Modeling of surface temperature distributions on powered e-textile structures using an artificial neural network

Abstract

An artificial neural network (ANN) model is constructed to derive the surface temperature of e-textile structures developed for cold weather clothing. A series of textile transmission lines made of different types of conductive yarns, insulated by using different types of seam tapes, were enclosed in a thermoplastic textile structure via hot air welding technology, and then they were powered with different levels of specific voltages in order to obtain different heating levels. The surface temperatures of the powered e-textile structures were measured using a thermal camera. The experimental input variables, sample type, temperature, feeding speed, resistance of samples, applied voltage and current were used to construct an ANN model and the outputs of surface temperature and electric power dissipated were used to test the prediction performance of the developed model. It was concluded that the ANN provided substantial predictive performance. Simulations based on the developed ANN model can estimate the surface temperature distributions of powered e-textile structures under different conditions. The ANN model developed for prediction of electric power dissipated was very successful and can be useful for e-textile product designers as well as textile manufacturers, particularly for cold weather protection products such as jackets, gloves and outdoor sleeping mats.

Keywords

e-textiles transmission lines electrical heating conductive yarns power welding artificial neural networks

Many prediction-related problems in textiles are computed by artificial neural network (ANN) architecture. ANN applications have mainly focused on fiber and yarn production issues, that is, prediction of cotton yield with supplemental nitrogen,¹ fiber identification with NIR (near-infrared) light without destruction of the fiber,^2–6 formulating physical/molecular structure property relationships of man-made fibers,^7,8 inspection of textured yarn packages,⁹ determination of the degree of spinnability from fiber and process properties with an accuracy of over 90%,¹⁰ yarn inspection to classify trash and neps with an accuracy of over 97%,¹¹ determination of the performance of a spinning unit from the number of yarns, ring frames, utilization levels of machines, etc., with an average error of around 1%¹² and prediction of the relationship between fiber properties, yarn count and/or yarn tensile and elongation properties with errors of less than 7% or correlation coefficients of over 0.84.^13–17 Moreover, ANN studies concerning fabric properties, including prediction of the relationship between fabric manufacturing process parameters and fabric properties, such as strength, bending rigidity, thickness, drape, shear, wrinkle, etc., with average absolute errors of around 2% and coefficient of correlation of 0.95,^18–26 fabric defect detection with a level of success around 90%,^27–34 fabric pattern classification³⁵ and prediction of stitch formation and sewability from fabric and sewing machine properties with a level of success of over 97%,^36–38 as well as those on prediction of color recipes/dyestuff concentration from absorbance,^39–44 were of great interest in the literature. However, it is apparent that there is not much work on ANN application to electronic textiles (e-textiles).

The concept of e-textiles explains interactive reactions that can sense the signals, process the information and give actuation with responses via the textile structure.^45,46 Today's innovative e-textile manufacturing approaches let the textiles and fabrics have applications mainly in wearable computing systems, medical and healthcare, sportswear and the military.^47–51 For instance, textile electrodes integrated into clothing that can be worn comfortably can gather clinical physiological data comparable to gel electrodes in order to monitor the health of the wearer in both sports and healthcare settings.⁵² Garment-based gesture controls or muscle fatigue detection can be realized by textile-based electromyography sensors.^53,54 Beyond sportswear or healthcare, the attention on connecting people through clothing-based wearable technology has been recently pioneered by incorporating micro-light emitting diode (LED) arrays to create wearable displays and tactile or touching textile sensors⁵⁵ and color-tunable textiles for illumination or decoration^56,57 in the fashion and entertainment sectors. Indeed, the product range and vision of the e-textile technology market is constantly improving and enabling more commercial product developments. For instance, electric resistance heating textile applications especially for cold weather protection like jackets, gloves and outdoor sleeping mats have attracted great interest in the market recently.⁵⁸

The interface layout or design of e-textiles is generally realized using conductive textile yarns or fibers. From a textile perspective, the textile-based transmission/power lines can be formed by designing conductive tracks on the textile structure based on different textile production techniques, such as printing, welding, knitting, weaving, embroidery, etc.^59–64 However, the current passing through these conductive tracks might cause a vital problem, since electrical energy is converted to heat on top of conductive tracks while transmitting power. On the other hand, based on the resistance values of conductive yarns with the current passing through, these conductive yarns can be used for heating purposes. In both cases, the surface temperature distribution on textile structures is critical and vital to be known as a design parameter before manufacturing e-textile structures, since they might cause burns on human skin. Another design constraint is the very high cost of e-textiles, which must be kept in mind at all times. Moreover, precise measurement of surface temperature and heat dissipation are time consuming, and demand highly sophisticated equipment as well as well-trained personnel.

The focus of this study is to offer an ANN-based model for prediction of final average temperature and heat dissipated on e-textile structures, including conductive yarns. Firstly, an experimental study has been conducted based on manufacturing of e-textile structures for heating purposes. Hot air welding technology has been used in order to insulate the conductive tracks over the textile structure, since it can offer a practical solution to prevent the short circuits caused by some of the disturbances resulting from the wearability concept of the textile structures. E-textile structures were used in electric power transmission and distribution to transmit and convert electrical energy into heat. Secondly, based on the experimental results, an ANN model has been designed in order to predict the average surface temperatures and heat dissipation of e-textile structures depending on conductivity level and applied voltage. Hence, the offered model can be a solution for product designers as well as textile manufacturers by eliminating the highly expensive pre-product design process for electric resistance heating product applications.

Experimental study

Electric resistance heating theory

In an electric resistance heating system, the electric current passing through the resistor generates an amount of heat depending on its level; however, the current flow is limited by the resistance and applied voltage in a circuit based on a basic principle of Ohm's law (Figure 1).^65,66

Figure 1.

Ohm's law illustration.

Production of e-textile structures for heating applications

A series of textile electrical resistance lines made of different types of two of the most popular commercially available conductive yarns supplied by Koolon ® and Shieldex® were manufactured using a hot air welding machine. The properties of conductive yarns used in the study are summarized in Table 1. In order to have rigid (less elastic) samples, woven fabric construction was chosen. Two different woven thermoplastic fabrics made of 100% polyester (PET) and 100% polyamide (PA) were used as the base material, with fabric weights of 200 and 180 g/m² and thicknesses of 0.24 and 0.30 mm, respectively. After placement of conductive yarns onto the base fabrics, an insulation layer was created over the base fabric by using three different types of seam tapes containing polyurethane (PU), namely PU/PU (ST-604 with two layers), PU/PU/nylon tricot (ST-306 with three layers) and PU/PU/ PET jersey (ST-318 with three layers).

Table 1.

The properties of the conductive yarns

Conductive yarn ID	Material type	Linear weight (g/m)	Yarn diameter (µm)	Yarn count (dtex)	Yarn twist (tours/m)	Linear resistance (Ω/m)
Y1	100% stainless steel	0.10	14	90f × 1	–	<65
Y2	100% stainless steel	0.51	12	275f × 2	41	<15
Y3	100% stainless steel	0.22	14	90f × 2	42	<30
Y4	100% stainless steel	0.76	12	275f × 3	31	<9
Y5	100% Silver plated PA	0.06	29	235f × 34 × 2 × 2	78	<50
Y6	100% Silver plated PA	0.11	20	235f × 34 × 2	30	<85

PA: polyamide

While preparing the samples, parameters such as temperature and feeding speed have been taken into consideration depending on the operational conditions of the hot air welding machine in accordance with all kinds of fabrics and welding tapes. The processing temperature was set to be 450 ℃, 500 ℃ or 550 ℃, depending on the seam tape and fabric type, while the feeding rate was adjusted to 1.5, 3, 6, 8 or 12 ft/min (see Table 2) in order to investigate the effect of processing temperature and sealing speed on manufacturability of the samples.

Table 2.

Processing conditions for manufacturing of e-textile power lines

Seam tape	Fabric type		Temperature (℃)	Speed (ft/min)	Pressure (bar)	Pressure of hot air (bar)	Wheel pressure of blow (bar)
ST 604	PA 100%	Set 1	450	8.0
		Set 2	450	12.0	0.1	0.58	3.92
		Set 3	500	8.0	0.1	0.58	3.92
		Set 4	500	12.0
	PET 100%	Set 1	500	8.0
		Set 2	500	12.0	0.1	0.58	3.92
		Set 3	550	8.0	0.1	0.58	3.92
		Set 4	550	12.0
ST 306 & ST 318	PA 100%	Set 1	500	1.5	0.1	0.58	3.92
		Set 2	500 550	3.0 1.5
		Set 3	550	1.5
		Set 4	550	3.0 1.5
	PET 100%	Set 1	500	3.0	0.1	0.58	3.92
		Set 2	500	6.0
		Set 3	550	3
		Set 4	550	6

PA: polyamide; PET: polyester

In Figures 2 and 3, the manufacturing process of e-textile structures using an H&H AI-001 hot air welding machine and final e-textile samples are shown, respectively. As shown in Figure 2, the seam tape is sealed on a thermoplastic fabric while passing through the nip point of a stationery roller (at the bottom) and a rotating circular roller (at the top), where the hot air is blown and pressure applied. Thus, thermoplastic fabric is covered by a seam tape, as shown in Figure 3.

Figure 2.

Manufacturing welded e-textile structures via hot air welding machine. (Color online only.)

Figure 3.

Some of the e-textile transmission line samples manufactured by using 100% polyester (yellow) and 100% polyamide (navy) thermoplastic fabrics, welding tapes and conductive yarns. (Color online only.)

Following the production of samples, the resistance (R) values were measured using a Keithley® multimeter.

Testing set up to measure the surface temperature of e-textile structures

In order to test the surface temperature of e-textile samples with high precision and repeatability, the set up shown in Figure 4 was constructed under standardized conditions in a dark room compatible with AATCC 128,⁶⁷ where any external heat and light reflections are minimized. The set up includes an opaque 45 ° inclined background block inside a dark testing cubicle. The emissivity value of the opaque background block is estimated to have a value of 0.95. E-textile samples are placed on the opaque block and the ends of samples are connected to the power supply. Measurements are taken by a thermal camera located 35 ± 2 cm far away from sample surface; thus, the thermal camera was perpendicular enough to the inclined background block. Measurements were taken via Fluke SmartView® software through a Fluke Ti200 9H2 digital thermal camera at applied voltages of 7.4, 9 and 12 V, respectively, and the average, minimum and maximum temperature values are acquired within a precision of ± 0.005. Following the thermal camera measurements, the actual voltage and current passing through conductive yarn over the sample connected to the power supply were measured using a Keithley multimeter. During observations while applying potential difference to samples, it is found that after around 2 minutes, a relatively stable flow regime is reached where the temperature of the sample becomes almost constant within a precision of ± 0.005. Therefore, with the acquired current value and voltage value, heat dissipation is calculated using Ohm's law, as illustrated in Figure 1.

Figure 4.

Testing set up to measure electric resistance heating.

For instance, the thermal image of the e-textile sample together with the sample's original image is shown in Figure 5.

Figure 5.

Original e-textile samples (100% polyester – yellow; 100% polyamide – navy) and their thermal images acquired by a thermal camera. (Color online only.)

Analysis of experimental uncertainty

The uncertainty analysis is taken into consideration by determining the main error sources in the experiments. Error sources are related to the measurements of the following quantities: electric voltage (potential) and current. The heat applied to the system is calculated by the multiplication of voltage and current. The absolute error in electric current measurement is 0.01 A and the accuracy of electric potential measurements is 0.2 V. For every experimental case, 20 readings are taken to collect necessary data from any sensor on the system after the system reaches the thermal steady state. For each one of the measurement points a relative convergence criterion of 3% between the consecutive temperature readings (about 5 minutes) is strictly sought. Once the propagation of error is investigated for data no 228 (Table 3), heat dissipation has 0.07911 Watt uncertainty. Dissipated power uncertainty is mostly due to the current measurement by 86.56% and 13.44% due to voltage measurements, which are evaluated with error sensitivity analysis. In all experiments, until the steady-state phase, various data is collected by the voltage readouts on the power supply and multimeter (voltage, current and resistance). For the present experiments, electrical power is fed by a Tektronix PWS 4602 Programmable direct current (DC) power supply in order to reduce magnetic interference, and in order to eliminate effects due to the inductance and capacitance of conductive yarns. The test sample is assumed to transform all electrical energy into thermal energy and dissipate it, which is provided by the supplied DC power, by the thermal precautions of opaque blocks and by the installed layered insulation tapes. The overall accuracy of the present thermal measurements is better than that of commercial thin-heat flux sensor⁶⁸ specs, which is given as 6.5 µV/Btu/Ft2Hr (equivalent to 2.06 µV per unit heat flux (W/m²), that is, 2.06 µV/(W/m²)) for the present small heat transfer area and low heat flux experiments.

Table 3.

A small experimental set

Data no.	INPUTS						OUTPUTS
Data no.	Sample type	Temperature (℃)	Feeding speed (ft/min)	Resistance of sample (Ω)	Applied voltage (V)	Current (A)	Tested surface temperature (℃)	Electric power dissipated (W)
1	Polyamide - ST318	500	1.5	18	7.4	0.41	29.04	3.04
2	Polyamide - ST318	500	1.5	18	9.0	0.50	32.59	4.50
37	Polyamide - ST604	450	8.0	18	7.4	0.41	29.55	3.04
38	Polyamide - ST604	450	8.0	18	9.0	0.50	35.26	4.50
85	Polyester - ST306	500	3.0	39	7.4	0.19	27.99	1.40
86	Polyester - ST306	500	3.0	39	9.0	0.23	30.56	2.07
177	Polyester - ST604	550	12	18	7.4	0.41	33.71	3.04
178	Polyester - ST604	550	12	18	9.0	0.50	38.43	4.50
217	Polyester - ST318	550	6.0	30	12.0	0.40	28.03	4.80
…	…	…	…	…	…	…	…	…
228	Polyester - ST318	550	6.0	51	7.4	0.14	39.39	1.07

Artificial neural network model

In this study, the Neural Network Toolbox of MATLAB® 7.10 (R2010a) mathematical software was used to predict average surface temperatures and electric power dissipations on e-textile structures under different applied voltage values. The input variables are Sample Type, Temperature (℃), Feeding Speed (ft/min), Resistance of Sample (Ω), Applied Voltage (V) and Current (A) and the outputs are Average Surface Temperature and Electric Power Dissipated. A small experimental set is shown in Table 3. Here it should be noted that the effects of selected inputs on each of the outputs is highly significant (p < 0.05). Experimental results mainly showed that as the linear resistance increases, the current passing through conductive yarns decreases when the voltage is constant, as expected. Although each input has a significant effect on the resulting average surface temperature and heat dissipated over the e-textile structure, a considerable effect is obtained with inputs of linear resistance and current passing through the samples, as well as applied voltage.

The input–output data can be actual or normalized. It is obvious that using normalized data leads to better results. The normalized dataset of laboratory experiments is obtained with Equation (1), where X is normalized data, Xi is actual data, Xmin is the minimum value of actual data and Xmax is the maximum value of actual data

X = (Xi - X min) / / (X max - X min)

(1)

Eighty percent of 228 experimental sets are used for training to the ANN and 20% of this set is used for validation of the proposed model. A small normalized experimental set is shown in Table 4.

Table 4.

A small normalized experimental set

Data no.	INPUTS						OUTPUTS
Data no.	Sample type	Temperature (℃)	Feeding speed (ft/min)	Resistance of sample (Ω)	Applied voltage (V)	Current (A)	Tested surface temperature (℃)	Electric power dissipated (W)
1	0.1	0.50	0.041667	0.276596	0.066667	0.141414	0.129143	0.078547
2	0.1	0.50	0.041667	0.276596	0.333333	0.181818	0.179857	0.134615
37	0.3	0.25	0.583333	0.276596	0.06667	0.141414	0.136428	0.078547
38	0.3	0.25	0.583333	0.276596	0.333333	0.181818	0.218000	0.134615
85	0.5	0.50	0.16667	0.723405	0.066667	0.040793	0.114142	0.015542
86	0.5	0.50	0.16667	0.723405	0.333333	0.059441	0.150857	0.041420
177	0.7	0.75	0.916667	0.276596	0.066667	0.141414	0.195857	0.078547
178	0.7	0.75	0.916667	0.276596	0.333333	0.181818	0.263285	0.134615
217	0.9	0.75	0.416667	0.531910	0.833333	0.136363	0.114714	0.146153
…	…	…	…	…	…	…	…	…
228	0.9	0.75	0.416667	0.9787234	0.066667	0.020499	0.277	0.002835

In this study, the feed forward back propagation ANN model is preferred in our multi-layered ANN structure since it is a commonly used approximator and best fitted with the dataset of the ANN model. Levenberg Marquardt⁶⁹ is used as a training algorithm in the proposed feed forward back propagation ANN model. The Gradient Descent with Momentum (GDM) learning algorithm is applied for the learning algorithm in MATLAB® software. Variables are normalized between 0 and 1, so LOGSIG (Log-Sigmoid), the most widely used transfer function (Equation (2)), is preferred for the ANN model

a = logsig (n) = \frac{1}{1 + e^{- n}}

(2)

As a result of tests and analysis, the network's optimum topology has been obtained with a specific iteration. The proposed ANN models consisted of six-neuron input layers that represent inputs, a hidden layer that is made of five and 10 neurons and finally the output layer that is made of two neurons. A sample structure that represents the ANN's input, output and hidden layers (consisting of five neurons) is indicated in Figure 6. Then two proposed ANN models have been trained. Results are simulated and compared with the actual experimental data.

Figure 6.

A sample structure of the proposed artificial neural network model (five neurons in the hidden layer).

Results

ANN outputs are compared with the actual values after training has been completed. During comparison of data, Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE) and Mean Square Error (MSE) were selected as a type of error and they are calculated according to Equations (3)–(5) for validation of the proposed model. Proposed model results are shown in Table 5

MAE = \frac{1}{n} \sum_{t = 1}^{n} | At - Ft |

(3)

MAPE = \frac{100 %}{n} \sum_{t = 1}^{n} | \frac{At - Ft}{At} |

(4)

MSE = \frac{1}{n} \sum_{t = 1}^{n} (At - Ft)^{2}

(5)

where At is actual data, Ft is forecast at time t and n is the number of samples.

Table 5.

Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE) and Mean Square Error (MSE) values of the two developed artificial neural network models

	Tested surface temperature (℃)		Electric power dissipated (W)
Ratio of training dataset	80%
Number of neurons in hidden layer	5	10	5	10
MAE	0.06037	0.05941	0.02045	0.01472
MAPE (%)	20.86000	6.67310	20.07000	14.19000
MSE	0.00943	0.00569	0.00081	0.00041

Figures 7 and 8 show comparisons between experimental and predicted values of the output variables. Correlation coefficient values (R²) and regression equations of the two proposed ANN models are given in the related figure.

Figure 7.

Correlation between experimental and predicted values for the tested surface temperature.

Figure 8.

Correlation between experimental and predicted values for electric power dissipated.

The plot in Figure 7 has correlation coefficients of 0.7399 and 0.8423 for five and 10 neurons, respectively, for prediction of the surface temperature. Moreover, the MSE values were found to be 0.009435 and 0.005692 for five and 10 neurons, respectively. On the other hand, correlation coefficients for prediction of electric power dissipated are about 0.996 for both five and 10 neurons, presenting an almost perfect fit with MSE values of lower than 0.001 (Figure 8). These results confirmed that the neural network model reproduces electric power dissipated values for this system within the experimental ranges adopted in the fitting model.

Conclusions

The potential surface temperature distribution on powered e-textile structures for electric resistance heating was investigated in this study. An ANN model is constructed to derive the surface temperature of e-textile structures. E-textile structures were created using different conductive yarns via welding technology. The measurements were conducted with a set up including a power supply and a thermal camera under different voltage values. Experimental results show that as the linear resistance increases, the current passing through conductive yarns decreases; thus, the resulting heat dissipation and surface temperature over the e-textile structure mainly depends on the resistance, current and applied voltage values. The real test results are used for modeling surface temperature distributions as well as electric power dissipated on powered e-textile structures using the ANN. The straight line obtained for the prediction of electric power dissipated model fits well with the experimental equilibrium data. The correlation coefficient values showed that the ANN model successfully predicted dissipated electric power and surface temperature on e-textile structures by applying a three-layered neural network architecture with 10 neurons in the hidden layer and using a feed forward back propagation algorithm. As a result, it can be concluded that since the values of the determination coefficient and the MSE were found to be 0.996 and less than 0.001, respectively, for both five and 10 neurons in the hidden layer, the ANN model developed for prediction of electric power dissipated was very successful and can be useful for e-textile product designers as well as textile manufacturers, particularly for cold weather protection products, such as jackets, gloves and outdoor sleeping mats. Depending on the end use of the e-textile, careful selection of conductive yarn considering its linear resistance value is crucial. With the ANN model developed, depending on the protection level against cold weather, suitable conductive yarn type and the required voltage can be identified and, thus, the heat dissipation on the textile surface can be predicted as part of the pre-design work.

Footnotes

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Ministry of Science, Industry and Technology, Republic of Turkey (Grant Agreement No. 0034.TGSD.2015-2) “Design of E-textile Based Thermal Heating Panels via Welding Technology”.

References

Wilkerson JB, Ruixiu S, Hart WE, et al. Development of a variable rate nitrogen applicator based on spectral reflectance characteristics of cotton plants. In: Dugger P (ed.) proceedings of the beltwide cotton conference, 5–9 January 1998, Vol.2, San Diego, CA: National Cotton Council of America, pp.439–443.

Leonard

Pirotte

Knott

. Classification of second hand textile waste based on near infrared analysis and neural network. Melliand Int 1998; 4: 242–244. .

Jasper

Kovacs

. Using neural networks and NIR spectrophotometry to identify fibers. Text Res J 1994; 64: 444–448.

She

Kong

Nahavandi

et al.

Animal fibre classification. Text Asia 2001; 32: 44–47.

Allan

. An automatic analysis system for natural fibres. In: Ghandi KL, Cusick GE (eds.) proceedings of world textile congress on natural polymer fibres, Huddersfield, UK: University of Huddersfield, 9–11 July 1997, pp. 220–230.

Sankaran V, Lee D, Hitchcock EM, et al. Automated fibre identification and analysis system. In: proceedings of the 78th world conference of the Textile Institute, Thessaloniki, Greece, 25–26 May 1997, Vol.2, Manchester, UK: The Textile Institute. pp.325–345.

Ulmer

Smith

Sumpter

et al.

Computational neural networks and the rational design of polymeric materials: the next generation polycarbonates. Comp Theo Poly Sci 1998; 8: 311–321.

De Weijer T. Prediction of yarn properties using natural computation. In: proceedings of 37th international man-made fibres congress, Dornbirn, Austria, 16–18 September 1998, pp.163–168. Österreichisches Chemiefaser-Institut Wien.

Chiu

Chen

et al.

Appearance analysis of false twist textured yarn packages using image processing and neural network technology. Text Res J 2001; 71: 313–317.

10.

Pynckels

Kiekens

Sette

et al.

Use of neural nets for determining the spinnability of fibres. J Text Inst 1995; 86: 425–437.

11.

Shiau

Tsai

Lin

. Classifying web defects with a back propagation neural network by color image processing. Text Res J 2000; 70: 633–640.

12.

Wu P, Fang SC, Nuttle HLW, et al. Guided neural network learning using a fuzzy controller, with applications to textile spinning, http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.39.6727&rep=rep1&type=pdf (accessed 10 December 2016).

13.

Zeng

Wang

. Predicting the tensile properties of air-jet spun yarns. Text Res J 2004; 74: 689–694.

14.

Ramesh

Rajamanickam

Jayaraman

. The prediction of yarn tensile properties using by using artificial neural networks. J Text Inst 1995; 86: 459–469.

15.

Cheng

Adams

. Yarn strength prediction using neural networks- Part 1: fibre properties and yarn strength prediction. Text Res J 1995; 65: 495–500.

16.

Ramesh

Rajamanickam

Jayaraman

. The prediction of yarn tensile properties by using artificial neural networks. J Text Inst 1995; 86: 459–569.

17.

Cheng

Adams

. Yarn strength prediction using neural networks. Part I: fibre properties and yarn strength relationship. Text Res J 1995; 65: 495–500.

18.

Fan

Hunter

. A worsted fabric expert system. Part I: system development. Text Res J 1998; 68: 680–686.

19.

Fan

Hunter

. A worsted fabric expert system. Part II: an artificial neural network model for predicting the properties of worsted fabrics. Text Res J 1998; 68: 763–771.

20.

Ertugrul

Ucar

. Predicting bursting strength of cotton plain knitted fabrics using intelligent techniques. Text Res J 2000; 70: 845–851.

21.

Mori

Komiyama

. Evaluating wrinkled fabrics with image analysis and neural networks. Text Res J 2002; 72: 417–422.

22.

Kim

. Objective evaluation of wrinkle recovery. Text Res J 1999; 69: 860–865.

23.

Okamoto

Nakajima

Hosokawa

. Measurement of fabric hand evaluation values by neural network based on PCA of drape images. Memoirs Facult Eng 1996; 37: 89.

24.

Fan

Newton

et al.

Predicting garment drape with a fuzzy-neural network. Text Res J 2001; 71: 605–608.

25.

Jeong

Kim

Hong

. Selecting optimal interlinings with a neural network. Text Res J 2000; 70: 1005–1010.

26.

Fan

Wang

Cheng

et al.

Fabric classification based on recognition using a neural network and dimensionality reduction. Text Res J 1998; 68: 179–185.

27.

Jasper

Potlapalli

. Image analysis of miss picks in woven fabrics. Text Res J 1995; 65: 683–692.

28.

Tsai

. Automatic inspection of fabric defects using ANN. Text Res J 1996; 66: 474–482.

29.

Chen

Liang

Yau

et al.

Classifying textile faults with a back propagation neural network using power spectra. Text Res J 1998; 68: 121–126.

30.

Beck

Jasper

Lee

et al.

Real time analysis and control of batch dyeing processes. Int Text Bull 1998; 98: 88–93.

31.

Faes

. Neural networks in colorimetry. Melliand Textilberichte 1998; 79: 462–465.

32.

Bezerra

Hawkyard

. Computer match prediction for fluorescent dyes by neural networks. J Soc Dye Col 2000; 116: 163–169.

33.

Huang

. Fuzzy neural network approach for classifying dyeing defects. Text Res J 2001; 71: 100–104.

34.

Banumathi

Sree

Priya

SPV

. Artificial intelligence techniques in textile fabric inspection. IJCSN 2015; 4: 793–798.

35.

Boong

. Automatic recognition of woven fabric patterns by artificial neural network. Text Res J 2003; 73: 645–650.

36.

Barrett

Clapp

Titus

. An on-line fabric classification technique using a wavelet based neural network approach. Text Res J 1996; 66: 521–528.

37.

Stylios

Sotomi

. A neuro-fuzzy control system for intelligent overlock sewing machines. Int J Cloth Sci Tech 1995; 7: 49–55.

38.

Stylios

Sotomi

. Thinking sewing machines for intelligent garment manufacture. Int J Cloth Sci Tech 1996; 8: 44–55.

39.

Westland

Bishop

Bushnell

et al.

An intelligent approach to colour recipe prediction. J Soc Dye Col 1991; 107: 235–237.

40.

Jasper

Kovacs

Berkstresser

. Using neural networks to predict dye concentrations in multiple-dye mixtures. Text Res J 1993; 63: 545–551.

41.

Vangheluwe

Sette

Pynckels

. Using neural networks to predict dye concentrations in multiple-dye mixtures. Text Res J 1994; 64: 182–183.

42.

Marjoniemi

Mantysalo

. Neuro-fuzzy modelling of spectroscopic data. Part A- modelling of dye solutions. J Soc Dye Col 1997; 113: 13–17.

43.

Marjoniemi

Mantysalo

. Neuro-fuzzy modelling of spectroscopic data. Part B - dye concentration prediction. J Soc Dye Col 1997; 113: 64–67.

44.

Elemen

Kumbasar

EPA

Yapar

. Modeling the adsorption of textile dye on organoclay using an artificial neural network. Dyes Pigm 2012; 95: 102–111.

45.

Leśnikowski

. Textile transmission lines in the modern textronic clothes. Fibres Text East Eur 2011; 19: 89–93.

46.

Chapter 2: conductive polymer yarns for electronic textiles. In: Dias

(ed). Electronic textiles, Cambridge, UK: Woodhead Publishing, 2015. .

47.

Uttam

. E-textiles: a review. Int J IT Eng Appl Sci Res 2014; 3: 8–10.

48.

Wilusz

Carta

Jourand

et al.

Military textiles, Cambridge: Woodhead Publishing, 2008.

49.

Vanfleteren

Puers

. Design and implementation of advanced systems in a flexible-stretchable technology for biomedical applications. Sens Act A Phys 2009; 156: 79–87.

50.

Berzowska

. Electronic textiles: wearable computers, reactive fashion and soft computation. Textile 2005; 3: 2–19.

51.

Patel PC, Vasavada DA and Mankodi HR. Applications of electrically conductive yarns in technical textiles. In: Nirmal N (ed.) IEEE international conference on power system technology (POWERCON), Auckland, New Zealand, 30 October–2 November 2012, Paper 306, Auckland, New Zealand: IEEE. pp.1–6.

52.

Merritt

Nagle

Grant

. Fabric-based active electrode design and fabrication for health monitoring clothing. IEEE Trans Inform Technol Biomed 2009; 13: 274–280.

53.

Shyr

Shie

Jiang

et al.

Textile-based wearable sensing device designed for monitoring the flexion angle of elbow and knee movements. Sensors 2014; 14: 4050–4059.

54.

Linz T, Gourmelon L and Langereis G. Contactless EMG sensors embroidered onto textile. In: Leonhardt S, Falck T, Mähönen P (eds.) 4th International Workshop on Wearable and Implantable Body Sensor Networks (BSN 2007). IFMBE Proceedings, vol 13. Berlin, Heidelberg: Springer, 2007, pp.29–34.

55.

Gorgutsa

Skorobogatiy

. A woven 2D touchpad sensor and a 1D slide sensor using soft capacitor fibers. Smart Mater Struct 2012; 21: 015010.

56.

Gauvreau

Guo

Schicker

et al.

Color-changing and color-tunable photonic bandgap fiber textiles. Opt Express 2008; 16: 15677–15693.

57.

Sayed

Berzowska

Skorobogatiy

. Jacquard-woven photonic bandgap fiber displays. Res J Text. Apparel 2010; 14: 97–105.

58.

EXO2 the heat inside. http://www.exo2theheatinside.com/technology.html (accessed 25 June 2017).

59.

et al.

A novel design method for intelligent clothing based on garment design and knitting technology. Text Res J 2009; 79: 1670–1679.

60.

Linz T, Kallmayer C, Aschenbrenner R, et al. Embroidering electrical interconnects with conductive yarn for the integration of flexible electronic modules into fabric. In: Shimizu E and Kidode M (eds.), Ninth IEEE international symposium on wearable computers, 18–21 October 2005, Osaka, Japan: IEEE. pp.86-91.

61.

Lee

Park

et al.

The use of carbon/dielectric fiber woven fabrics as filters for electromagnetic radiation. Carbon 2009; 47: 1896–1904.

62.

Locher

. Enabling technologies for electrical circuits on a woven monofilament hybrid fabric. Text Res J 2010; 78: 583–594.

63.

Alagirusamy

Eichhoff

Gries

et al.

Coating of conductive yarns for electro-textile applications. J Text Inst 2013; 104: 270–277.

64.

Kursun Bahadir

Kalaoglu

Jevsnik

. The use of hot air welding technologies for manufacturing e-textile transmission lines. Fiber Polym 2015; 16: 1384–1394.

65.

Mbise

Dias

Hurley

Chapter 6: design and manufacture of heated textiles. In: Dias

(ed.). Electronic textiles, Cambridge, UK: Woodhead Publishing, 2015. .

66.

Paynter

Boydell

BJT

. Introduction to electricity, Upper Saddle River, NJ/London: Prentice Hall/Pearson Education, 2011.

67.

AATCC 128. Wrinkle recovery of fabrics: appearance method, 2013.

68.

Omega. https://www.omega.com/pptst/HFS-3_HFS-4.html (accessed 5 October 2017).

69.

Moré

. The Levenberg-Marquardt algorithm: Implementation and theory. Numerical analysis, Berlin, Heidelberg: Springer, 1978, pp. 105–116.