Prediction of physical textile parameters from the tactile
assessment of fabrics
Shary Kock Student ID: 5994918 Final version 3 May 2016
Supervisor: Hein Daanen Signature:
Second Examiner: Frank Nack Signature:
Master Thesis Information Science Human Centered Multimedia
University of Amsterdam Faculty of Science
Prediction of physical textile parameters from the tactile
assessment of fabrics
Shary Kock
Student ID: 5994918
Amsterdam, Netherlands
sharykock@gmail.com
ABSTRACTIt is of interest to investigate if the assessment of textiles can be executed without mechanical intervention to make this process less dependent on expensive equipment for fashion designers, buyers and manufacturers. To investigate the manual assessment of textiles in comparison with the mechanical approach we first conducted a pilot to extract the interactions with fabric handling and terms used to describe them. We measured the objective data with the FAST system and the subjective data was acquired through a survey with participants. We measured the relations between FAST and the 18 semantic terms used to describe fabric properties in the survey by the means of a correlation matrix and multiple regressions. Bending (FAST) could be predicted by ‘thick’ and ‘stiff’ (subjective) with 68% explained variance; extensibility with 77% explained variance from ‘elastic’, shear with 67% explained variance from ‘stiff’ or ‘thick’ and formability with almost 60% explained variance from ‘elastic’. The most difficult to predict was compression. In summary, when no systems are available to measure physical textile properties, a rough estimation can be derived from subjective panel scores. For future studies we recommend to involve more subjects for the subjective assessment and include more samples varying in composition and appearance.
Author Keywords
Fashion; technology; fabric hand; tactile; interaction; mechanical properties; subjectivity.
ACM Classification Keywords
H.5.2. (D.2.2, H.1.2, I.3.6): Evaluation/methodology; 1. INTRODUCTION
Textiles are constantly evolving due to the development of new fibers and the innovative ways of rendering textile materials [1]. Clothing can be seen as an interface between the body and the environment. When the choice for a garment is made to wear around the body different aspects are at play. It can be related to the comfort [2] of the wearer, the activity the wearer is participating in, the cultural background of the wearer, or the identity the individual wishes to express towards the outside world. Therefore humans dress themselves with comfort in mind, while protecting them from the environment (hot or cold weather) and optionally keeping esthetics in mind (culture, fashion).
With such a vast availability of textiles and their different
ways of use and expressiveness [3] it becomes a laborious task to ensure the qualities and tailorability of a garment. The evaluation of fabrics is already part of the manufacturing process1. However, as the role of fabrics might change it is important to investigate materials beyond qualities upheld by the industry with regards to colorfastness, durability and fiber analysis. Though it is necessary to know these fabric parameters to establish a feasible production line, it is also important to involve other stakeholders, namely the fashion designer and the end user. Whether it is a buyer, inspecting the garments before mass consumption and production, or a fashion designer deciding on textiles for a new collection, they must inspect and therefore interact with the fabrics in order to make an informed decision to proceed with manufacturing. For the manufacturers it is important to receive feedback from buyers and designers to uphold the industry’s standards with regards to quality.
The dominant method of acquiring fabric properties is mechanical. The two most used systems for the mechanical approach are Kawabata2 (KES-F) and Sirofast3 (FAST). Recently the Fabric Touch Tester4 (FTT) impacted in this process. KES-F is the most common method for the objective evaluation of fabric hand and drape [4]; while FAST is able to predict how a fabric will behave once converted into a garment. Both systems basically measure the same parameters, but use different test methods. The KES-F system is more elaborate in the sense that it measures the resistance and recovery of fabric deformation. On the contrary the FAST system is less expensive, simpler and more robust to use since it only measures the resistance of fabric to deformation. Though both approaches support the new way of garment prototyping in form of 3D virtual garment and pattern making [5], where the emphasis lies on the reuse of pattern and material descriptions, their process of measuring samples is lengthy as the samples need to be marked, cut out and conditioned at least 24 hours 1 http://www.manufacturingsolutionscenter.org/testing-services.html 2 http://nptel.ac.in/courses/116102029/55 3http://www.itecinnovation.com/productDetails.php?id=65 4 http://www.sdlatlas.com/product/478/FTT-Fabric-Touch-Tester
beforehand. Moreover, the analysis has to be performed by an expert, as only the precise performance of an evaluation with respect to placement of samples and the duration of the measurement, provides reliable values. Afterwards the testing on the machines is preferably carried out by an expert, to assure consistency of the standards held by the industry. The tests need to be carried out a number of times on the 3 different FAST machines, since measuring occurs in different directions such as the warp, weft and bias [6]. An additional problem is the accessibility of the machines. In the Netherlands for example, it is only accessible at the Amsterdam Fashion Institute5 (AMFI) in Amsterdam and the Vlisco6 lab in Helmond. Finally, even in the process of machine evaluation some manual processes need to be performed. For example, once the machine evaluation is done, the designer has to take the results and distribute them manually into the 3D environment of choice, e.g. the Lectra Modaris V7R27.
A different approach towards fabric testing is based on subjective assessment through human touch [7], where the material under investigation is touched, squeezed, rubbed, or treated otherwise. The problem here lies in the growing collection of textiles and lengthy process of measuring their properties and the transferability of the collected data in existing 3D virtual garment and pattern environments. The latter shares process problems as already outlined for FAST, only that the Lectra at the moment works with values, where a manual approach will result in semantic terms. If a mapping would be possible then experts in the field could make use of manual fabric evaluation, freeing them from the burden of using difficult to access machinery.
This study aims to determine whether subjective assessments of fabric samples can be used as an alternative to objective measuring and how these interactions with fabrics can be used in the future for the support of virtual garment design and prototyping tools.
In this paper we first discuss previous work related to this study. We then describe the FAST system in more detail and elaborate further on the most relevant properties for fabric design and evaluation. We then outline our comparative study between FAST and touch-based measuring, its results and the discussion of those. The paper closes with a conclusion and a view on future work. 2. RELATED WORK
There is a sound body of work on fabric experience (fabric hand) and perception.
Peirce [8] defined fabric hand as “being the judgement of the buyer, which depends on time, place, season, fashion
5http://amfi.nl/
6http://www.vlisco.com 7http://www.lectra.com/en
and personal predilections” [8, p.377]. Owen [9] defined fabric hand as “all the sensations that are felt by the fingers when the cloth is handled”. In an early definition Schwartz [10] and Brand [11] defined fabric hand as a subjective property as evaluated by its customers. A more general approach towards defining fabric hand is “the total sensation experienced when a fabric is touched or manipulated with both fingers [12, p.1]”.
Fabric hand consists of different fabric properties, rendering it a complex parameter, consisting of flexibility, compressibility, elasticity, resilience, density, surface contour (roughness, smoothness), surface friction and thermal character [13].
There are different processes at work when forming a subjective perception of a garment. There are physical processes between clothing and the environment, physiological processes between the body and skin and psychological processes that are processed by the brain. For the subjective hand analysis technique the measurement of the psychological processes is most relevant, which is the human perception measured in a subjective way. This perception is the direct measure of a person’s opinion. In order to carry out this direct measuring psychological scaling need to be applied; this is the process of making judgements based on the scale of individual words and language [14]. A study by Slater [15] encountered problems with this type of technique due to 1) it relies on the honesty of the subject, 2) human opinions have an extended variety, 3) it is difficult to perform statistics on subjective data since they are not real numbers and the mental calibration by each subject might not be the same. In a later study Slater [16] elaborated on those findings by stating that also inconsistencies arise as opinions can be influenced by psychological, physiological, social and environmental factors.
Dillon et. al [17] encountered that blindfolded subjects had a stronger tactile perception than non-blindfolded participants. However, when they combined tactile perception with vision the overall perception was enhanced. Some fabrics feel different when touched by hand than when experienced on the body [18].
Kayseri et. al [19] and Kweon et. al [20] showed that gender provides different results in subjective perception, as females responds in a more delicate and sensitive manner. Philippe [21] conducted a tactile sensory analysis with 11 evaluators for 7 cotton samples containing different finishing treatments. A total of 15 attributes were evaluated on a linear scale. The results showed the evaluators were able to distinguish between the treated samples.
Luible [22] used 3 different evaluation cycles to measure subjectivity with the 2 finger assessment (instead of both hands). A rating method was applied with 2 references on a
5 point scale. The ratings showed medium correlations with the objective measurements done with KES-F.
Tactile perception is also a complex parameter. The reason we touch is to obtain information of an object. According to [23] the feel of an object depends on a combination of its perceptual properties. Although there are different parameters at play that form our opinion the brain receives and processes this as one parameter. Chen et al. [24] show that touch perception is often associated with more than one property. Therefore if there are more properties at play these cannot be separated or distinguished against each other. The brain perceives this combination of properties as a singular sensation. This renders the determination of fabric handle complex. Therefore there is a need for fabric hand analysis techniques, the mechanical technical specifications and the sensory judgment of a fabric. 3. COMPARISON STUDY
In this study we are interested to see if simple touch-based measuring can establish similar values of fabric description as performed with mechanical measurement processes. The idea is that touch-based methods are to be performed easier with respect to access to material and availability to the technique.
The methodology used for this research consists of a modeling of manual fabric evaluation and then use this model applied to denim evaluation to compare its results with raw, objective data obtained from the FAST machines. The assessment of this comparison should allow us to see if the established descriptions are similar and hold relations that can be used to establish a basis for ease of inclusion of manual results into 3D fabric design tools.
We decided for the FAST environment out of opportunistic means, simply because it is available at AMFI. We are aware of the fact that FAST uses simpler methods than KES-F, but as it measures the same parameters it is sufficiently sophisticated as a tool.
We decided for denim as plot samples due to its versatility in use and availability in different blends (variations in thickness and stretch). This provides us with a broad range of choice to experiment. We also had, similar to the decision for the FAST environment, a pragmatic reason: in the study program at AMFI there is a minor dedicated to denim and so students could benefit from the results of this study.
The study consists of two parts. We first conducted a pilot study in order to explore how fashion designers judge fabrics; how they interact with these and the terms they use to describe them. The outcome of this study should provide us with the set of semantic terms we could use within the main evaluation, namely to find correlations between those semantic terms and objectively measured fabric values, so that we can investigate if it is possible to let designers or users evaluate material by hand similarly accurate as the measurements provided by the FAST environment. We also
hope with this evaluation to provide insights how a translation from touch measurements into objective values can be performed.
We first outline the procedures of the evaluation and then present the results and discussion of the final comparison study in section 4.
3.1 Pilot study
A pilot survey was conducted with 4 fashion designers from AMFI. They were all female and between 24-25 years old. They were recruited through the teachers here at AMFI. They were assessed on the spot during one of their classes and had to fill in a survey containing 25 questions concerning demographics and several properties of textiles for 5 samples (different in color, content and print) as depicted in Figure 1.
Figure 1. Pilot samples.
The pilot study was of a more explorative nature to gain access and insight in how fashion designers judge fabrics. On the one hand we looked at the manner of interaction in which they handle (touch, press, rub or handle otherwise) the fabrics and on the other hand we observed the terms (e.g. soft, warm) used when assessing the fabrics. By extracting these individual and manual fabric assessments it provided us with a guideline for the setup of the subjective measuring method. Although the AATCC8 (American Association of Textile Chemists and Colorists) test method EP5 fabric hand is considered a standard guideline it is not commonly accepted and various variations exist when it comes to the subjective assessment of fabric hand.
The observations showed that all four designers make use of the same handlings to assess the properties of fabric. For example for the term elastic they held the sample between both hands and performed a little pull to test the degree in which it stretches. To determine questions related to the surface of the sample they rubbed it against the palm of their hand. For general impressions they rubbed the sample between index finger and thumb, pinched the sample or letting it hang on one end (between index and thumb) and subtly wave it. Both left and right hand were used for these actions.
8
A semantic word cloud analysis of those descriptions in combination with the terminology provided by Chen et al. [24], we established a list of 13 terms that describe the process as well as value parameters for the manual evaluation of fabric. The relevant terms for the evaluation values are: warm, dry, stiff, rough, irritating, smooth, synthetic, pleasant, thick, solid, elastic, limp, and greasy. Those terms formed the basis for the evaluation of the manual part of the final evaluation.
3.2 FAST Environment
For being able to compare the manual approach with the mechanic approach of fabric evaluation, we also performed an analysis of the parameter set provided by the FAST machines.
There are four FAST components, namely FAST-1 which measures the compression, FAST-2 which measures the bending, FAST-3 which measures the extensibility and FAST-4 measures the dimensional stability. The latter is not an instrument but a test method and the only component that is carried out manually. FAST-4 is left out of the scope for this study. These 3 components also return values for properties. These are the shear rigidity, formability and the bending rigidity. Shear rigidity is calculated from the bias extensibility (FAST-3); formability is obtained from the extensibility (FAST-3) + the bending rigidity (FAST-2) and the bending rigidity is calculated as follows: bending length * weight.
The list of properties is: Compression
Compression looks at the fabric and the surface thickness and predicts problems with pressing.
Bending
Bending looks at the bending length and bending rigidity (related to perceived stiffness), both in warp and weft direction. Predicts problems with cutting and automated handling.
Extensibility
Extensibility covers the extension in the warp and weft direction. Problems are predicted with regards to laying up, pattern matching, moulding, over feed seams and sleeve insertion.
Formability
Formability is a measure of the ability of a fabric to absorb compression on its own plane without buckling and is calculated from FAST-2 + FAST-3. Predicts problems with seam puckering.
Shear Rigidity
Shear rigidity is how easy a fabric can be distorted by a trellising action (related to fabric looseness) and is
calculated from the bias extensibility (FAST-3). The same problems arise as mentioned under Tensile Extension.
Bending Rigidity
The bending rigidity is derived from the bending length (FAST-2) * weight.
We then performed a semantic comparison between those 6 FAST concepts with the derived 13 terms for manual assessment. It turned out that the 13 terms could be distributed among the 6 FAST terms. However, in particular compression was not adequately covered, and so we enhance the list of 13 terms with an additional 5 terms, namely hairy, prickly, soft, strong, and grainy (those were also part of the word cloud but at the time considered as not strong enough).
3.3 Environment Comparison
As the potential correlations were formed we then evaluated 5 different types of denim with the FAST environment and by means of manual evaluation.
In both environments we used the same type of 5 different denims, as presented in Figure 2. They differ with respect to weight, structure, and content. The denims are further referred to as D1 (the lightest) up to D5 (the heaviest).
Figure 2. Five types of denim samples, where the most left is the lightest and the most right the heaviest type of denim.
3.3.1 Fast Environment
In order to measure the physical textile parameters the objective measuring was conducted at the Vlisco lab in Helmond with the help of an expert. This was necessary as the FAST environment at AMFI was not accessible due to water damage.
The samples were emailed previously by post to be conditioned at least 24 hours beforehand according to standard atmospheric regulations of 20°C temperature, 65% relative humidity. The samples were cut according to the measurement standards stated in the FAST manual. Six samples (130 x 50 mm) per denim for the compression and bias and 6 samples (200 x 50 mm) per denim for bending and extension of which 3 were used for the warp direction and the other 3 were used for the weft direction. Normally a fabric sample cutter is used to cut cylindrical samples (100mm2) of a fabric to measure the weight 3 times in total and calculate the mean of these 3 values. For this part of the research we deviated from this sample cutting method. This is due to the fact that the razors were very blunt and provide
an inaccurate sample. Also, the pad on which it is cut contains leftover fibers from previously cut samples, which can alter the weight of the current sample. Therefore we made the choice to cut out samples with the dimensions of 10 x 10 cm. The different samples can be seen in Figure 2. 3.3.2 Manual Environment
The preparations for the subjective method were of similar nature as with the objective method. They were also conditioned at least 24 hours beforehand according to standard atmospheric regulations of 20°C temperature, 65% relative humidity.
The participants for the manual evaluation were recruited in a purpose-driven selective manner. A total of 9 subjects aged between 21-28 years old, of which 8 were female and 1 male, participated. They had backgrounds in fashion and design and fashion and management.
Figure 3. Subjective assessment set up at the lab. During the evaluation, which took place in the AMFI lab 133 in form of individualized sessions (see Figure 3), the participants were first informed about the research goals and the procedure of the session. They had to agree that the session was videotaped and that their measurements and additional statements will be used for this research only. Each participant had to perform two basic evaluation processes, namely first a sorting task performed on all denims, and then a second task, in which each denim had to be evaluated individually. The denim sets for each participant were new sets, so that all could work under the same condition.
The first task for each participant was to inspect the 5 denim samples and then order them based on their weight (from the lightest to the heaviest). With this test we wanted to evaluate if the participants were in the position to establish “value-based” evaluations on the provided samples that were similar in kind to what the FAST machines could establish. Once done, a participant had to fill in the first part of the survey, containing demographic information and the result of the ranking task.
After that each participant received a set of the 5 different denim samples (10 by 10 cm). The subject was allowed full freedom to inspect the samples through touching, rubbing,
squeezing, folding or by handling them otherwise. No reference material was used and the subjects were not blindfolded. No time restriction was provided for the assessment of each denim type.
After each assessment they had to complete the second part of the digital survey (we used Google Form Survey Tools9). For each denim they had to provide 25 assessments related to thermal properties, bending, compression, extensibility/formability/shear, and surface/texture. The categories were set-up in this manner to determine how the participants grouped the provided 18 semantic terms under the 6 FAST categories and how they evaluate each semantic concept with respect to its relevance. For the latter we used a 5 point Likert scale, where 1 stated “Strongly Irrelevant” and 5 denotes “Strongly Relevant”.
Once the participant had filled in the survey, he or she was rewarded with 5 Euros and a lollipop for their contribution. The sessions per subject took between 20 to 30 minutes. 4. RESULTS AND DISCUSSION
In this section we provide the results from the different evaluation processes and our discussions of the findings. 4.1 Results
In this section we first look at the values provided by the FAST machines and then present the findings of the manual evaluation.
4.1.1 FAST
The results from the FAST machines are described in table 1 to 3 and Figure 4.
Figure 4. All values measured with FAST.
Figure 4 shows that denim sample 1 was the thinnest fabric due having the lowest compression value. Denim sample 3 came out as the most elastic with the highest values for Extension. Denim sample 4 came out as the stiffest with the highest values for Bending.
Table 1 presents the basic properties of the denim samples with a variety in weight, structure and content. The weight
is determined by the mean of the 3 square samples and the structure was determined with the help of a textile expert. The samples D1, D2 and D3 are blended fabrics, while D4 and D5 consist of natural fibers.
Sample Fiber content Structure Weight (g/m2) D1 93% cotton 3% silk No Stretch Right hand twill (RHT) 2-1 139 D2 80% cotton 18% bamboo 2% lycra Stretch Right hand twill (RHT) 2-1 281 D3 52% cotton 21% rayon 2% lycra 26% polyester Stretch Left hand twill (LHT) 2-1 352 D4 100% cotton FC No Stretch Broken twill 2-1 482 D5 100% cotton NTB 400-69 No Stretch Right hand twill (RHT) 3-1 345
Table 1. Basic properties of the denim samples.
For the FAST-1 (Compression), FAST-2 (Bending) and FAST-3 (Extension) Table 2 provides an overview of the mean values of the warp and weft direction for each of the denim samples.
Table 2. FAST-1, FAST-2, FAST-3 properties.
Furthermore the FAST system also provides us with derived properties (Table 3) such as the Formability and the Shear Rigidity.
Sample Shear Rigidity (N/m) Formability (mm2)
D1 44.5 0.2
D2 67.7 1.2
D3 134.1 3.8
D4 321.2 2.7
D5 50.2 1.3
Table 3. Derived properties from FAST. 4.1.2 Manual
Figure 5 describes the average distribution of the 18 semantic concepts assigned by the 9 participants to each denim type. This distribution forms the basis for the correlation analysis we then performed.
We performed an analysis of the mean, median, and mode distribution of ranks for each semantic parameter for every denim type. What we found is that the semantic concepts stiff, thick, pleasant, elastic, prickly, strong, irritating, and hairy showed consistent representations over the evaluations of the subjects within and over denim types. The other showed consistencies for one particular denim but mainly demonstrated high standard deviations.
For the sorting task they were asked to inspect all samples manually and to determine the correct order of the denim weights by ranking them from lightest to heaviest. The sorting task was carried out flawlessly, besides 1 participant who switched samples D2 and D5. This gives us an indication that the subjects were able to determine the weight parameter by hand. Not by exact measurements, but by feeling and estimation.
It needs to be mentioned that we received additional statements in the free text part of the questionnaire, providing additional conceptual descriptions. We discuss the relevance of those later in the discussion section. The comments for sample 1 were: “nice feeling fabric - fabric memory seems poor on bias”. For sample 2: one side feels soft while the other (right side) feels grainy. For sample 3: “good fabric memory”. For sample 4: “The inside of the fabric feels soft, but the outside feels a little rough...”; “fabric feels very strong, grainy on one side and soft and smooth on the underside, feels like some mechanical stretch has been added in the weave”. For sample 5: “very rough denim stretch only really on the bias, doesn't have very good memory when stretched or pulled in all directions.” 4.1.3 Correlations
We compared the relationships between the FAST and manual measurements and established the correlation matrix. In table 4 we present the essential correlation matrix that provides relevant correlations between semantic terms and FAST measurement values over all denims.
Sample Compression (mm) Bending length (mm) Extension (%) D1 0.4 14.8 1.8 D2 0.6 19.0 3.2 D3 1.0 23.0 2.1 D4 0.9 30.0 1.0 D5 1.0 18.0 1.9
Figure 5. Subjective values for the 18 semantic terms.
Table 4. Correlation matrix.
Looking at the individual set of correlations we get the following measurements.
For the bending parameter we notice a positive correlation (p >0.6) with the parameters of stiff (0.78) and thick (0.79), which implies that the stiffer and thicker a fabric is the more difficult it is to bend. The fabric needs more time and length to go over the lever resulting in a bigger angle that the sensor needs to detect. Figure 6 and 7 shows us the graphic representation of the bending estimation performed by the subjects against the actual measured bending values.
Figure 6. Stiff plot for Bending.
Figure 7. Thick plot for Bending.
For compression a positive correlation was found with the grainy parameter with a value of 0.61, see Figure 8. The matrix also returned negative correlations with the parameters smooth (-0.68), pleasant (-0.62) and soft (-0.61). These values account for an R2 value of 0.54, which means that as a set of parameters they account for 54% of the variance. See Appendix A for the negative line fit plots.
Figure 8. Grainy plot for Compression.
Extensibility had a positive correlation with the elastic parameter with a value of 0.88, see Figure 9. Negative correlations occurred with parameters stiff (-0.70) and thick (-0.73). This set of parameters accounts for a 78% of variance. See Appendix B for the negative line fit plots. y = 26.788x - 24.031 R² = 0.8978 -50.00 0.00 50.00 100.00 150.00 0 2 4 6 B en d in g ( m ic ro N * m ) stiff stiff Line Fit Plot
Bending Predicted Bending Linear (Bending ) Linear (Predicted Bending ) y = 27.478x - 23.397 R² = 0.9086 -50.00 0.00 50.00 100.00 150.00 0 2 4 6 B en d in g ( m ic ro N * m ) thick thick Line Fit Plot
Bending Predicted Bending Linear (Bending ) Linear (Predicted Bending ) y = 0.1084x + 0.4573 R² = 0.6736 0.00 0.20 0.40 0.60 0.80 1.00 1.20 0 2 4 6 C o m p re ssi o n ( m m ) grainy grainy Line Fit Plot
Compression Predicted Compression Linear (Compression) Linear (Predicted Compression) Com-pression Bending Extens-ibility Shear Formability stiff 0.78 -0.70 0.78 0.66 rough 0.65 smooth -0.68 pleasant -0.62 thick 0.79 -0.73 0.78 0.66 elastic 0.88 0.77 grainy 0.61 soft -0.61 0.00 1.00 2.00 3.00 4.00 5.00 Denim 1 Denim 2 Denim 3 Denim 4 Denim 5
Figure 9. Elastic plot for Extensibility.
For the derived properties Shear Rigidity (Figure 10 and 11) and Formability (Figure 12-15) we notice only positive correlations. Shear correlated with parameters stiff and thick with an equal value of 0.78. The explained variance for shear rigidity is at 67%. For formability explained variance had a percentage of 64.
Figure 10. Stiff plot for Shear Rigidity.
Figure 11. Thick plot for Shear Rigidity.
Figure 12. Stiff plot for Formability.
Figure 13. Rough plot for Formability.
Figure 14. Thick plot for Formability.
Figure 15. Elastic plot for Formability. y = 0.8507x + 0.3816 R² = 0.9876 0.00 2.00 4.00 6.00 0 2 4 6 E x te n si o n elastic elastic Line Fit Plot
Extension Predicted Extension Linear (Extension ) Linear (Predicted Extension ) y = 57.495x - 27.24 R² = 0.897 0.00 50.00 100.00 150.00 200.00 250.00 300.00 350.00 0 2 4 6 S h ea r (G) stiff stiff Line Fit Plot
Shear (G) Predicted Shear (G) Linear (Shear (G)) Linear (Predicted Shear (G)) y = 59.029x - 26.016 R² = 0.9094 0.00 50.00 100.00 150.00 200.00 250.00 300.00 350.00 0 2 4 6 S h ea r (G) thick
thick Line Fit Plot Shear (G)
Predicted Shear (G) Linear (Shear (G)) Linear (Predicted Shear (G)) y = 0.4481x + 0.695 R² = 0.3904 0.00 1.00 2.00 3.00 4.00 5.00 0 2 4 6 Fo rm a b il it y ( m m 2 ) stiff stiff Line Fit Plot
Formability Predicted Formability Linear (Formability ) Linear (Predicted Formability ) y = 0.1392x + 1.5081 R² = 0.0352 0.00 1.00 2.00 3.00 4.00 5.00 0 2 4 6 Fo rm a b il it y ( m m 2 ) rough rough Line Fit Plot
Formability Predicted Formability Linear (Formability ) Linear (Predicted Formability ) y = 0.5554x + 0.4632 R² = 0.5768 0.00 1.00 2.00 3.00 4.00 5.00 0 2 4 6 Fo rm a b il it y ( m m 2 ) thick thick Line Fit Plot
Formability Predicted Formability Linear (Formability ) Linear (Predicted Formability ) y = 0.4072x + 0.7932 R² = 0.387 0.00 1.00 2.00 3.00 4.00 5.00 0 2 4 6 Fo rm a b il it y ( m m 2 ) elastic elastic Line Fit Plot
Formability Predicted Formability Linear (Formability ) Linear (Predicted Formability )
Formability correlated with four parameters of which equal values of 0.66 for stiff and thick, 0.65 for rough and 0.77 for elastic.
4.2 Discussion
We are aware of the fact that the findings will only tell us something about the correlation between machine-based and human-based semantic evaluations for one particular fabric, namely denim. Thus, different fabrics might result in different types of correlations.
Despite this uncertainty regarding the scalability of the findings we can state that all the provided semantic terms could be assigned by the participants to one machine category and that participants are consistent there.
The first problem that occurs is that the two compared measurement types are of different nature. Where the machinery values provide one certain value, the semantic terms provide a not specified range. This becomes apparent in our analysis of the mode, median, and mean values, which in half of the cases showed large differences in made assignments. This indicates that a potential mapping matrix that facilitates a mapping from manual evaluation to objective values requires a range to adapt to idiosyncratic interpretation of what a semantic term actually means. We also saw that only half of the provided measurements for the 18 semantic terms by the participants showed consistency and that again only half of those also provided relevant correlations between the two types of measurements. Those were stiff, thick, elastic, and pleasant. From those stiff and thick showed correlations for 4 machine categories (bending, extensibility, shear, and formability) and those correlations also make sense in general material behavior. Elastic, the semantic term that provides the highest correlations, provides two (extensibility and formability), which were expected. Pleasant only provided correlations for the category compression but we had expected that at least it would also correlate with extensibility. All the other semantic terms show also limited correlations but those that are expected. Thus, it seems that a few semantic terms can be easier defined and aligned to the responding FAST measurement set. As correlations do not provide causation, we cannot state that the established correlations clearly indicate the relevance of the semantic terms but it seems that the established set in table 4 is the one to be investigated further. In this context it is also apparent that these 8 terms are distributed among the FAST categories and hence should be able to describe a fabric. Additional tests need to identify, however, if this set is indeed sufficient.
When taking all denim samples into account the parameters stiff and thick resulted as the most influential. Both affect all FAST values except for compression. Compression measures (surface) thickness and it is interesting to see that the subjective parameter thick has no direct influence on its own. This implies that the concept of touch is indeed of a
complex nature. The brain and the human perception do not only rely on one parameter for a certain impression. Instead compression is influenced negatively by smooth, pleasant and soft and only positively by grainy, which can be expected as it is the opposite of the other parameters. The grainy parameter has to do with the surface properties. Since FAST-1 also looks at the surface thickness these findings indicate it is best to look at both thickness and surface properties on their own. Already by sensing the sample and stating that it is grainy the user has an indication that the compression values will be affected without being measured with the FAST system.
Considering the already mentioned findings, it seems that the presented evaluation approach is too shortsighted. Considering the value type problem between the two measurement methods, it seems to be better to let designers define fabrics in environments like Lectra, then produce the new fabric and then let designers evaluate the material with the by then established semantic terms. In that way one could get a better understanding of the idiosyncratic processes in fabric imagination and final perception, which can help to really understand the range set necessary for a mapping matrix.
Finally, a shortcoming of our analysis is that we have not considered to look if there are differences with respect to gender and age of the participants. Those parameters are considered relevant in literature and even though we have the data, we did not find the time to investigate in this direction. It is our assumption, though, that this might indicate some answers to the value diversity among the semantic terms, meaning that in case of different clustering more coherence might be achieved.
5. CONCLUSION AND FUTURE WORK
Both types of fabric hand analysis techniques, FAST and subjective assessments, have valuable contributions. FAST data provides us with raw data that predicts how a fabric will behave in terms of quality and tailorability. Subjective parameters, on the other hand, provide a human perception on the touch of fabrics and can be applied to the customer’s wishes. The valuable contribution fashion designers deliver here is that they are able to provide a good estimation of fabric properties by hand omitting the use of expensive and time-consuming systems like KES-F and FAST.
What we can state is that certain parameters can be mapped to the measurements retrieved from fabric evaluation systems, namely stiff, rough, smooth, pleasant, thick, elastic, grainy, and soft. The advantage of this set is that the established 8 parameters do address and cover all machine-based categories. However, as there are still large inconsistencies in the application and validation of those terms it cannot be stated that a mapping matrix could be established. As the mapping would require ranges on the side of the semantic terms (which term covers which ranges of measurement values), it is important to test what these ranges really are. In that context it also needs to be
investigated, if the current value differences are based due to the gender and age of the fashion designers.
It is also necessary to establish a clearer picture with respect to the core set of semantic terms with respect to different type of fabrics.
At the same time we see that the use of 3D virtual garment prototyping allows the designer to test different materials at the same time on 1 type of garment. It can be the case that none of the available materials is apt for the design at hand. Now the possibility exists to communicate back to the manufacturer to order a customized fabric that will adhere to the specific properties needed. This approach is altering the design process where the designers acquire more say in the manufacturing process. For the manufacturers this means a more efficient way of working and meeting the client’s wishes immediately by providing them with useful materials. This means an increase in the quality standards that were uphold until so far by the industry and less waste of materials.
By taking these fabric hand assessments into consideration for the potential they have to serve as an alternative this might be used for future haptic feedback systems. By providing a virtual simulation of the feel of the fabric that the user can “touch” while designing in 3D, properties such as bending can already be inferred on how it will impact the garment in process.
ACKNOWLEDGEMENTS
My dearest gratitude and appreciation are extended to my supervisor H. Daanen for his guidance and support, and to my second reader F. Nack for his attention and support. Furthermore thanks to S. Kuijpers, A. Vink, L. Duncker, L. Vonk and the participants for their valuable contributions. REFERENCES
1. Luible, C. Derivation of innovative fabric
characteristics from 3D virtual garment simulations. 2. Daanen, H. A. M. "Kleding en mode: een kritisch
wetenschappelijke beschouwing." (2014). ISBN 978 90 5629 757 2.
3. Lam Po Tang, S., & Stylios, G. K. (2006). An
overview of smart technologies for clothing design and engineering. International Journal of Clothing Science and Technology, 18(2), 108-128.
4. Kawabata, S., & Niwa, M. (1989). Fabric performance in clothing and clothing manufacture. Journal of the Textile Institute, 80(1), 19-50.
5. Siersema, I. (2015, November). The influence of 3D simulation technology on the fashion design process and the consequences for higher education. In Digital Fashion Conference 2015 (p. 9).
6. Haghighat, E., Najar, S. S., Ahmadloo, E., & Rastgoo, A. (2012). Estimation of the Initial Shear Modulus of Twill Woven Denim Fabrics Based on Fabric
Mechanical and Geometrical Properties. Fibres & Textiles in Eastern Europe.
7. Behery, H. (Ed). (2005). Effect of mechanical and physical properties on fabric hand. Elsevier. 8. Peirce, F. T. (1930). 26—THE “HANDLE” OF
CLOTH AS A MEASURABLE QUANTITY. Journal of the Textile Institute Transactions, 21(9), sT377-T416.
9. Owen JD. Handle and drape of fabrics. Shipley Link 1971; 4: 18.
10. Winn, L. J., & Schwarz, E. R. (1939). Technical Evaluation of Textile Finishing Treatments Flexibility and Drape as Measurable Properties of Fabric. Textile Research Journal, 10(1), 5-16.Niwa, M., Yamada, T., & Kawabata, S. (1983). Objective Evaluation of Apparel Fabrics. Textile Machinery Society of Japan, 67.
11. Brand, R. H. (1964). Measurement of fabric aesthetics analysis of aesthetic components. Textile Research Journal, 34(9), 791-804.
12. Luible, C., Varheenmaa, M., Magnenat-Thalmann, N., & Meinander, H. (2007, October). Subjective fabric evaluation. In Cyberworlds, 2007. CW'07.
International Conference on (pp. 285-291). IEEE. 13. Mäkinen, M., Meinander, H., Luible, C., &
Magnenat-Thalmann, N. (2005, December). Influence of physical parameters on fabric hand. In submitted to the
Proceedings of this Conference (Vol. 9).
14. El Mogahzy, Y. E., Kilinc, F. S., & Hassan, M. (2005). Developments in measurement and evaluation of fabric hand. Effect of mechanical and physical properties on fabric hand, 45.
15. Slater, K. (1985). Human comfort (Vol. 1). Springfield, Ill., USA: CC Thomas.
16. Slater, K. (1986). Discussion paper the assessment of comfort. Journal of the textile institute, 77(3), 157-171. 17. Dillon, P., Moody, W., Bartlett, R., Scully, P., Morgan, R., & James, C. (2001). Sensing the fabric: To simulate sensation through sensory evaluation and in response to standard acceptable properties of specific materials when viewed as a digital image. In Haptic Human-Computer Interaction (pp. 205-218). Springer Berlin Heidelberg.
18. Brewster, S., & Murray-Smith, R. (Eds.).
(2003).Haptic Human-Computer Interaction: First International Workshop, Glasgow, UK, August 31-September 1, 2000, Proceedings(Vol. 2058). Springer. 19. Kayseri, G. Ö., Mengüç, G. S., & Özdil, N.
(2012). Sensorial comfort of textile materials. INTECH Open Access Publisher.
20. Kweon, S. A., Lee, E. K., & Choi, J. M. (2004). A comparative study on the subjective fabric hand
according to gender for winter sleepwear fabrics. Fibers and Polymers, 5(1), 6-11.
21. Philippe, F., Schacher, L., Adolphe, D. C., &
Dacremont, C. (2004). Tactile feeling: Sensory analysis applied to textile goods. Textile research
journal,74(12), 1066-1072.
22. Luible, C., Varheenmaa, M., Magnenat-Thalmann, N., & Meinander, H. (2007, October). Subjective fabric evaluation. In Cyberworlds, 2007. CW'07.
International Conference on (pp. 285-291). IEEE. 23. Heller, M. A., & Schiff, W. (2013). The psychology of
touch. Psychology Press.
24. Chen, X., Shao, F., Barnes, C., Childs, T., & Henson, B. (2009). Exploring relationships between touch perception and surface physical properties. International Journal of Design, 3(2), 67-76.
APPENDICES
Appendix A – Negative plots for Compression parameters
Appendix B – Negative plots for Extensibility parameters y = -0.1357x + 1.1605 R² = 0.8572 0.00 0.20 0.40 0.60 0.80 1.00 1.20 0 5 10 Comp ress io n (m m) smooth smooth Line Fit Plot
Compression Predicted Compression Linear (Compression) y = -0.1346x + 1.2651 R² = 0.6851 0.00 0.20 0.40 0.60 0.80 1.00 1.20 0 2 4 6 C o m p re ssi o n ( m m ) pleasant pleasant Line Fit Plot
Compression Predicted Compression Linear (Compression) y = -0.1185x + 1.1541 R² = 0.6888 0.00 0.50 1.00 1.50 0 2 4 6 C o m p re ssi o n ( m m ) soft soft Line Fit Plot
Compression Predicted Compression Linear (Compression) y = -0.1661x + 3.0667 R² = 0.0314 0.00 1.00 2.00 3.00 4.00 5.00 6.00 0 2 4 6 E x te n si o n ( % ) stiff stiff Line Fit Plot
Extension Predicted Extension Linear (Extension ) y = -0.0358x + 2.7218 R² = 0.0014 0.00 1.00 2.00 3.00 4.00 5.00 6.00 0 2 4 6 E x te n si o n ( % ) thick thick Line Fit Plot
Extension Predicted Extension Linear (Extension ) Linear (Predicted Extension )
