length lindividual fibers in filters after each washing step for the fabric groups B, C, and D. Fiber fragment length of lmax branch in filters after each washing step for the fabric groups B, C, and D. Fiber bundle length lfibre bundles in filters after each washing step for the fabric groups B, C, and D. The analysis indicates that the length of released fibre fragments captured on the filter and identified as individual objects l individual fibers is characterized by median values of the original data ranging from 48 to 229 µm. For the longest branch in the skeleton representation l max branch, median values range from 45 to 3,040 µm. In contrast, the median of the longest shortest path within fibre bundle structures l fiber bundles varies from 41.9 to 5,262 µm. Differences among individual samples were not statistically significant at the 95% confidence level, as indicated by overlapping confidence intervals. When the Gamma distribution is used to estimate the mean and variance of length-related parameters, it generally yields lower mean values and higher variance than the bootstrap approach, reflecting sensitivity to distributional assumptions in highly skewed datasets. Overall, the evaluated parameters suggest broadly comparable trends in fibre release behavior across most samples. However, local deviations are observed for l individual fibers in samples DB (2nd wash), LC (1st wash), and LD (1st wash), where reduced values are evident compared to the remaining dataset. For l max branch, increased values are observed in DB (1st wash) and LD (3rd wash), while LC (3rd wash) shows a reduction. A similar pattern is observed for l fiber bundles, with increases in DB (1st wash) and LD (3rd wash), and a decrease in LC (3rd wash), indicating sample-dependent variability in fragmentation behavior. In the initial analysis, all fiber fragments were considered as a single dataset. However, due to the strong heterogeneity of fragment sizes and the highly skewed distributions, a global evaluation did not allow sufficient morphological discrimination. Therefore, a size-based stratification was introduced to improve interpretability of fragmentation behavior across different scales. The lower threshold (500 µm) was selected in analogy to the fine particulate (dust-like) fraction commonly used in textile contamination studies, while the upper threshold (2000 µm) emerged empirically from the observed break in the size distribution, which indicated a natural separation between the dominant fragment population and a distinct group of rare long fiber fragments. Importantly, data points that initially appeared as statistical outliers in the global dataset were not excluded from the analysis; instead, they were reclassified into a separate “long fragment” category based on this empirically identified threshold, and evaluated independently in order to preserve their physical relevance and avoid loss of extreme morphological information. These categories correspond to micro-, meso-, and macro-scale regimes: ultrashort fiber fragments (< 500 µm), short fiber fragments (500–2000 µm), and long fiber fragments (> 2000 µm). In all cases, the number of observations in the l individual fibers 1 dataset for the ultrashort fragment category represents the minimum dataset available for statistical evaluation. Data for the l individual fibers 2 short fragment subgroup are missing in samples DB 2nd wash, DC 1st wash, and LD 1st wash. The occurrence of l individual fibers 3 long fiber fragments is sporadic; therefore, no statistical evaluation was performed for this category. Fiber fragments longer than 2000 µm were excluded from comparative graphical analyses due to their very low frequency of occurrence. Comparative graphs for ultrashort and short fiber fragments are presented in Figs. 10 and 11. In most cases, the l individual fibers 1 distributions across fabric groups B, C, and D show comparable ranges, including median and mean estimates obtained using both approaches (A – Gamma distribution and B – bootstrap method). Lower values are observed for samples DB 2nd wash, GB 2nd wash, LC 1st wash, and LD 1st wash.Fig. 10Fiber fragment length l individual fibers 1 in filters after each washing step for the fabric groups B, C, and D.Fig. 11Fiber fragment length l individual fibers 2 in filters after each washing step for the fabric groups B, C, and D. Fiber fragment length l individual fibers 1 in filters after each washing step for the fabric groups B, C, and D. Fiber fragment length l individual fibers 2 in filters after each washing step for the fabric groups B, C, and D. Similarly, the l individual fibers 2 distributions across fabric groups B, C, and D also show broadly comparable ranges, with consistent median and mean estimates obtained using both statistical approaches. Overall, individual fiber fragment lengths appear to be generally consistent across fabric groups and washing cycles, irrespective of the applied micro-, meso-, or macro-scale classification. To assess differences among the investigated samples and the length characteristics of fibre fragments (l individual fibers 1, l individual fibers 2, l max branch, l fiber bundles), a nonparametric Kruskal–Wallis test was applied. The choice of this method was motivated by non-normal (right-skewed) data distributions, the presence of extreme values, and unequal sample sizes across groups. The tested hypotheses were defined as follows: H₀: all groups originate from the same distribution (equal medians); H₁: At least one group differs significantly in its median. The sum of squares SS, degrees of freedom df, mean square MS, including the Kruskal–Wallis test χ2 statistic, and the associated p-value are listed in Tables 2, 3, 4, 5.Table 2Nonparametric one-way ANOVA Kruskal–Wallis test results for l individual fibers 1.Original dataSourceSSdfMSχ2 statisticpGroups7.02689E + 114415,970,206,03010,067.750Error1.31714E + 1228,89545,583,618.74Total2.01983E + 1228,939Filtered dataSourceSSdfMSχ2 statisticpGroups7.58718E + 114417,243,581,48211,331.650Error1.13907E + 1228,30040,249,905.49Total1.89779E + 1228,344Table 3Nonparametric one-way ANOVA Kruskal–Wallis test results for l individual fibers 2.Original dataSourceSSdfMSχ2 statisticpGroups105,693,707.6412,577,895.31120.639.01941E-10Error2,733,909,9973200854,346.87Total2,839,603,7053241Filtered dataSourceSSdfMSχ2 statisticpGroups110,248,209.4412,688,980.72147.446.30346E-14Error2,128,520,2182953720,799.26Total2,238,768,4272994Table 4Nonparametric one-way ANOVA Kruskal–Wallis test results for l max branch.Original dataSourceSSdfMSχ2 statisticpGroups23,863,125,83744542,343,7692985.010Error54,425,287,29997495,582,653.33Total78,288,413,1379793Filtered dataSourceSSdfMSχ2 statisticpGroups22,253,074,32644505,751,689.23042.510Error46,257,504,20493234,961,654.425Total68,510,578,5309367Table 5Nonparametric one-way ANOVA Kruskal–Wallis test results for l fiber bundles.Original dataSourceSSdfMSχ2 statisticpGroups23,293,147,85244529,389,723.93282.070Error42,191,708,52391834,594,545.2Total65,484,856,3759227Filtered dataSourceSSdfMSχ2 statisticpGroups20,957,511,27044476,307,074.33380.280Error32,510,657,91285803,789,120.97Total53,468,169,1838624 Table 2Nonparametric one-way ANOVA Kruskal–Wallis test results for l individual fibers 1.Original dataSourceSSdfMSχ2 statisticpGroups7.02689E + 114415,970,206,03010,067.750Error1.31714E + 1228,89545,583,618.74Total2.01983E + 1228,939Filtered dataSourceSSdfMSχ2 statisticpGroups7.58718E + 114417,243,581,48211,331.650Error1.13907E + 1228,30040,249,905.49Total1.89779E + 1228,344 Nonparametric one-way ANOVA Kruskal–Wallis test results for l individual fibers 1. Table 3Nonparametric one-way ANOVA Kruskal–Wallis test results for l individual fibers 2.Original dataSourceSSdfMSχ2 statisticpGroups105,693,707.6412,577,895.31120.639.01941E-10Error2,733,909,9973200854,346.87Total2,839,603,7053241Filtered dataSourceSSdfMSχ2 statisticpGroups110,248,209.4412,688,980.72147.446.30346E-14Error2,128,520,2182953720,799.26Total2,238,768,4272994 Nonparametric one-way ANOVA Kruskal–Wallis test results for l individual fibers 2. Table 4Nonparametric one-way ANOVA Kruskal–Wallis test results for l max branch.Original dataSourceSSdfMSχ2 statisticpGroups23,863,125,83744542,343,7692985.010Error54,425,287,29997495,582,653.33Total78,288,413,1379793Filtered dataSourceSSdfMSχ2 statisticpGroups22,253,074,32644505,751,689.23042.510Error46,257,504,20493234,961,654.425Total68,510,578,5309367 Nonparametric one-way ANOVA Kruskal–Wallis test results for l max branch. Table 5Nonparametric one-way ANOVA Kruskal–Wallis test results for l fiber bundles.Original dataSourceSSdfMSχ2 statisticpGroups23,293,147,85244529,389,723.93282.070Error42,191,708,52391834,594,545.2Total65,484,856,3759227Filtered dataSourceSSdfMSχ2 statisticpGroups20,957,511,27044476,307,074.33380.280Error32,510,657,91285803,789,120.97Total53,468,169,1838624 Nonparametric one-way ANOVA Kruskal–Wallis test results for l fiber bundles. In all evaluated cases, statistically significant differences among groups were observed (p < 0.05), with most p-values far below 0.001, indicating strong evidence against the null hypothesis. The relatively high χ2 values, particularly for l individual fibers 1, l max branch, and l fiber bundles are attributed to pronounced data skewness, large dataset sizes, and differences in the number of observations across evaluated parameters, as the amount of detected fibre fragments and bundles varied between experimental groups. For l individual fibers 1 and l individual fibers 2, pronounced inter-group differences were observed. Filtering of extreme values led to an increase in χ2 values, indicating that the observed differences are not driven by outliers but reflect systematic differences in the underlying distributions. The parameter l max branch exhibited statistically significant differences among groups while remaining relatively stable between original and filtered datasets, indicating robustness to extreme values and suitability as a morphological descriptor. Similarly, l fiber bundles showed strong and consistent inter-group differences across both datasets, confirming that structural differences between samples persist independently of data filtering. Mechanistic interpretation in the context of previous studies Direct comparison of the mass and length distributions of released fiber fragments with previous studies is constrained by substantial differences in experimental design, particularly regarding textile structure, material composition, and analytical methodology. In addition, no equivalent hybrid woven construction based on rotor-spun yarns with a high proportion of recycled cotton fibers (29.5 tex) is reported in the literature, further limiting direct comparability. In the previously published study on these yarns7 and in Part 2 of this work, lint generation was assessed under different mechanical loading conditions. Under dry-state abrasion using the Lawson–Hemphill system, dyed rotor-spun yarns consistently exhibited lower lint generation than gray variants, indicating reduced fiber release under friction-dominated conditions. In Part 2, intensified laundering using 25 steel balls introduced combined wet friction and impact loading. Under these conditions, differences between gray and dyed yarns were reduced but remained detectable, particularly during the first washing cycle. These results suggest that the initial response to mechanical loading is governed by the yarn surface state and the presence of loosely anchored short fiber ends. This behavior is attributed to progressive mobilization of protruding fiber segments under mechanical action. Under dry conditions, shear and tensile stresses at the yarn surface progressively extract partially embedded fibers, increasing fuzz and hairiness, as previously reported13 and 14. Comparable changes in hairiness during production and finishing stages have also been reported for woven structures, where processing significantly modifies surface fiber protrusion and hairiness evolution15. Fiber release appears to depend on the interplay between inter-fiber friction, structural anchoring, and external mechanical loading, which collectively determine resistance to fiber pull-out. Representative SEM micrographs of yarn surfaces before and after the first laundering cycle are presented in Fig. 12. Compared to unwashed yarns, the washed samples exhibit localized abrasion damage and surface disruption of fibers located on the yarn surface, indicating progressive mechanical wear under intensified laundering conditions with steel balls. These observations qualitatively support the proposed mechanism of progressive fiber mobilization during repeated laundering cycles.Fig. 12Representative SEM micrographs of 29.5 tex rotor-spun yarn surfaces before and after the first laundering cycle: (a) yarn GB before washing, (b) yarn DB before washing, (c) yarn GD before washing, (d) yarn DD before washing, (e) yarn GB after washing, (f) yarn DB after washing, (g) yarn GD after washing, and (h) yarn DD after washing. Representative SEM micrographs of 29.5 tex rotor-spun yarn surfaces before and after the first laundering cycle: (a) yarn GB before washing, (b) yarn DB before washing, (c) yarn GD before washing, (d) yarn DD before washing, (e) yarn GB after washing, (f) yarn DB after washing, (g) yarn GD after washing, and (h) yarn DD after washing. These observations are consistent with studies on fiber friction and cohesion, which show that yarn compactness and inter-fiber interactions control fiber slippage. Repeated cyclic loading (bending, tensile, and shear) progressively degrades fiber anchoring until a critical threshold is reached, resulting in either fiber pull-out or rupture. Under wet conditions, additional mechanisms contribute to fiber mobilization. Washing promotes swelling of cellulosic fibers, increases water penetration into amorphous regions, and weakens intermolecular interactions within cellulose structures16. The resulting more open yarn structure facilitates migration and detachment of mechanically weakened fibers. Although swelling may locally reduce frictional resistance, combined shear and impact loading during laundering enhances fuzz formation and promotes subsequent fiber rupture and release. These findings are consistent with higher levels of residual trash and impurities reported for undyed cotton materials3, which, together with a higher proportion of short fibers in gray yarns, increase surface heterogeneity and the density of weakly anchored fiber ends. Natural waxes present in untreated cotton fibers may further influence inter-fiber friction and detachment behavior6. Overall, the results demonstrate that yarn production technology, fiber composition, and wet processing jointly govern fiber cohesion, structural integrity, and fiber release mechanisms4 and5. The proposed methodology provides a suitable framework for systematic fiber release studies and enables more consistent comparison of shedding behavior across processing states and material systems. Conclusion This final part of the study successfully integrated the quantitative findings from Part 1 with the advanced image-analysis methodology adapted and set in Part 2, creating a unified analytical framework for evaluating fiber shedding from recycled cotton textiles. By transitioning from a weight-based assessment to a detailed morphological characterization, this research provides a deeper understanding of the mechanisms governing contaminant release in woven structures. Part 3 focuses on the evaluation of particulate and fibrous contaminant release from woven fabrics produced with recycled cotton weft yarns (types B, C, and D) subjected to repeated laundering under intensified mechanical stress over five washing cycles. The number of cycles was selected intentionally, as previous results from Part 1 demonstrated that the woven fabric structure reaches a geometrically stable and dimensionally equilibrated state after this period. The present study therefore evaluates fibre release within the structurally stabilized state of the textile system. The washing conditions, including the use of 25 steel balls, were selected in accordance with the methodological framework established in Part 1, where this setup was defined as an intensified abrasion-assisted laundering protocol aimed at accelerating and quantifying fibre release under controlled laboratory conditions. The results confirm that the laundering process induces the systematic emission of dust particles, individual fibres, and fibre bundles, with clear influence of the recycled cotton raw material type on fibre shedding behavior, while yarn manufacturing and fabric construction parameters were kept constant across all samples. The highest release was observed during the first washing and drying cycle, particularly for fabrics containing gray weft yarn. This behavior is consistent with the findings of Part 1 and Part 2 of this study, where gray yarns exhibited higher levels of loosely bound impurities and dry lint generation prior to washing. The observed difference between gray and dyed yarns reflects the influence of pre-treatment history, where gray yarn retains a higher proportion of surface contaminants that are removed during the initial wash. In contrast, dyed yarns undergo prior rewinding and cleaning operations, followed by wet processing, which partially removes loosely bound impurities and modifies the fibre surface condition, resulting in a shifted release profile. Subsequent washing cycles generally showed lower release values, indicating progressive depletion of easily detachable material and a transition toward a more stable structural state, in which fibre detachment is increasingly influenced by mechanical fatigue and structural resistance. Fabrics containing weft yarn type B, produced from highly contaminated regenerated cotton, consistently exhibited slightly higher release of individual fibres and fibre bundles compared to types C and D. This behavior is consistent with the higher proportion of mechanically weakened and short fibres in type B yarn, which reduces structural cohesion and increases susceptibility to fibre detachment under cyclic mechanical loading. Morphological analysis showed that dust particle sizes feretmax were comparable across all fabric types (23.75–56.67 µm), while larger trash particles (> 500 µm) occurred only sporadically. Fibre fragment length parameters (l individual fibers 1, l individual fibers 2, l max branch, l fiber bundles) displayed overlapping median confidence intervals in descriptive statistics, while Kruskal–Wallis tests confirmed statistically significant inter-group differences (p < 0.001). Extremely high χ2 values observed for individual fibre parameters reflect strong heterogeneity in fragment distributions and differences in sample population sizes rather than measurement inconsistency. The classification of individual fibres into three empirically defined length-based subgroups (ultra-short, short, and long fibre fragments) provided a more detailed description of fragmentation behavior and enabled the separation of previously identified extreme values into a physically meaningful category rather than treating them as statistical outliers. Overall, the results indicate that fiber release appears to be a cyclic process influenced by fiber alignment, fixation, and fatigue within the hierarchical fabric structure. The experimental design captured the release behavior up to the point of structural stabilization of the textile system, which was the primary objective of this study. Based on the observed trends, it can be hypothesised that fabric hierarchical structure, particularly yarn interlacement density and the resulting number of fibre–fibre contact points, plays a key role in limiting fibre detachment into the aqueous environment by mechanically stabilizing the fibre network. This work therefore establishes a methodological and conceptual basis for future studies focused on long-term ageing behavior and extended fibre release kinetics beyond structural stabilization. Such investigations are necessary to fully describe time-dependent degradation processes and emission behavior under prolonged service conditions.