Deep learning driven thermochromic color-changing flexible temperature sensor and method of making

Abstract

The application discloses a deep learning driven thermochromic flexible temperature sensor and a preparation method thereof, and steps are as follows: 1) dividing a target temperature into three intervals, and selecting three phase change materials matching phase change temperatures; 2) mixing the three phase change materials with leuco agents and color developing agents respectively to prepare red, green and blue thermochromic materials, and encapsulating the thermochromic materials into reversible thermochromic microcapsules with core-shell structures having transparent shell layers through an in-situ polymerization method; 3) blending the three-color reversible thermochromic microcapsules, coating a fabric surface through silk screen printing, drying, or weaving after melt spinning of composite fibers, and forming FSF; 4) shooting color change pictures of the FSF at different temperatures; 5) inputting picture data into a deep learning model for training; and 6) assembling the FSF with the trained model to obtain an FTS. The application takes the colorful color change microcapsules as a core, combines two fabric preparation processes, realizes multi-color signal recognition and temperature quantitative monitoring through deep learning, and has the advantages of flexibility, stability, a wide temperature range, high-precision visual temperature measurement capability and the like.

Description

Technical Field

[0001] This invention relates to the field of flexible temperature sensor technology, and in particular to a deep learning-driven thermochromic flexible temperature sensor and its fabrication method. Background Technology

[0002] With the rapid development of smart wearable devices, medical and health monitoring, and personalized protection, flexible temperature sensors have become an important research direction in smart textiles and flexible electronics. Temperature is a crucial parameter for human health, environmental safety, and equipment operating status; the real-time and accurate monitoring of temperature is of great significance for disease early warning, heat risk protection, and the development of smart clothing functions. Traditional electronic temperature sensors rely on electrical signal transmission. While they offer high accuracy, they generally suffer from rigid structures, poor comfort, high power consumption, and susceptibility to electromagnetic interference, making it difficult to meet the application requirements of fabric-based, flexible, and long-term monitoring.

[0003] In recent years, thermochromic materials have attracted widespread attention due to their ability to produce intuitive color responses during temperature changes. By introducing thermochromic microcapsules into flexible substrates or fabrics, passive and visualized temperature monitoring can be achieved. However, existing thermochromic systems generally suffer from problems such as limited color rendering range, single response signal, insufficient resolution, and susceptibility to ambient light interference, making it difficult to meet the needs of refined temperature identification. Summary of the Invention

[0004] In view of the aforementioned defects or deficiencies in the existing technology, it is desirable to provide a deep learning-driven thermochromic flexible temperature sensor and its fabrication method. The sensor uses thermochromic microcapsules as the core functional unit, wherein the thermochromic microcapsules employ a transparent shell to encapsulate the thermochromic core material, reducing the shell's obstruction and scattering of the color signal and improving color saturation and distinguishability. A temperature-responsive layer is constructed by combining a flexible substrate material with screen printing technology. Alternatively, the thermochromic microcapsules can be blended with a polymer matrix material, melt-spun to prepare composite fibers, and then processed into fabrics through weaving. Finally, a deep learning algorithm is used to achieve intelligent recognition of multi-color signals and quantitative temperature monitoring, maintaining the fabric's flexibility and stability while achieving wide-range, continuous, and high-precision visualized temperature measurement.

[0005] The deep learning-driven thermochromic flexible temperature sensor and its fabrication method of the present invention include the following steps: 1) Set the target temperature range and divide it into three temperature intervals; obtain the three phase change materials whose phase change temperatures correspond one-to-one with the three temperature intervals; 2) Mix the three phase change materials with the leuco and chromogenic agents in a predetermined ratio, and stir them in a constant temperature water bath until uniform to prepare red, green and blue thermochromic materials for later use. Use in-situ polymerization to encapsulate the red, green and blue thermochromic materials with transparent shells to form red, green and blue reversible thermochromic microcapsules with complete core-shell structures. 3) Red, green and blue reversible thermochromic microcapsules are blended in a predetermined ratio to obtain thermochromic multicolor microcapsules. The thermochromic multicolor microcapsules are printed on the surface of fabric and dried to obtain FSF; or thermochromic multicolor microcapsules are melt-spun with a polymer matrix material in a certain ratio to obtain composite fiber, and then FSF is prepared by weaving process. 4) Place the FSF at different temperatures and take thermochromic images of the FSF for later use; 5) Input the experimental data of thermochromic images into the deep learning model for model training; 6) Assemble the FSF and the trained deep learning model to obtain the deep learning-driven FTS.

[0006] Furthermore, in step 2), The phase change material is selected from at least one of the following: fatty alcohols, paraffins, fatty acids, and polyethylene glycols; the phase change material obtains a phase change temperature within a corresponding temperature range by adjusting the type and / or proportion of its raw materials. The colorimetric agent is selected from any one of bisphenol A, bisphenol F, bisphenol S, bisphenol AF, and hydroquinone; The leuco agent includes: a red leuco agent, a green leuco agent, and a blue leuco agent; wherein the red leuco agent is selected from any one of: thermored red, cresol red, pH-3, pH-5, and F-3; the green leuco agent is selected from any one of: malachite green lactone, thermored green, pH-2, F-4, F-5, and F-9; and the blue leuco agent is selected from any one of: crystal violet lactone, pH-1, pH-4, and pyridine blue. The color developer, the first phase change material, and any one of the red leuco, green leuco, and blue leuco are combined to obtain the first thermochromic material, wherein the phase change temperature of the first phase change material is in the lowest range among the selected phase change materials. The color developer, the second phase change material, and any two of the remaining two leuco agents among the red, green, and blue leuco agents are combined to obtain the second thermochromic material, wherein the phase change temperature of the second phase change material is higher than that of the first phase change material. The color developer, the third phase change material, and the last of the red, green, and blue leuco agents are combined to obtain a third thermochromic material, wherein the phase change temperature of the third phase change material is higher than that of the second phase change material, and the phase change temperature of the third phase change material is in the highest range among the selected phase change materials.

[0007] Furthermore, in step 2), the preparation of the reversible thermochromic microcapsules includes the following steps: 21) Add the thermochromic material to the styrene-maleic anhydride copolymer emulsifier and stir to form an O / W solution system; 22) The transparent shell material is selected from any one of the following: polymethyl methacrylate, polyethylene copolymer resin, melamine-formaldehyde resin, melamine resin, urea-formaldehyde resin, polyurethane, and silica; 23) The O / W solution system was added to the melamine-formaldehyde resin prepolymer solution, stirred and heated to polymerize, and then filtered, washed and dried to obtain reversible thermochromic microcapsules.

[0008] Furthermore, in step 3), the preparation of the thermochromic microcapsules and FSF includes the following steps: 311) Red, green, and blue reversible thermochromic microcapsules were blended according to different application scenarios to obtain thermochromic multicolor microcapsules, and the mass percentage of red, green, and blue reversible thermochromic microcapsules in the thermochromic multicolor microcapsules was 20%~50%; 312) The binder, thickener, deionized water and the thermochromic microcapsules are mixed to obtain thermochromic ink; 313) Thermochromic ink is printed onto the surface of the fabric using a screen printing plate and then dried to obtain FSF.

[0009] Furthermore, in step 3), the preparation of the composite fiber and FSF includes the following steps: 321) The polymer matrix material is selected from any one of the following: polyamide, polyester, polypropylene, polyethylene, polystyrene, polycarbonate, polylactic acid, thermoplastic polyurethane, and ethylene-vinyl alcohol copolymer; 322) 5 wt.% of thermochromic microcapsules were blended with 95 wt.% of polymer matrix material and composite fibers were prepared by melt spinning process; 323) FSF is prepared by stretching and twisting composite fibers and then processing them into fabrics through weaving or knitting processes.

[0010] Furthermore, in step 4), the steps for capturing the thermochromic image of the FSF are as follows: 41) Attach the FSF to the surface of the hot stage, and then place the hot stage with the FSF attached under a standard light source; Specifically, when the FSF presents a mixed effect of three colors, the heating stage begins to heat up; when the FSF presents a white or colorless effect, the heating stage stops heating up. 42) The heating stage is gradually heated by 0.1℃ each time. After each temperature increase, the temperature is stabilized for 5 minutes, and then the FSF is sampled and photographed using a camera.

[0011] Furthermore, in step 5), the deep learning model can identify multiple color levels and display the corresponding temperature within 50ms, with an accuracy of over 99%; model training includes the following steps: 51) By constructing different lighting conditions through multiplicative brightness perturbation and adding Gaussian noise, affine transformation is used to generate rotated samples from thermochromic images; 52) Input the rotated samples into the deep learning model and construct the feature extraction backbone. Insert a 512-dimensional bottleneck layer into the classification head and use the Hardswish activation function to enhance the nonlinear representation. 53) In the initial 50 rounds of freezing the feature extraction layer, the deep learning model uses a learning rate of 1e -4 Fine-tune the classification header; 54) After unfreezing the entire deep learning model, switch to the Adam optimizer and use a cosine annealing scheduler to dynamically adjust the learning rate. 55) Introduce deep learning models into label smoothing techniques, and then enhance the generalization of decision boundaries by softening the distribution of real labels.

[0012] Furthermore, the FSF undergoes graded color changes within corresponding temperature ranges, including the following color change ranges: When the temperature is below the phase change temperature range of the first phase change material, the red, green and blue reversible thermochromic microcapsules are all in a colored state, and the FSF exhibits a mixed effect of the three colors. When the temperature is within the phase change temperature range of the first phase change material, the reversible thermochromic microcapsules in the corresponding temperature range fade, and the FSF exhibits a mixed effect of the remaining two colors. When the temperature is within the phase change temperature range of the second phase change material, the reversible thermochromic microcapsules in the corresponding temperature range fade, and the FSF exhibits the effect of the last color. When the temperature is within the phase change temperature range corresponding to the third phase change material, the reversible thermochromic microcapsules in the corresponding temperature range fade, and FSF exhibits a white or colorless effect. As the temperature decreases, the colors of the thermochromic microcapsules gradually revert, and FSF exhibits a mixed effect of three colors.

[0013] Furthermore, in step 6), The FTS can achieve graded reversible color-changing response of red, green and blue within the target temperature range; The FTS has a temperature resolution of 0.1°C, and after 10,000 color development and fading cycles, the color change rate does not exceed 5%. Furthermore, the color response and signal recognition capability of the FTS remain stable when bent between 0 and 180°.

[0014] In addition, the present invention also provides a thermochromic flexible temperature sensor prepared by the above-described deep learning-driven thermochromic flexible temperature sensor preparation method.

[0015] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention prepares thermochromic microcapsules with different color rendering temperature ranges and screen prints the thermochromic microcapsules onto flexible fabrics to obtain FSF; or blends thermochromic microcapsules with polymer matrix materials, prepares composite fibers through melt spinning, and further weaves them into fabric-type FSF, realizing a wide temperature range and multi-level color gradient response, significantly improving temperature resolution and the ability to identify subtle temperature differences; wherein, the microcapsules are coated with a highly transparent shell material to cover the thermochromic core material, which effectively reduces the shading and distortion of the color rendering light signal by the shell while ensuring the structural stability and environmental resistance of the microcapsules, making the color change more realistic, clear and repeatable; the FSF is combined with a trained deep learning model to perform intelligent analysis of complex multicolor signals, and can achieve high-precision and real-time temperature monitoring under various environmental conditions.

[0016] (2) The FSF prepared by the present invention maintains stable color development and recognition performance under bending, stretching and multiple thermal cycles, while also having excellent flexibility and wearability, making it suitable for applications such as intelligent health monitoring, personalized protection and intelligent textiles.

[0017] It should be understood that the description in the Summary of the Invention is not intended to limit the key or essential features of the embodiments of the present invention, nor is it intended to restrict the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0018] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 Flowchart of the fabrication process for a thermochromic flexible temperature sensor; Figure 2 A real-world image showing the color change of a thermochromic flexible temperature-sensing fabric at 30-50℃. Figure 3 A physical image of a thermochromic flexible temperature sensor assembled from a thermochromic fabric and a deep learning model. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0021] Please refer to Figures 1-3 The present invention provides a deep learning-driven thermochromic flexible temperature sensor and its fabrication method, comprising the following steps: 1) Set the target temperature range and divide it into three temperature intervals; obtain the three phase change materials whose phase change temperatures correspond one-to-one with the three temperature intervals; 2) Mix the three phase change materials with the leuco and chromogenic agents in a predetermined ratio, and stir them in a constant temperature water bath until uniform to prepare red, green and blue thermochromic materials for later use. Use in-situ polymerization to encapsulate the red, green and blue thermochromic materials with transparent shells to form red, green and blue reversible thermochromic microcapsules with complete core-shell structures. 3) Red, green and blue reversible thermochromic microcapsules are blended in a predetermined ratio to obtain thermochromic multicolor microcapsules. The thermochromic multicolor microcapsules are printed on the surface of a fabric and dried to obtain FSF (thermochromic flexible temperature sensing fabric); or thermochromic multicolor microcapsules are melt-spun with a polymer matrix material in a certain ratio to obtain composite fibers, and then FSF is prepared by weaving process. 4) Place the FSF (thermochromic flexible temperature sensing fabric) at different temperatures and take thermochromic images of the FSF (thermochromic flexible temperature sensing fabric) for later use; 5) Input the experimental data of thermochromic images into the deep learning model for model training; 6) Assemble the FSF (thermochromic flexible temperature sensing fabric) and the trained deep learning model to obtain the deep learning-driven FTS (thermochromic flexible temperature sensor).

[0022] In step 2), The phase change material is selected from at least one of the following: fatty alcohols, paraffins, fatty acids, and polyethylene glycols; the phase change material obtains a phase change temperature within a corresponding temperature range by adjusting the type and / or proportion of its raw materials. The colorimetric agent is selected from any one of bisphenol A, bisphenol F, bisphenol S, bisphenol AF, and hydroquinone; The leuco agent includes: a red leuco agent, a green leuco agent, and a blue leuco agent; wherein the red leuco agent is selected from any one of: thermored red, cresol red, pH-3, pH-5, and F-3; the green leuco agent is selected from any one of: malachite green lactone, thermored green, pH-2, F-4, F-5, and F-9; and the blue leuco agent is selected from any one of: crystal violet lactone, pH-1, pH-4, and pyridine blue. The color developer, the first phase change material, and any one of the red leuco, green leuco, and blue leuco are combined to obtain the first thermochromic material, wherein the phase change temperature of the first phase change material is in the lowest range among the selected phase change materials. The color developer, the second phase change material, and any two of the remaining two leuco agents among the red, green, and blue leuco agents are combined to obtain the second thermochromic material, wherein the phase change temperature of the second phase change material is higher than that of the first phase change material. The color developer, the third phase change material, and the last of the red, green, and blue leuco agents are combined to obtain a third thermochromic material, wherein the phase change temperature of the third phase change material is higher than that of the second phase change material, and the phase change temperature of the third phase change material is in the highest range among the selected phase change materials.

[0023] In step 2), the preparation of the reversible thermochromic microcapsules includes the following steps: 21) Add the thermochromic material to the styrene-maleic anhydride copolymer emulsifier and stir to form an O / W solution system; 22) The transparent shell material is selected from any one of the following: polymethyl methacrylate, polyethylene copolymer resin, melamine-formaldehyde resin, melamine resin, urea-formaldehyde resin, polyurethane, and silica; 23) The O / W solution system was added to the melamine-formaldehyde resin prepolymer solution, stirred and heated to polymerize, and then filtered, washed and dried to obtain reversible thermochromic microcapsules.

[0024] In step 3), the preparation of thermochromic microcapsules and FSF (thermochromic flexible temperature sensing fabric) includes the following steps: 311) Red, green, and blue reversible thermochromic microcapsules were blended according to different application scenarios to obtain thermochromic multicolor microcapsules, and the mass percentage of red, green, and blue reversible thermochromic microcapsules in the thermochromic multicolor microcapsules was 20%~50%; 312) The binder TF-321F, the thickener TF-321D, deionized water and the thermochromic microcapsules are mixed to obtain thermochromic ink; 313) Thermochromic ink is printed onto the surface of the fabric using a screen printing plate and then dried to obtain FSF (thermochromic flexible temperature sensing fabric). In step 4), the preparation of the composite fiber and FSF includes the following steps: 321) The polymer matrix material is selected from any one of the following: polyamide, polyester (PET, PBT, PTT), polypropylene, polyethylene, polystyrene, polycarbonate, polylactic acid, thermoplastic polyurethane, and ethylene-vinyl alcohol copolymer; 322) 5 wt.% of thermochromic microcapsules were blended with 95 wt.% of polymer matrix material and composite fibers were prepared by melt spinning process; 323) FSF is prepared by stretching and twisting composite fibers and then processing them into fabrics through weaving or knitting processes.

[0025] In step 4), the steps for capturing the thermochromic images of the FSF (thermochromic flexible temperature-sensing fabric) are as follows: 41) Attach FSF (thermochromic flexible temperature sensing fabric) to the surface of the hot stage, and then place the hot stage with FSF attached under a standard light source. Specifically, when the FSF (thermochromic flexible temperature sensing fabric) exhibits a mixed effect of three colors, the heating platform begins to heat up; when the FSF (thermochromic flexible temperature sensing fabric) exhibits a white or colorless effect, the heating platform stops heating up. 42) The heating stage is gradually heated by 0.1℃ each time. After each temperature increase, the temperature is stabilized for 5 minutes, and then the FSF is sampled and photographed using a camera.

[0026] In step 5), the deep learning model can identify multiple color levels and display the corresponding temperature within 50ms, with an accuracy of over 99%. Model training includes the following steps: 51) By constructing different lighting conditions through multiplicative brightness perturbation and adding Gaussian noise, affine transformation is used to generate rotated samples from thermochromic images; 52) Input the rotated samples into the deep learning model and construct the feature extraction backbone. Insert a 512-dimensional bottleneck layer into the classification head and use the Hardswish activation function to enhance the nonlinear representation. 53) In the initial 50 rounds of freezing the feature extraction layer, the deep learning model uses a learning rate of 1e -4 Fine-tune the classification header; 54) After unfreezing the entire deep learning model, switch to the Adam optimizer and use a cosine annealing scheduler to dynamically adjust the learning rate. 55) Introduce deep learning models into label smoothing techniques, and then enhance the generalization of decision boundaries by softening the distribution of real labels.

[0027] The FSF (thermochromic flexible temperature sensing fabric) undergoes graded color changes within corresponding temperature ranges, including the following color-changing ranges: When the temperature is below the phase change temperature range of the first phase change material, the red, green and blue reversible thermochromic microcapsules are all in a colored state, and the FSF (thermochromic flexible temperature sensing fabric) presents a mixed effect of the three colors. When the temperature is within the phase change temperature range of the first phase change material, the reversible thermochromic microcapsules in the corresponding temperature range fade, and the FSF (thermochromic flexible temperature sensing fabric) exhibits a mixed effect of the remaining two colors. When the temperature is within the phase change temperature range of the second phase change material, the reversible thermochromic microcapsules in the corresponding temperature range fade, and the FSF (thermochromic flexible temperature sensing fabric) presents the effect of the last color. When the temperature is within the phase change temperature range corresponding to the third phase change material, the reversible thermochromic microcapsules in the corresponding temperature range fade, and the FSF (thermochromic flexible temperature sensing fabric) appears white or colorless. As the temperature decreases, the colors of the thermochromic microcapsules gradually revert, and the FSF (thermochromic flexible temperature sensing fabric) exhibits a mixed effect of three colors.

[0028] In step 6), The FTS (thermochromic flexible temperature sensor) can achieve graded reversible color-changing response of red, green and blue within the target temperature range; The FTS (thermochromic flexible temperature sensor) has a temperature resolution of 0.1°C. After 10,000 color development and fading cycles, the color change rate does not exceed 5%. Furthermore, the color response and signal recognition capability of the FTS remain stable when bent between 0 and 180°.

[0029] In addition, the present invention also provides a thermochromic flexible temperature sensor prepared by the above-described deep learning-driven thermochromic flexible temperature sensor preparation method.

[0030] In this embodiment, thermochromic microcapsules with different color rendering temperatures are prepared, and the thermochromic microcapsules are compositely screen-printed onto flexible fabrics to obtain FSF (thermochromic flexible temperature sensing fabric). Alternatively, thermochromic microcapsules are blended with a polymer matrix material, melt-spun to prepare composite fibers, and further woven into fabric-type FSF (thermochromic flexible temperature sensing fabric). This achieves a wide temperature range and multi-level color gradient response, significantly improving temperature resolution and the ability to identify subtle temperature differences. The microcapsules are encapsulated with a highly transparent shell material that has high light transmittance. This ensures the structural stability and environmental resistance of the microcapsules while effectively reducing the obstruction and distortion of the color rendering light signal by the shell, making the color changes more realistic, clear, and repeatable. The FSF (thermochromic flexible temperature sensing fabric) combined with a trained deep learning model performs intelligent analysis of complex multicolor signals, enabling high-precision, real-time temperature monitoring under various environmental conditions.

[0031] The prepared FSF (thermochromic flexible temperature sensing fabric) maintains stable color development and recognition performance under bending, stretching and multiple rounds of thermal cycling conditions, while also having excellent flexibility and wearability, making it suitable for applications such as smart health monitoring, personalized protection and smart textiles.

[0032] Example 1

[0033] In this embodiment, three reversible thermochromic microcapsules (green, blue, and red) were prepared using in-situ polymerization. Thermosensitive green, bisphenol A, and dodecanol were uniformly stirred in a 70°C water bath; crystal violet lactone, bisphenol A, and hexadecyl alcohol were uniformly stirred in a 70°C water bath; simultaneously, thermosensitive red, bisphenol A, and docosyl alcohol were uniformly stirred in a 70°C water bath. The green, blue, and red ternary mixtures were added to deionized water, followed by the addition of a styrene-maleic anhydride copolymer solution. The mixtures were stirred in a 70°C water bath for 30 minutes (3500 rpm / min) to obtain an O / W emulsion system. Finally, melamine-formaldehyde resin prepolymer was added dropwise to the green, blue, and red O / W emulsions at a pump rate of 1.2 mL / min. The temperature was rapidly increased to 80°C and held for 60 minutes; the temperature was then further increased to 85°C and held for 60 minutes before stopping the reaction. The pH of the solution was adjusted to approximately 7, and after multiple water washings, the precipitate was dried.

[0034] Three reversible thermochromic microcapsules, comprising 40 wt.% green, 40 wt.% blue, and 20 wt.% red, were mixed to obtain thermochromic multicolor microcapsules. Then, the thermochromic multicolor microcapsules, binder TF-321F, thickener TF-321D, and deionized water were mixed to obtain thermochromic multicolor ink. Next, a 200-mesh screen was fixed onto a white polyester (PET) substrate fabric, and the thermochromic multicolor ink was transferred to the PET substrate using a scraper. The substrate was then heated and dried to prepare FSF (thermochromic flexible temperature sensing fabric).

[0035] The FSF (thermochromic flexible temperature sensing fabric) was attached to the surface of the heating stage. The heating stage was initially heated to 30°C and placed under a standard light source. The temperature was gradually increased by 0.1°C each time. After each temperature increase, the temperature was stabilized for 5 minutes. The FSF was then sampled and photographed using a digital camera. Finally, the temperature of the heating stage was increased to 50°C, and the surface of the FSF turned completely white.

[0036] The original dataset collected above is expanded using a multimodal enhancement pipeline to address issues such as illumination fluctuations and viewpoint shifts in practical applications: Bionic transformations are used to generate ±15° rotation samples to simulate pattern deformation during human movement; different illumination conditions are constructed using multiplicative brightness perturbations (coefficients 0.7~1.3); and Gaussian noise with a standard deviation of 5~20 is added to simulate sensor noise. A feature extraction backbone is built based on MobileNetV3-small, and the parameters of the first eight inverse residual modules are frozen to control overfitting risk. A dual module is introduced at the network's end: firstly, a 512-dimensional bottleneck layer is inserted into the classification head, using a Hardswish activation function to enhance nonlinear representation; secondly, a channel attention mechanism is added after the global pooling layer to increase the weight coefficients of temperature-sensitive regions.

[0037] The initial 50 rounds freeze the feature extraction layer, using only a learning rate of 1e. -4 Fine-tune the classification head to avoid gradient anomalies on small datasets. After unfreezing the entire network, switch to the Adam optimizer and dynamically adjust the learning rate (initial value 5e) using a cosine annealing scheduler. -5 Minimum decay to 1 e-6 The loss function incorporates label smoothing technology (smoothing factor 0.1) to enhance the generalization of the decision boundary by softening the true label distribution. Finally, after the MobileNetV3-small model is trained, it is combined with FSF (thermochromic flexible temperature sensing fabric) to form a deep learning-driven FTS (thermochromic flexible temperature sensor).

[0038] Example 2

[0039] The only difference between this embodiment and Embodiment 1 is that green, blue and red reversible thermochromic microcapsules are mixed according to their mass fractions, wherein the content of each reversible thermochromic microcapsule is 20wt.% to 50wt.%, and the sum of the mass fractions of the three is 100wt.%, thereby obtaining thermochromic multicolor microcapsules.

[0040] Example 3

[0041] The only difference between this embodiment and Example 1 is that 40 wt.% green, 40 wt.% blue, and 20 wt.% red reversible thermochromic microcapsules are mixed to obtain thermochromic multicolor microcapsules. Subsequently, 5 wt.% of the thermochromic multicolor microcapsules and 95 wt.% PET polymer are added to a high-speed mixer, stirred, and pulverized. Then, the mixture is melt-extruded and granulated using a micro twin-screw extruder. The operating temperatures of the preheating zone, melting zone, pressurization zone, and die heating zone of the micro twin-screw extruder are 160℃, 175℃, 190℃, and 195℃, respectively. The resulting masterbatch is then melt-spun at the same temperature to prepare composite phase change fibers. The resulting composite fibers are then drawn, twisted, and woven to obtain FSF (thermochromic flexible temperature sensing fabric).

[0042] Example 4

[0043] The difference between this embodiment and Embodiment 3 is that the three types of reversible thermochromic microcapsules—green, blue, and red—are mixed according to their mass fractions, with each type of reversible thermochromic microcapsule having a content of 20 wt.% to 50 wt.%, and the sum of their mass fractions being 100 wt.%, thereby obtaining thermochromic multicolor microcapsules.

[0044] In the description of this specification, the terms "one embodiment," "some embodiments," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0045] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for fabricating a deep learning-driven thermochromic flexible temperature sensor, characterized in that, Includes the following steps: 1) Set the target temperature range and divide it into three temperature intervals; obtain the three phase change materials whose phase change temperatures correspond one-to-one with the three temperature intervals; 2) Mix the three phase change materials with the leuco and chromogenic agents in a predetermined ratio, and stir them in a constant temperature water bath until uniform to prepare red, green and blue thermochromic materials for later use. Use in-situ polymerization to encapsulate the red, green and blue thermochromic materials with transparent shells to form red, green and blue reversible thermochromic microcapsules with complete core-shell structures. 3) Red, green and blue reversible thermochromic microcapsules are blended in a predetermined ratio to obtain thermochromic multicolor microcapsules. The thermochromic multicolor microcapsules are printed on the surface of fabric and dried to obtain FSF; or thermochromic multicolor microcapsules are melt-spun with a polymer matrix material in a certain ratio to obtain composite fiber, and then FSF is prepared by weaving process. 4) Place the FSF at different temperatures and take thermochromic images of the FSF for later use; 5) Input the experimental data of thermochromic images into the deep learning model for model training; 6) Assemble the FSF and the trained deep learning model to obtain the deep learning-driven FTS.

2. The method for fabricating a deep learning-driven thermochromic flexible temperature sensor according to claim 1, characterized in that, In step 2), The phase change material is selected from at least one of the following: fatty alcohols, paraffins, fatty acids, and polyethylene glycols; the phase change material obtains a phase change temperature within a corresponding temperature range by adjusting the type and / or proportion of its raw materials. The colorimetric agent is selected from any one of bisphenol A, bisphenol F, bisphenol S, bisphenol AF, and hydroquinone; The leuco agent includes: a red leuco agent, a green leuco agent, and a blue leuco agent; wherein the red leuco agent is selected from any one of: thermored red, cresol red, pH-3, pH-5, and F-3; the green leuco agent is selected from any one of: malachite green lactone, thermored green, pH-2, F-4, F-5, and F-9; and the blue leuco agent is selected from any one of: crystal violet lactone, pH-1, pH-4, and pyridine blue. The color developer, the first phase change material, and any one of the red leuco, green leuco, and blue leuco are combined to obtain the first thermochromic material, wherein the phase change temperature of the first phase change material is in the lowest range among the selected phase change materials. The color developer, the second phase change material, and any two of the remaining two leuco agents among the red, green, and blue leuco agents are combined to obtain the second thermochromic material, wherein the phase change temperature of the second phase change material is higher than that of the first phase change material. The color developer, the third phase change material, and the last of the red, green, and blue leuco agents are combined to obtain a third thermochromic material, wherein the phase change temperature of the third phase change material is higher than that of the second phase change material, and the phase change temperature of the third phase change material is in the highest range among the selected phase change materials.

3. The method for fabricating a deep learning-driven thermochromic flexible temperature sensor according to claim 1, characterized in that, In step 2), the preparation of the reversible thermochromic microcapsules includes the following steps: 21) Add the thermochromic material to the styrene-maleic anhydride copolymer emulsifier and stir to form an O / W solution system; 22) The transparent shell material is selected from any one of the following: polymethyl methacrylate, polyethylene copolymer resin, melamine-formaldehyde resin, melamine resin, urea-formaldehyde resin, polyurethane, and silica; 23) The O / W solution system was added to the melamine-formaldehyde resin prepolymer solution, stirred and heated to polymerize, and then filtered, washed and dried to obtain reversible thermochromic microcapsules.

4. The method for fabricating a deep learning-driven thermochromic flexible temperature sensor according to claim 1, characterized in that, In step 3), the preparation of thermochromic microcapsules and FSF includes the following steps: 311) Red, green, and blue reversible thermochromic microcapsules were blended according to different application scenarios to obtain thermochromic multicolor microcapsules, and the mass percentage of red, green, and blue reversible thermochromic microcapsules in the thermochromic multicolor microcapsules was 20%~50%; 312) The binder, thickener, deionized water and the thermochromic microcapsules are mixed to obtain thermochromic ink; 313) Thermochromic ink is printed onto the surface of the fabric using a screen printing plate and then dried to obtain FSF.

5. The method for fabricating a deep learning-driven thermochromic flexible temperature sensor according to claim 1, characterized in that, In step 3), the preparation of the composite fiber and FSF includes the following steps: 321) The polymer matrix material is selected from any one of the following: polyamide, polyester, polypropylene, polyethylene, polystyrene, polycarbonate, polylactic acid, thermoplastic polyurethane, and ethylene-vinyl alcohol copolymer; 322) 5 wt.% of thermochromic microcapsules were blended with 95 wt.% of a polymer matrix material and composite fibers were prepared by melt spinning. 323) FSF is prepared by stretching and twisting composite fibers and then processing them into fabrics through weaving or knitting processes.

6. The method for fabricating a deep learning-driven thermochromic flexible temperature sensor according to claim 1, characterized in that, In step 4), the steps for capturing the thermochromic image of FSF are as follows: 41) Attach the FSF to the surface of the hot stage, and then place the hot stage with the FSF attached under a standard light source; Specifically, when the FSF presents a mixed effect of three colors, the heating stage begins to heat up; when the FSF presents a white or colorless effect, the heating stage stops heating up. 42) The heating stage is gradually heated by 0.1℃ each time. After each temperature increase, the temperature is stabilized for 5 minutes, and then the FSF is sampled and photographed using a camera.

7. The method for fabricating a deep learning-driven thermochromic flexible temperature sensor according to claim 1, characterized in that, In step 5), the deep learning model can identify multiple color levels and display the corresponding temperature within 50ms, with an accuracy of over 99%. Model training includes the following steps: 51) By constructing different lighting conditions through multiplicative brightness perturbation and adding Gaussian noise, affine transformation is used to generate rotated samples from thermochromic images; 52) Input the rotated samples into the deep learning model and construct the feature extraction backbone. Insert a 512-dimensional bottleneck layer into the classification head and use the Hardswish activation function to enhance the nonlinear representation. 53) In the initial 50 rounds of freezing the feature extraction layer, the deep learning model uses a learning rate of 1e -4 Fine-tune the classification header; 54) After unfreezing the entire deep learning model, switch to the Adam optimizer and use a cosine annealing scheduler to dynamically adjust the learning rate. 55) Introduce deep learning models into label smoothing techniques, and then enhance the generalization of decision boundaries by softening the distribution of real labels.

8. The method for fabricating a deep learning-driven thermochromic flexible temperature sensor according to claim 6, characterized in that, The FSF undergoes graded color changes within corresponding temperature ranges, including the following color change ranges: When the temperature is below the phase change temperature range of the first phase change material, the red, green and blue reversible thermochromic microcapsules are all in a colored state, and the FSF exhibits a mixed effect of the three colors. When the temperature is within the phase change temperature range of the first phase change material, the reversible thermochromic microcapsules in the corresponding temperature range fade, and the FSF exhibits a mixed effect of the remaining two colors. When the temperature is within the phase change temperature range of the second phase change material, the reversible thermochromic microcapsules in the corresponding temperature range fade, and the FSF exhibits the effect of the last color. When the temperature is within the phase change temperature range corresponding to the third phase change material, the reversible thermochromic microcapsules in the corresponding temperature range fade, and FSF exhibits a white or colorless effect. As the temperature decreases, the colors of the thermochromic microcapsules gradually revert, and FSF exhibits a mixed effect of three colors.

9. The method for fabricating a deep learning-driven thermochromic flexible temperature sensor according to claim 1, characterized in that, In step 6), The FTS can achieve graded reversible color-changing response of red, green and blue within the target temperature range; The FTS has a temperature resolution of 0.1°C, and after 10,000 color development and fading cycles, the color change rate does not exceed 5%. Furthermore, the color response and signal recognition capability of the FTS remain stable when bent between 0 and 180°.

10. A thermochromic flexible temperature sensor prepared by the method described in any one of claims 1 to 9, which is a deep learning-driven thermochromic flexible temperature sensor.

Images