Supercritical foaming silica gel robot skin quality traceability management method

Abstract

The application discloses a kind of supercritical foaming silicone robot skin quality traceability management methods, belong to quality control technical field, specifically include: the full life cycle database of each piece of silicone sheet is established;Based on service sensitivity sequence and contact stress integral value calculation normalization attenuation coefficient;According to initial residual quantity partition, filter attenuation coefficient below threshold value's superior product sheet, extract its average temperature and shrinkage rate, form superior product shrinkage rate mapping parameter set;To each interval, time-temperature equivalence relationship fitting is carried out, and the shrinkage kinetics combined parameter set is established;With average temperature and shrinkage rate as input attenuation coefficient as output, train feedforward control strategy model;The initial residual quantity of new sheet is obtained, and the target setting temperature is determined according to the joint parameter set and model output of the interval. The application realizes the quantitative traceability and closed-loop optimization from the setting process to the service life, and improves the skin quality consistency.

Description

Technical Field

[0001] This invention relates to the field of quality control technology, specifically to a method for traceability and control of the quality of supercritical foamed silicone robotic skin. Background Technology

[0002] Supercritical foamed silicone materials are widely used as elastic base layers or protective layers for robotic tactile sensing interfaces due to their advantages such as lightweight, high elasticity, chemical stability, and wide temperature range adaptability. These materials, formed by physically foaming silicone rubber in a supercritical CO2 environment to create a microporous structure, effectively buffer external impacts and transmit tactile signals, making them a key component for achieving compliant robot interaction. As the requirements for tactile sensing accuracy and long-term stability in humanoid robots and service robots continue to increase, quality control of supercritical foamed silicone robot skin, especially its predictability and consistency in service life, has become a key focus of the industry.

[0003] Currently, quality traceability methods for supercritical foamed silicone robotic skin mainly focus on recording foaming process parameters, sampling inspection of cell structure (such as pore size and porosity), and tactile performance testing after sensor integration. Existing technologies typically control parameters such as foaming temperature, pressure, and depressurization rate within preset ranges, and perform sampling inspections on the product after foaming, recording its initial tactile sensitivity, linearity, and other indicators. They also link raw material and operator information through batch numbers. Some improved solutions also introduce basic recording of the ambient temperature during the settling stage and incorporate the settling time as a fixed parameter into the process specifications to achieve limited traceability of the production process. Furthermore, existing technologies collect fault feedback data during product service to make post-production adjustments.

[0004] However, existing quality traceability systems still have shortcomings in practical applications. First, the post-foaming aging stage is treated as a passive waiting process, only recording the ambient temperature or fixed duration, without measuring the initial residual amount of dissolved carbon dioxide inside the sheet, nor establishing a correlation between this residual amount and the subsequent overall thickness shrinkage rate. In fact, the initial CO2 residual amount varies significantly between different batches or different locations within the same batch due to factors such as depressurization rate and material thickness. This difference, along with the aging ambient temperature, determines the final shrinkage degree of the sheet, thus affecting the stress state of the cell walls and the rate of tactile performance degradation under long-term service. Second, existing technologies lack a closed-loop mechanism to optimize aging process parameters from actual service life data. Even if the service life of different sheets varies considerably, it is impossible to pinpoint the root cause through traceability data and make targeted adjustments to key parameters such as aging temperature.

[0005] Therefore, how to effectively record and utilize the coupling relationship between initial residual amount, resting temperature and shrinkage rate during the production process, and establish a quantitative traceability and feedback optimization method from resting process to service life, has become an urgent technical problem to be solved. Summary of the Invention

[0006] The purpose of this invention is to provide a method for traceability and control of skin quality in a supercritical foamed silicone robot, addressing the following technical problems:

[0007] Existing quality traceability systems do not consider the coupled effects of initial carbon dioxide residue and settling temperature on shrinkage rate and service life, and lack a closed-loop mechanism to optimize settling process parameters from the perspective of service life.

[0008] The objective of this invention can be achieved through the following technical solutions: A method for traceability and control of the skin quality of a supercritical foamed silicone robot includes the following steps: S1: Based on the static process data and service feedback data of supercritical foamed silicone sheets, historical sheet records are extracted to establish a full life cycle database of silicone skin; the static process data includes the unique identifier of the sheet, the initial residual amount of carbon dioxide, the static environment temperature sequence, and the overall thickness shrinkage rate at the end of static treatment; the service feedback data includes the initial tactile sensitivity, the service sensitivity sequence, and the contact stress integral value. S2: Extract the shrinkage rate evolution characteristics and restore the working condition attenuation coefficient from the full life cycle database of silicone skin, and construct a normalized attenuation coefficient based on the service sensitivity sequence and contact stress integral value; S3: In different initial residual amount ranges, select high-quality sheet samples with normalized decay coefficients lower than the preset threshold, extract the average static temperature and thickness shrinkage rate of the high-quality sheet samples, and integrate them to form a high-quality shrinkage rate mapping parameter set. S4: Deconstruct the high-quality shrinkage rate mapping parameter set, fit the time-temperature equivalent relationship between the average static temperature and the thickness shrinkage rate in different initial residual amount ranges, and establish a joint parameter set for shrinkage dynamics. S5: Construct the attribution mapping relationship between the joint parameter set of contraction dynamics and the normalized decay coefficient, and train and output a feedforward control strategy model based on contraction characteristics; S6: Obtain the real-time initial residual amount of the new batch of sheets and determine its range. Input the strategy model according to the joint parameter set corresponding to the range, and set the target static environment temperature of the new sheets according to the model output.

[0009] As a further aspect of the present invention: the specific process of establishing a silicone skin full life cycle database in S1 is as follows: A unique identifier is attached to the surface of each supercritical foamed silicone sheet. The intensity of the characteristic absorption peak of carbon dioxide collected by a near-infrared spectroscopy detection device is converted into the initial residual amount, and the initial residual amount is associated with and stored with the unique identifier. The sheet is placed in a static area with adjustable ambient temperature for a fixed time. During the static period, the ambient temperature value is collected at fixed intervals to form a static ambient temperature sequence, and the sequence is associated with and stored with the unique identifier. After the static period, the sheet thickness is measured and the overall thickness shrinkage rate is calculated. The overall thickness shrinkage rate is associated with and stored with the unique identifier. The output voltage to pressure ratio of the skin product made from the sheet is measured on a test bench as the initial tactile sensitivity, and the initial tactile sensitivity is associated with and stored with the unique identifier. After the finished skin is installed on the robot, the real-time sensitivity value of the finished skin is collected at a fixed sampling period to form a real-time sensitivity sequence. At the same time, the contact torque is collected, converted into contact stress, and integrated to obtain the contact stress integral value. The real-time sensitivity sequence and the contact stress integral value are associated and stored with the unique identifier. The unique identifier and its associated stored initial residual amount, static ambient temperature sequence, overall thickness shrinkage rate, initial tactile sensitivity, real-time sensitivity sequence, and contact stress integral value are summarized into a data entry. The collection of all data entries is the silicone skin full life cycle database.

[0010] As a further aspect of the present invention: the specific construction process of the normalized attenuation coefficient in S2 is as follows: The service sensitivity sequence and contact stress integral value are read from each data entry in the silicone skin life cycle database. The service sensitivity sequence contains multiple sensitivity sampling values ​​collected at fixed time intervals starting from the initial sensitivity value. The contact stress integral value contains the cumulative contact stress value corresponding to each sensitivity sampling time. The initial sensitivity value in the service sensitivity sequence is subtracted from the sensitivity value of the i-th sampling point to obtain the sensitivity decrease at the i-th sampling point. The sensitivity decrease is divided by the contact stress integral value corresponding to the i-th sampling point to obtain the sensitivity attenuation per unit contact stress at the i-th sampling point. The arithmetic mean of the sensitivity attenuation per unit contact stress at all sampling points is calculated, and this arithmetic mean is used as the normalized attenuation coefficient of the sheet.

[0011] As a further aspect of the present invention: the specific process for obtaining the superior shrinkage rate mapping parameter set in step S3 is as follows: The initial residual amount of each historical sheet is obtained and the initial residual amount is divided into multiple continuous intervals according to a preset interval. Within each initial residual amount interval, the normalized attenuation coefficient of the corresponding sheet is extracted, and the sheet with the normalized attenuation coefficient lower than the preset threshold is selected as the superior sheet sample of that interval. Extract the static environmental temperature sequence recorded during the static stage of the premium sheet sample, calculate the arithmetic mean of the static environmental temperature sequence as the static temperature mean, and extract the overall thickness shrinkage rate of the premium sheet sample at the end of static stage as the thickness shrinkage rate; combine and store the static temperature mean and thickness shrinkage rate of all premium sheet samples within the same initial residual amount range to form a premium shrinkage rate mapping parameter set indexed by the initial residual amount range.

[0012] As a further aspect of the present invention: the specific process of establishing the joint parameter set of contraction dynamics in S4 is as follows: Obtain all interval records in the premium shrinkage rate mapping parameter set indexed by the initial residual amount interval identifier; for each initial residual amount interval, extract the average static temperature and thickness shrinkage rate corresponding to all premium sheet samples within that interval, and combine the average static temperature and thickness shrinkage rate of each sample into a temperature shrinkage rate data pair to obtain the temperature shrinkage rate data pair set for that initial residual amount interval; for each initial residual amount interval temperature shrinkage rate data pair set, perform curve fitting operation to solve the fitting parameter set corresponding to that initial residual amount interval; associate and store the interval identifier of each initial residual amount interval with the corresponding fitting parameter set to establish a shrinkage kinetic joint parameter set.

[0013] As a further aspect of the present invention: in step S5, the specific generation process of the feedforward control strategy model is as follows: Access the shrinkage kinetics joint parameter set, extract the initial residual amount interval identifiers and the corresponding fitting parameter sets stored in the parameter set; access the premium shrinkage rate mapping parameter set, extract the static temperature mean, thickness shrinkage rate and the initial residual amount interval identifier of each premium sheet sample stored in the parameter set; access the silicone skin full life cycle database, extract the normalized decay coefficient corresponding to each premium sheet sample. For each high-quality sheet sample, the average static temperature of the sample is used as the first input feature value, the thickness shrinkage rate of the sample is used as the second input feature value, and the normalized attenuation coefficient corresponding to the sample is used as the output label value to construct a training data record; all high-quality sheet samples are traversed, and the training data records are summarized to form a training sample set. The training sample set is input into a preset regression learner, with the first input feature value and the second input feature value as input variables and the output label value as the fitting target. The model training iteration is performed to obtain a feedforward control strategy model with the mean static temperature and thickness shrinkage rate as inputs and the normalized attenuation coefficient as output.

[0014] As a further aspect of the present invention: the specific process for obtaining the target static ambient temperature in step S6 is as follows: Collect the real-time initial residual amount of the new batch of supercritical foamed silicone sheets, compare the real-time initial residual amount with the upper and lower boundaries of each initial residual amount interval, determine the initial residual amount interval to which the real-time initial residual amount belongs, and extract the interval identifier of the initial residual amount interval. Access the joint parameter set of contraction dynamics based on the interval identifier, and extract the fitting parameter set corresponding to the interval identifier; input the fitting parameter set and the preset normalized decay coefficient threshold into the feedforward control strategy model, perform the model inverse solution operation, and obtain the recommended average static temperature that satisfies the preset normalized decay coefficient threshold; calibrate the recommended average static temperature as the target static environment temperature.

[0015] As a further aspect of the present invention: S6 further includes obtaining the actual static environment temperature sequence and actual overall thickness shrinkage rate of the new batch of sheets during the actual static stage, and obtaining the actual initial tactile sensitivity, actual service sensitivity sequence and actual contact stress integral value of the new batch of sheets during the actual service stage, and calculating the actual normalized attenuation coefficient. After associating the actual static ambient temperature sequence, actual overall thickness shrinkage rate, actual initial tactile sensitivity, actual service sensitivity sequence, actual contact stress integral value, and actual normalized attenuation coefficient with the unique identifier of the new batch of sheets, the data is written back to the silicone skin full life cycle database to obtain the updated silicone skin full life cycle database. Based on the updated silicone skin full life cycle database, the premium shrinkage rate mapping parameter set, shrinkage dynamics joint parameter set, and feedforward control strategy model are updated sequentially.

[0016] The beneficial effects of this invention are: 1) This invention establishes a full lifecycle database of silicone skin, including initial residual amount, static temperature sequence, and overall thickness shrinkage rate. Based on the initial residual amount range and the selection of high-quality sheet samples, a joint parameter set for shrinkage kinetics is established and a feedforward control strategy model is trained. This allows the static temperature of new batches of sheets to be differentiated according to their real-time initial residual amount. It is understandable that the amount of carbon dioxide dissolved in the silicone matrix during supercritical foaming and its residual amount after depressurization directly affect the subsequent gas escape rate and the matrix molecular chain rearrangement process. Sheets with higher initial residual amounts have a stronger gas escape driving force and a faster thickness shrinkage rate under the same static conditions. If a uniform fixed temperature is used for static treatment, sheets with high initial residual amounts may shrink excessively, leading to cell collapse, while sheets with low initial residual amounts may shrink insufficiently, resulting in an unstable cell structure. The above scheme establishes a mapping relationship between the initial residual amount and the settling temperature, so that the settling temperature is adjusted in reverse according to the initial residual amount. A high initial residual amount corresponds to a lower settling temperature to suppress the excessively fast shrinkage rate, while a low initial residual amount corresponds to a higher settling temperature to promote the shrinkage process. As a result, sheets with different initial residual amounts all reach a similar shrinkage state with stable cell structure after a fixed settling time, which effectively reduces the dispersion of service sensitivity decay caused by differences in initial state between batches.

[0017] 2) This invention extracts the service sensitivity sequence and contact stress integral value, calculates the sensitivity attenuation under unit contact stress as a normalized attenuation coefficient, and incorporates the cumulative contact stress effect during actual service into the sensitivity attenuation evaluation system. It is understandable that the tactile sensitivity attenuation of silicone skin in actual robot use scenarios is affected by two factors: firstly, the intrinsic pore stability of the material determined by the initial residual amount and the static process; and secondly, the cumulative effect of external stress determined by the robot's operating frequency and contact intensity. If the absolute value of sensitivity attenuation is directly used to judge skin durability, the difference in intensity will mask the difference in the intrinsic quality of the material, leading to a distorted evaluation. It is also understandable that the contact frequency and contact pressure of robot skin vary significantly in different application scenarios. The sensitivity attenuation under high-frequency, high-load conditions is necessarily greater than that under low-frequency, light-load conditions. If the absolute value of attenuation is directly used to judge skin durability, the difference in operating conditions will mask the difference in the intrinsic quality of the material. This invention introduces the integral value of contact stress as a normalized denominator, converting the total attenuation into the attenuation contribution per unit contact stress. This eliminates the interference factor of working condition differences from the evaluation system, making the intrinsic durability of skin products under different working conditions comparable. Based on this, the selected superior sheet samples have lower normalized attenuation coefficients, indicating stronger sensitivity retention under the same contact stress, i.e., better cell structure stability. The model is trained using the mean static temperature and thickness shrinkage rate corresponding to these samples, ensuring that the input-output mapping relationship learned by the model directly points to the optimization direction of the material's intrinsic durability, providing an unbiased data foundation for subsequent feedforward recommendations of static temperature.

[0018] 3) This invention establishes a joint parameter set for shrinkage kinetics by fitting the relationship between the average resting temperature and the thickness shrinkage rate of high-quality sheet samples within each initial residual amount range. It then trains a feedforward control strategy model with the average resting temperature and thickness shrinkage rate as input and the normalized attenuation coefficient as output. When the real-time initial residual amount of a new batch of sheet is measured and its range is determined, the system can automatically reverse-engineer the model and set a recommended average resting temperature that ensures the normalized attenuation coefficient meets a preset threshold. The advantage of this closed-loop mechanism lies in that it no longer relies on a fixed preset resting temperature value, but rather allows the data itself to determine "what resting temperature should be used for the current initial residual amount to ensure the shrinkage rate falls within the high-quality range, thereby obtaining a lower normalized attenuation coefficient." As the actual service data of new batches of sheet is continuously written back to the full lifecycle database, the high-quality shrinkage rate mapping parameter set and the joint parameter set for shrinkage kinetics are updated, and the recommendation accuracy of the feedforward control strategy model continuously improves with the accumulation of sample size. As a result, sheets with different initial residual amounts can be matched to a static temperature that is compatible with their internal gas escape characteristics, significantly reducing the dispersion of service durability among sheets and realizing a leap in quality control paradigm from "post-event record traceability" to "pre-event process control". Attached Figure Description

[0019] The invention will now be further described with reference to the accompanying drawings.

[0020] Figure 1 This is a schematic diagram of a method for tracing and controlling the skin quality of a supercritical foamed silicone robot according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 As shown, this invention is a method for traceability and control of the skin quality of a supercritical foamed silicone robot, comprising the following steps: S1: Based on the static process data and service feedback data of supercritical foamed silicone sheets, historical sheet records are extracted to establish a full life cycle database of silicone skin; the static process data includes the unique identifier of the sheet, the initial residual amount of carbon dioxide, the static environment temperature sequence, and the overall thickness shrinkage rate at the end of static treatment; the service feedback data includes the initial tactile sensitivity, the service sensitivity sequence, and the contact stress integral value. Near-infrared spectroscopy equipment was deployed at the supercritical foamed silicone sheet production line station. A diffuse reflection fiber optic probe was attached to the sheet surface, and the wavenumber range was 4000 cm⁻¹. - ¹ to 7000cm - ¹ Near-infrared diffuse reflectance spectra, extracting carbon dioxide molecules at 4960 cm⁻¹ - ¹Near combination frequency absorption peak and 6960 cm⁻¹ -¹Absorbance value of the nearby overtone absorption peak. Based on the pre-established absorbance-concentration standard curve of carbon dioxide in the silicone matrix, the absorbance value is converted into the mass concentration of dissolved carbon dioxide in a unit volume of silicone sheet, and then multiplied by the sheet volume to obtain the initial residual amount of the sheet. Each sheet is simultaneously assigned a unique identifier, and a multi-point thermocouple array is deployed in the resting area to record the resting environment temperature sequence at a preset sampling period. After resting, a laser displacement sensor is used to measure the thickness of the sheet at multiple points and the average value is taken. This average value is compared with the initial thickness average value at the time of foaming to calculate the overall thickness shrinkage rate. The unique identifier of the sheet, the initial residual amount, the resting environment temperature sequence, and the overall thickness shrinkage rate are used as data for the resting process. The initial tactile sensitivity of the finished skin is measured at the sensor integration station. During the robot's service phase, an embedded tactile sensor array is used to record the service sensitivity sequence at a preset sampling period, and a pressure sensor installed on the surface of the robot joint collects the contact stress signal and integrates it over the service time to obtain the contact stress integral value. Initial tactile sensitivity, service sensitivity sequence, and contact stress integral value are used as service feedback data, which are then linked to the unique sheet identifier and static process data to form a full life cycle database for silicone skin.

[0023] S2: Extract the shrinkage rate evolution characteristics and restore the working condition attenuation coefficient from the full life cycle database of silicone skin, and construct a normalized attenuation coefficient based on the service sensitivity sequence and contact stress integral value; Extract the service sensitivity sequence and contact stress integral value corresponding to the unique identifier of the same sheet from the database. Take the difference between the last sampled value and the initial tactile sensitivity in the service sensitivity sequence as the total sensitivity attenuation. Divide the total sensitivity attenuation by the contact stress integral value to obtain the sensitivity attenuation under unit contact stress, which is marked as the normalized attenuation coefficient. At the same time, extract the arithmetic mean of the static environment temperature sequence of the sheet as the static temperature mean, and extract the overall thickness shrinkage rate as the thickness shrinkage rate, and store them in association with the normalized attenuation coefficient.

[0024] S3: In different initial residual amount ranges, select high-quality sheet samples with normalized decay coefficients lower than the preset threshold, extract the average static temperature and thickness shrinkage rate of the high-quality sheet samples, and integrate them to form a high-quality shrinkage rate mapping parameter set. The initial residual amount is divided into multiple continuous intervals according to preset intervals. Within each initial residual amount interval, the normalized attenuation coefficient of the corresponding sheet is extracted. Sheets with a normalized attenuation coefficient lower than a preset threshold are designated as high-quality sheet samples for that interval. The average resting temperature and thickness shrinkage rate of these samples are extracted. The average resting temperature and thickness shrinkage rate of all high-quality sheet samples within the same initial residual amount interval are combined and stored to form a high-quality shrinkage rate mapping parameter set indexed by the initial residual amount interval identifier.

[0025] S4: Deconstruct the high-quality shrinkage rate mapping parameter set, fit the time-temperature equivalent relationship between the average static temperature and the thickness shrinkage rate in different initial residual amount ranges, and establish a joint parameter set for shrinkage kinetics; For each initial residual amount range, the mean static temperature and thickness shrinkage rate of all high-quality sheet samples within that range are extracted. These are then combined to form a temperature-shrinkage rate data pair. An Arrhenius-type equation describing the gas escape rate in the polymer is used to perform curve fitting on this data pair set, solving for the corresponding fitting parameter set for that range. The fitting parameter set includes the apparent activation energy coefficient and the pre-exponential factor. Each initial residual amount range identifier is associated with and stored with its corresponding fitting parameter set, establishing a joint parameter set for shrinkage kinetics.

[0026] S5: Construct the attribution mapping relationship between the joint parameter set of contraction dynamics and the normalized decay coefficient, and train and output a feedforward control strategy model based on contraction characteristics; The fitting parameter sets corresponding to each initial residual amount interval are extracted, and the mean static temperature, thickness shrinkage rate, and normalized attenuation coefficient of each high-quality sheet sample are extracted. A training sample set is constructed using the mean static temperature and thickness shrinkage rate as input features and the normalized attenuation coefficient as the output label. The training sample set is input into a support vector regression model, and a radial basis function kernel function is used for nonlinear mapping. The penalty coefficient and kernel function parameters are optimized using a grid search method to train a feedforward control strategy model with the mean static temperature and thickness shrinkage rate as input variables and the normalized attenuation coefficient as the output variable.

[0027] S6: Obtain the real-time initial residual amount of the new batch of sheets and determine its range. Input the strategy model according to the joint parameter set corresponding to the range, and set the target static environment temperature of the new sheets according to the model output.

[0028] Near-infrared spectroscopy was performed on the new batch of supercritical foamed silicone sheets. The real-time initial residual amount was calculated according to the method described in S1. The real-time initial residual amount was compared with the upper and lower boundaries of each initial residual amount interval to determine the interval identifier. Based on the interval identifier, the shrinkage kinetic joint parameter set was accessed, and the corresponding fitting parameter set was extracted. The fitting parameter set and the preset normalized attenuation coefficient threshold were input into the feedforward control strategy model, and the model inverse solution operation was performed to obtain the recommended average static temperature that meets the threshold condition. This recommended average static temperature was written as the target static environment temperature into the static area temperature control device, and a fixed static time was maintained.

[0029] It is worth noting that the difference in the initial residual amount of carbon dioxide in supercritical foamed silicone sheets directly affects the gas escape rate and the matrix molecular chain rearrangement process during the resting stage. The higher the initial residual amount, the greater the gas concentration gradient inside the sheet, the stronger the escape driving force, and the faster the thickness shrinkage rate at the same resting temperature; conversely, the lower the initial residual amount, the slower the shrinkage rate. If a fixed temperature is used for resting, sheets with high initial residual amounts are prone to collapse of the cell structure due to excessive shrinkage, while sheets with low initial residual amounts suffer from unstable cells due to insufficient shrinkage. Both will exhibit rapid sensitivity decay during service. This invention establishes a joint parameter set for shrinkage kinetics, solidifies the time-temperature equivalent relationship between resting temperature and thickness shrinkage rate under different initial residual amount ranges in the form of a fitted parameter set, and establishes a mapping between the mean resting temperature, thickness shrinkage rate, and normalized decay coefficient through a feedforward control strategy model. Once the initial residual amount of a new batch of sheets is measured, the model can solve the normalized attenuation coefficient by back-calculating the set of fitting parameters corresponding to that interval, thereby achieving dynamic matching between the settling process parameters and the initial residual amount. This ensures that different batches of sheets reach similar stable cell states after a fixed settling time, reducing the dispersion of service sensitivity attenuation caused by differences in initial state.

[0030] In a preferred embodiment of the present invention, the specific process of establishing a silicone skin lifecycle database in step S1 is as follows: A unique identifier is attached to the surface of each supercritical foamed silicone sheet. The intensity of the characteristic absorption peak of carbon dioxide collected by a near-infrared spectroscopy detection device is converted into the initial residual amount, and the initial residual amount is associated with and stored with the unique identifier. The sheet is placed in a static area with adjustable ambient temperature for a fixed time. During the static period, the ambient temperature value is collected at fixed intervals to form a static ambient temperature sequence, and the sequence is associated with and stored with the unique identifier. After the static period, the sheet thickness is measured and the overall thickness shrinkage rate is calculated. The overall thickness shrinkage rate is associated with and stored with the unique identifier. The output voltage to pressure ratio of the skin product made from the sheet is measured on a test bench as the initial tactile sensitivity, and the initial tactile sensitivity is associated with and stored with the unique identifier. After the finished skin is installed on the robot, the real-time sensitivity value of the finished skin is collected at a fixed sampling period to form a real-time sensitivity sequence. At the same time, the contact torque is collected, converted into contact stress, and integrated to obtain the contact stress integral value. The real-time sensitivity sequence and the contact stress integral value are associated and stored with the unique identifier. The unique identifier and its associated stored initial residual amount, static ambient temperature sequence, overall thickness shrinkage rate, initial tactile sensitivity, real-time sensitivity sequence, and contact stress integral value are summarized into a data entry. The collection of all data entries is the silicone skin full life cycle database.

[0031] In another preferred embodiment of the present invention, the specific construction process of the normalized attenuation coefficient in step S2 is as follows: The system retrieves all records corresponding to a specific sheet from the silicone skin's full lifecycle database. These records contain a service sensitivity sequence collected from the moment the sheet is installed on the robot's surface and begins service until the current moment, along with a time-corresponding sequence of contact stress integral values. The service sensitivity sequence is collected by recording the first sensitivity value at the moment the robot skin is first powered on and completes tactile calibration. This initial sensitivity value is stored in the database. Subsequently, at pre-set fixed time intervals, such as every 24 hours, the robot's built-in tactile signal acquisition system reads the real-time sensitivity value at the current moment, recording it sequentially as the sensitivity values ​​for the first, second, and subsequent sampling points, forming a string of sensitivity values ​​arranged chronologically. The contact stress integral values ​​are obtained by converting contact stress into an electrical signal when the robot skin comes into contact with an external object. This electrical signal is continuously integrated over time from the start of service, and the cumulative integration result at each sampling moment is recorded as the contact stress integral value for that sampling moment. Therefore, each sensitivity sampling point has a contact stress integral value corresponding to its timestamp. For any sampling point in the service sensitivity sequence, the sensitivity value at that sampling point is subtracted from the initial sensitivity value to obtain the total decrease in sensitivity relative to the initial state. This total decrease represents the absolute magnitude of the attenuation of the sheet's tactile response capability since the start of service. The total decrease in sensitivity corresponding to that sampling point is divided by the integral value of the contact stress corresponding to that sampling point. The quotient is the sensitivity attenuation at that sampling point under unit contact stress. This value physically eliminates the influence of the cumulative contact stress on the attenuation magnitude, making the attenuation levels comparable between different sampling times or between different sheets under different external contact stress conditions. This subtraction and division operation is performed on all sampling points in the service sensitivity sequence of the sheet, obtaining the sensitivity attenuation per unit contact stress for each sampling point. The arithmetic mean of the sum of the sensitivity attenuations per unit contact stress for all sampling points is obtained and divided by the total number of sampling points. This arithmetic mean is stored in the database as the normalized attenuation coefficient of the sheet and associated with the sheet's unique identifier.

[0032] It is understandable that the sensitivity decay of silicone skin during actual robot service is influenced by the coupling effect of both the evolution of the material's intrinsic structure and the accumulation of external mechanical stress. If the absolute value of sensitivity decrease is directly used as an indicator of skin durability, skin used under high-frequency, high-load conditions will inevitably exhibit a greater degree of sensitivity decay. A significant portion of this decay is attributed to differences in external usage intensity rather than the quality of the material's own pore structure stability, leading to a serious distortion in the evaluation of skin quality. By dividing the total sensitivity decrease by the integral value of contact stress, the contribution of the interference factor of external stress accumulation is effectively removed from the total decay. This allows the calculated normalized decay coefficient to reflect the sensitivity decay generated by the sheet material under each unit of contact stress, i.e., the material's intrinsic tolerance to external mechanical forces. Furthermore, the arithmetic mean of all sampling points is taken instead of just the value at one endpoint because sensitivity decay may exhibit non-linear, phased characteristics as service time progresses. A single sampling point cannot comprehensively characterize the average decay behavior of the sheet throughout its entire service life. The mean of all sampling points can smooth out the random errors caused by local fluctuations, making the normalized decay coefficient more robustly represent the overall durability level of the sheet throughout the entire observation period. The establishment of this normalized decay coefficient provides a scientific and objective evaluation criterion for selecting superior sheet samples in subsequent steps. This allows sheets with different initial residual amounts, different settling processes, and different service conditions to be compared on a unified benchmark, thereby ensuring that the selected superior sheet samples are indeed individuals exhibiting excellent durability due to the full stabilization of their cell structure, rather than simply showing a low decay rate due to a smaller total amount of external contact during service. Therefore, the shrinkage dynamics joint parameter set and feedforward control strategy model established based on the high-quality samples can truly reflect the causal relationship between the static process parameters and the intrinsic durability of the material, and ultimately achieve precise control of the static environment temperature for the initial residual amount difference of new batches of sheets, so as to reduce the degree of service durability dispersion between batches.

[0033] In another preferred embodiment of the present invention, the specific process for obtaining the high-quality shrinkage rate mapping parameter set in step S3 is as follows: All historical sheet records with completed service data collection were extracted from the silicone skin full life cycle database. Each record includes the initial residual amount, normalized decay coefficient, static environmental temperature sequence, and overall thickness shrinkage rate of the sheet. First, the numerical distribution range of the initial residual amount of all historical sheets was observed. The span between the minimum and maximum values ​​of this distribution range was divided into multiple continuous intervals at pre-set fixed intervals. For example, if the initial residual amount of all historical sheets ranged from a few milligrams per cubic centimeter to several milligrams, then the first interval, the second interval, and finally the last interval covering the maximum value were sequentially divided at fixed intervals, starting from the minimum value. Each interval had a clear lower and upper limit boundary, and the boundaries between adjacent intervals were continuous and non-overlapping. The initial residual amount of each historical sheet was compared with the upper and lower boundaries of each interval. When the initial residual amount of the sheet was greater than or equal to the lower limit boundary of an interval but less than the upper limit boundary, the sheet was assigned to that initial residual amount interval. Thus, all historical sheets were allocated to their respective continuous intervals according to their initial residual amount. After dividing the material into intervals, for each initial residual amount interval, the normalized attenuation coefficients of all historical sheets belonging to that interval are collected. These normalized attenuation coefficients are then compared with a pre-set normalized attenuation coefficient threshold, which is an upper limit of the normalized attenuation coefficient determined based on the service durability requirements of the robotic skin target. When the normalized attenuation coefficient of a sheet is less than the threshold, it means that the sensitivity attenuation of the sheet under unit contact stress is at a low level, that is, the intrinsic tolerance of the sheet to external mechanical action is better than the expected standard. Therefore, the sheet is marked as a superior sheet sample within that initial residual amount interval. Conversely, when the normalized attenuation coefficient is greater than or equal to the threshold, the sheet is not included in the statistical range of superior sheet samples. After screening within each initial residual amount range, the static environmental temperature sequence recorded during the static setting phase for each premium sheet sample within that range is extracted. This sequence consists of multiple temperature samples collected at fixed time intervals during the static setting period. The arithmetic mean of the static environmental temperature sequence is calculated by summing all temperature samples from the start to the end of the static setting and dividing by the total number of samples. This arithmetic mean is used as the average static temperature of the premium sheet sample, reflecting the overall thermal environment level experienced by the sheet during the static setting phase. Simultaneously, the overall thickness shrinkage rate measured at the end of the static setting phase is extracted and used as the thickness shrinkage rate of the sample.The overall thickness shrinkage rate is measured as follows: at the end of foaming, the initial thickness values ​​at multiple preset points on the sheet surface are measured and the average initial thickness is calculated. At the end of the settling period, the thickness values ​​at the same points are measured again and the average thickness after settling is calculated. The difference between the initial average thickness and the average thickness after settling is divided by the initial average thickness; the quotient is the overall thickness shrinkage rate. This value reflects the degree of macroscopic dimensional change of the sheet due to carbon dioxide emission and matrix molecular chain rearrangement during the settling stage. The average settling temperature and thickness shrinkage rate of all high-quality sheet samples within the same initial residual amount range are paired one-to-one. Each sample forms a set of data pairs of average settling temperature and thickness shrinkage rate. All data pairs within the initial residual amount range are aggregated and stored uniformly using the range identifier as the retrieval basis, ultimately forming a high-quality shrinkage rate mapping parameter set indexed by the initial residual amount range.

[0034] It is understandable that the initial residual amount of supercritical foamed silicone sheets determines the initial driving force for carbon dioxide escape. Sheets with higher initial residual amounts have a faster gas escape rate and a more severe thickness shrinkage process at the same resting temperature. Therefore, the initial residual amount is the primary influencing factor determining resting behavior. If sheets with different initial residual amounts are mixed together for analysis, the relationship between resting temperature and thickness shrinkage rate will be obscured by the difference in initial residual amounts, failing to reveal the true relationship between resting temperature and shrinkage rate under a specific initial residual amount condition. By dividing the analysis into intervals, the sheets within each interval are kept at similar levels regarding the key variable of initial residual amount, allowing resting temperature to be used as the main independent variable to examine its relationship with thickness shrinkage rate in subsequent analyses. Secondly, within each interval, high-quality sheet samples are selected based on a normalized attenuation coefficient lower than a preset threshold. It is understandable that the normalized attenuation coefficient is a sensitivity attenuation index after normalization of the contact stress integral value. This index has eliminated the interference caused by differences in contact frequency and contact intensity under service conditions, and can more purely reflect the intrinsic durability level of the material's own cell structure. The lower the normalized attenuation coefficient of the sheet, the stronger its sensitivity retention capability under the same unit contact stress, indicating that its cell structure has reached a relatively ideal stable state during the resting stage. Selecting these sheets means that the subsequent extracted average resting temperature and thickness shrinkage rate data are derived from successful samples that have been proven to have excellent durability, eliminating interference from samples with poor cell structure and ultimately poor service performance due to insufficient or excessive resting. Extracting the average resting temperature instead of directly using the raw sampled values ​​of the resting temperature sequence is because the resting process typically lasts from several hours to tens of hours, during which the ambient temperature may fluctuate slightly. The temperature value of a single sampling point is insufficient to represent the overall thermal environment level of the sheet during the entire resting stage. The arithmetic mean can synthesize all temperature sampling information during the resting stage, obtaining a representative value characterizing the overall thermal intensity of that stage, resulting in a more robust correspondence between the input features and the stage-specific result of thickness shrinkage rate during subsequent modeling. The average resting temperature and thickness shrinkage rate of all high-quality samples within the same initial residual amount range are stored together. The purpose is to provide a data subset stratified by initial residual amount for fitting the time-temperature equivalence relationship in subsequent steps. This allows the fitting operation to be performed separately in a subspace with relatively uniform initial residual amount conditions, thereby obtaining a specific quantitative relationship between resting temperature and thickness shrinkage rate under each initial residual amount range. This lays the data foundation for finally recommending differentiated resting temperatures for new batches of sheets with different initial residual amounts.

[0035] In another preferred embodiment of the present invention, the specific process of establishing the joint parameter set of contraction dynamics in step S4 is as follows: Obtain all interval records in the premium shrinkage rate mapping parameter set indexed by the initial residual amount interval identifier; for each initial residual amount interval, extract the average static temperature and thickness shrinkage rate corresponding to all premium sheet samples within that interval, and combine the average static temperature and thickness shrinkage rate of each sample into a temperature shrinkage rate data pair to obtain the temperature shrinkage rate data pair set for that initial residual amount interval; for each initial residual amount interval temperature shrinkage rate data pair set, perform curve fitting operation to solve the fitting parameter set corresponding to that initial residual amount interval; associate and store the interval identifier of each initial residual amount interval with the corresponding fitting parameter set to establish a shrinkage kinetic joint parameter set.

[0036] It is understandable that the thickness shrinkage of supercritical foamed silicone sheets during the static stage is essentially a volume relaxation process driven by the outward escape of carbon dioxide gas from the inside of the silicone matrix. The rate of gas escape is determined by two core factors: first, the concentration gradient of carbon dioxide inside the silicone matrix; the larger the concentration gradient, the stronger the driving force for the diffusion of gas molecules from the inside to the outside; and second, the thermal activity of the molecular chain segments of the silicone matrix. The higher the temperature, the more intense the activity of the molecular chain segments, and the more and smoother the free volume channels provided for the transition of gas molecules.

[0037] The initial residual amount is the key variable determining the initial magnitude of the concentration gradient. Sheets with higher initial residual amounts have much higher internal carbon dioxide concentrations than the ambient concentration, resulting in a steeper concentration gradient and stronger diffusion driving force. Even at the same static temperature, their gas escape rate and thickness shrinkage process will inevitably be faster than sheets with lower initial residual amounts. Therefore, if data from all initial residual amount ranges are combined for uniform fitting, the resulting set of fitting parameters is actually an averaged compromise expression of the temperature shrinkage relationship under various diffusion driving force levels. This expression cannot accurately characterize the rapid response characteristics of high initial residual amount ranges under high driving force, nor can it accurately describe the slow response characteristics of low initial residual amount ranges under low driving force. This will inevitably lead to deviations between subsequent temperature recommendations based on this expression and actual conditions. The approach of independently fitting each initial residual amount interval is equivalent to acknowledging the primary influence of the initial residual amount on the shrinkage kinetics process. Within each interval, all high-quality sheet samples are similar under the key premise of initial residual amount. Therefore, the difference between the average resting temperature and the thickness shrinkage rate of each sample within the interval mainly comes from the difference in resting temperature. At this time, the fitting parameter set extracted by curve fitting can truly depict the specific quantitative relationship between temperature and shrinkage rate at this initial residual amount level.

[0038] The first parameter in the fitting parameter set characterizes the sensitivity of the thickness shrinkage rate to temperature changes within the specified interval, while the second parameter characterizes the shrinkage baseline level of that interval under baseline conditions. The combination of these two parameters forms a complete temperature-shrinkage rate relationship curve. This curve enables feedforward calculation of process parameters for new batches of sheets in subsequent steps. Specifically, after the real-time initial residual amount of a new batch of silicone sheet is measured, the system can first determine the initial residual amount interval to which the sheet belongs. Then, it retrieves the fitting parameter set corresponding to that interval. Based on the precise mathematical relationship between temperature and shrinkage rate determined by this parameter set, and combined with the mapping relationship between shrinkage rate and normalized attenuation coefficient expressed by the feedforward control strategy model, it reverse-engineers the average static temperature that should be set for the sheet. This process ensures that the setting of the settling temperature is no longer based on experience or fixed specifications, but on quantitative calculations based on the shrinkage kinetics that perfectly match the initial state of the sheet. This ensures that the thermal intensity experienced by each sheet during the settling stage can guide its thickness shrinkage rate to a range that is conducive to the full stability of the cell structure. Ultimately, this results in a lower normalized decay coefficient during the service stage, achieving the goal of reducing batch-to-batch differences in service durability and improving the overall quality consistency of the finished skin products.

[0039] In another preferred embodiment of the present invention, the specific generation process of the feedforward control strategy model in step S5 is as follows: Each record is sequentially read from the storage structure of the shrinkage kinetics joint parameter set. Each record contains an initial residual amount interval identifier and a set of fitting parameters corresponding to that interval identifier. The fitting parameter set includes at least parameters reflecting the temperature sensitivity of the shrinkage process and parameters reflecting the baseline shrinkage rate. Simultaneously, detailed data for all high-quality sheet samples are read from the high-quality shrinkage rate mapping parameter set. Each sample data includes the sample's mean static temperature, thickness shrinkage rate, and the initial residual amount interval identifier to which the sample belongs. Furthermore, the normalized attenuation coefficient corresponding to each high-quality sheet sample is read from the silicone skin full lifecycle database. This normalized attenuation coefficient was calculated in the previous steps by dividing the total sensitivity attenuation by the integral value of the contact stress and taking the average of all sampling points. The data from these three sources are aligned using the initial residual amount interval identifier and the sheet's unique identifier, ensuring that each high-quality sheet sample can be simultaneously associated with its own mean static temperature, thickness shrinkage rate, and normalized attenuation coefficient—three core values. Next, training data records are constructed using a single high-quality sheet material sample as the basic unit. The construction method involves using the average static temperature of the sample as the first input feature, the thickness shrinkage rate as the second input feature, and the normalized decay coefficient as the output label. This forms a complete training data record consisting of two input features and one output label. This process is repeated for all high-quality sheet material samples included in the high-quality shrinkage rate mapping parameter set. Finally, the training data records for all samples are aggregated to form a complete training sample set. Each record in this training sample set represents the correspondence between the thermodynamic conditions experienced by a sheet material with proven excellent service durability during the static stage and its final normalized decay coefficient. The entire training sample set is then fed into a pre-defined regression learner for model training. A regression learner is a mathematical framework that automatically learns the mapping rules from the corresponding data of input features and output labels. During training, the regression learner reads each record in the training sample set, using the first and second input feature values ​​of each record as independent variables and the output label value as the observed target of the dependent variable. Through an internal iterative algorithm, it continuously adjusts its internal parameters, gradually narrowing the gap between the predicted output value calculated based on the input features and the actual output label value of the record. The training process ends when the gap narrows to a preset convergence criterion or the number of iterations reaches a preset upper limit. At this point, the mapping relationship solidified by the regression learner's internal parameters constitutes the feedforward control strategy model.The model uses the mean static temperature and thickness shrinkage rate as input variables and the normalized attenuation coefficient as output variable. When any combination of the mean static temperature and thickness shrinkage rate is provided to the model, the model can output the predicted value of the normalized attenuation coefficient that can be achieved under the static conditions.

[0040] It is understandable that in the entire process of manufacturing and serving supercritical foamed silicone robotic skin, the temperature setting during the resting stage and the resulting thickness shrinkage rate are key process nodes that determine whether the internal cell structure of the silicone matrix can be sufficiently stabilized. The stability of the cell structure directly determines the ability of the finished skin to maintain stable tactile sensitivity when subjected to repeated contact stress during subsequent service. This ability is quantitatively characterized by the normalized attenuation coefficient. However, the relationship between the average resting temperature and the thickness shrinkage rate and the final normalized attenuation coefficient is not a simple linear causal relationship, but a complex mapping involving the coupling of multiple physicochemical mechanisms such as carbon dioxide diffusion kinetics, viscoelastic relaxation of the silicone matrix, and mechanical strengthening of the cell walls. Simply relying on theoretical derivation is insufficient to establish an analytical expression that accurately covers the mapping relationship under different initial residual conditions. However, a data-driven approach, using high-quality sheet samples with proven low normalized attenuation coefficients as training objects, and employing the actual correspondence between their mean resting temperature, thickness shrinkage rate, and normalized attenuation coefficient as training samples, allows the model to directly learn the statistical regularities between input and output from empirical data without explicitly solving all intermediate physical processes. Each sample in the training set originates from high-quality sheet material, meaning that all the mapping relationships learned by the model correspond to combinations of process parameters that produce excellent service performance. During training, the model automatically eliminates the influence of failure cases where improper resting temperature settings lead to shrinkage rates deviating from the reasonable range, resulting in excessively high normalized attenuation coefficients. This ensures that the trained feedforward control strategy model itself inherently possesses a convergence orientation towards the high-quality range. By combining the trained feedforward control strategy model with the joint parameter set of shrinkage kinetics, a complete feedforward calculation chain for process parameters can be formed. Once the real-time initial residual amount of a new batch of sheets is measured and its corresponding initial residual amount range is determined, the fitting parameter set for that range first describes the definite relationship curve between the average resting temperature and the thickness shrinkage rate under that initial residual amount condition. Subsequently, under the constraint of this relationship curve, the feedforward control strategy model uses a preset target normalized attenuation coefficient threshold as the expected output to solve in reverse for the recommended average resting temperature value that simultaneously satisfies the shrinkage kinetics and service durability objectives. This mechanism ensures that the setting of the resting temperature no longer depends on empirical parameters or fixed specifications, but is based on systematic learning and quantitative modeling of historical high-quality data. Each new sheet can obtain a personalized resting temperature condition that precisely matches its initial residual amount, ultimately achieving the overall invention objective of reducing the dispersion of service durability between batches.

[0041] In another preferred embodiment of the present invention, the specific process of obtaining the target static ambient temperature in step S6 is as follows: Collect the real-time initial residual amount of the new batch of supercritical foamed silicone sheets, compare the real-time initial residual amount with the upper and lower boundaries of each initial residual amount interval, determine the initial residual amount interval to which the real-time initial residual amount belongs, and extract the interval identifier of the initial residual amount interval. Access the joint parameter set of contraction dynamics based on the interval identifier, and extract the fitting parameter set corresponding to the interval identifier; input the fitting parameter set and the preset normalized decay coefficient threshold into the feedforward control strategy model, perform the model inverse solution operation, and obtain the recommended average static temperature that satisfies the preset normalized decay coefficient threshold; calibrate the recommended average static temperature as the target static environment temperature.

[0042] In another preferred embodiment of the present invention, S6 further includes obtaining the actual static environment temperature sequence and actual overall thickness shrinkage rate of the new batch of sheets during the actual static stage, and obtaining the actual initial tactile sensitivity, actual service sensitivity sequence and actual contact stress integral value of the new batch of sheets during the actual service stage, and calculating the actual normalized attenuation coefficient. After associating the actual static ambient temperature sequence, actual overall thickness shrinkage rate, actual initial tactile sensitivity, actual service sensitivity sequence, actual contact stress integral value, and actual normalized attenuation coefficient with the unique identifier of the new batch of sheets, the data is written back to the silicone skin full life cycle database to obtain the updated silicone skin full life cycle database. Based on the updated silicone skin full life cycle database, the premium shrinkage rate mapping parameter set, shrinkage dynamics joint parameter set, and feedforward control strategy model are updated sequentially.

[0043] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for traceability and control of the skin quality of a supercritical foamed silicone robot, characterized in that, Includes the following steps: S1: Based on the static process data and service feedback data of supercritical foamed silicone sheets, historical sheet records are extracted to establish a full life cycle database of silicone skin; the static process data includes the unique identifier of the sheet, the initial residual amount of carbon dioxide, the static environment temperature sequence, and the overall thickness shrinkage rate at the end of static treatment; the service feedback data includes the initial tactile sensitivity, the service sensitivity sequence, and the contact stress integral value. S2: Extract the shrinkage rate evolution characteristics and restore the working condition attenuation coefficient from the full life cycle database of silicone skin, and construct a normalized attenuation coefficient based on the service sensitivity sequence and contact stress integral value; S3: In different initial residual amount ranges, select high-quality sheet samples with normalized decay coefficients lower than the preset threshold, extract the average static temperature and thickness shrinkage rate of the high-quality sheet samples, and integrate them to form a high-quality shrinkage rate mapping parameter set. S4: Deconstruct the high-quality shrinkage rate mapping parameter set, fit the time-temperature equivalent relationship between the average static temperature and the thickness shrinkage rate in different initial residual amount ranges, and establish a joint parameter set for shrinkage kinetics; The specific process for establishing the joint parameter set of contraction dynamics is as follows: Obtain all interval records in the premium shrinkage rate mapping parameter set indexed by the initial residual amount interval identifier; for each initial residual amount interval, extract the average static temperature and thickness shrinkage rate corresponding to all premium sheet samples within that interval, and combine the average static temperature and thickness shrinkage rate of each sample into a temperature shrinkage rate data pair to obtain the temperature shrinkage rate data pair set for that initial residual amount interval; for each initial residual amount interval temperature shrinkage rate data pair set, perform curve fitting operation to solve the fitting parameter set corresponding to that initial residual amount interval; associate and store the interval identifier of each initial residual amount interval with the corresponding fitting parameter set to establish a shrinkage kinetic joint parameter set; S5: Construct the attribution mapping relationship between the joint parameter set of contraction dynamics and the normalized decay coefficient, and train and output a feedforward control strategy model based on contraction characteristics; The specific generation process of the feedforward control strategy model is as follows: Access the shrinkage kinetics joint parameter set, extract the initial residual amount interval identifiers and the corresponding fitting parameter sets stored in the parameter set; access the premium shrinkage rate mapping parameter set, extract the average static temperature, thickness shrinkage rate and the initial residual amount interval identifier of each premium sheet sample stored in the parameter set. Access the silicone skin full life cycle database and extract the normalized decay coefficients corresponding to each premium sheet sample; For each sample of premium sheet material, the average static temperature of the sample is used as the first input feature value, the thickness shrinkage rate of the sample is used as the second input feature value, and the normalized attenuation coefficient of the sample is used as the output label value to construct a training data record. Traverse all high-quality sheet material samples and summarize each training data record to form a training sample set; The training sample set is input into a preset regression learner, with the first input feature value and the second input feature value as input variables and the output label value as the fitting target. The model training iteration is performed to obtain a feedforward control strategy model with the mean static temperature and thickness shrinkage rate as inputs and the normalized attenuation coefficient as output. S6: Obtain the real-time initial residual amount of the new batch of sheets and determine its range. Input the strategy model according to the joint parameter set corresponding to the range, and set the target static environment temperature of the new sheets according to the model output.

2. The method for traceability and control of the skin quality of a supercritical foamed silicone robot according to claim 1, characterized in that, In step S1, the specific process of establishing a full life-cycle database for silicone skin is as follows: A unique identifier is attached to the surface of each supercritical foamed silicone sheet. The intensity of the characteristic absorption peak of carbon dioxide collected by a near-infrared spectroscopy detection device is converted into the initial residual amount, and the initial residual amount is associated with and stored with the unique identifier. The sheet is placed in a static area with adjustable ambient temperature for a fixed time. During the static period, the ambient temperature value is collected at fixed intervals to form a static ambient temperature sequence, and the sequence is associated with and stored with the unique identifier. After the static period, the sheet thickness is measured and the overall thickness shrinkage rate is calculated. The overall thickness shrinkage rate is associated with and stored with the unique identifier. The output voltage to pressure ratio of the skin product made from the sheet is measured on a test bench as the initial tactile sensitivity, and the initial tactile sensitivity is associated with and stored with the unique identifier. After the finished skin is installed on the robot, the real-time sensitivity value of the finished skin is collected at a fixed sampling period to form a real-time sensitivity sequence. At the same time, the contact torque is collected, converted into contact stress, and integrated to obtain the contact stress integral value. The real-time sensitivity sequence and the contact stress integral value are associated and stored with the unique identifier. The unique identifier and its associated stored initial residual amount, static ambient temperature sequence, overall thickness shrinkage rate, initial tactile sensitivity, real-time sensitivity sequence, and contact stress integral value are summarized into a data entry. The collection of all data entries is the silicone skin full life cycle database.

3. The method for traceability and control of the skin quality of a supercritical foamed silicone robot according to claim 1, characterized in that, In S2, the specific construction process of the normalized attenuation coefficient is as follows: The service sensitivity sequence and contact stress integral value are read from each data entry in the silicone skin life cycle database. The service sensitivity sequence contains multiple sensitivity sampling values ​​collected at fixed time intervals starting from the initial sensitivity value. The contact stress integral value contains the cumulative contact stress value corresponding to each sensitivity sampling time. Subtract the initial sensitivity value in the service sensitivity sequence from the sensitivity value at the i-th sampling point to obtain the sensitivity decrease at the i-th sampling point; Divide the decrease in sensitivity by the integral value of the contact stress corresponding to the i-th sampling point to obtain the sensitivity attenuation per unit contact stress at the i-th sampling point; Calculate the arithmetic mean of the unit contact stress sensitivity attenuation at all sampling points, and use this arithmetic mean as the normalized attenuation coefficient of the sheet.

4. The method for traceability and control of the skin quality of a supercritical foamed silicone robot according to claim 1, characterized in that, In step S3, the specific process for obtaining the high-quality shrinkage rate mapping parameter set is as follows: The initial residual amount of each historical sheet is obtained and the initial residual amount is divided into multiple continuous intervals according to a preset interval. Within each initial residual amount interval, the normalized attenuation coefficient of the corresponding sheet is extracted, and the sheet with the normalized attenuation coefficient lower than the preset threshold is selected as the superior sheet sample of that interval. Extract the static environmental temperature sequence recorded during the static stage of the premium sheet sample, calculate the arithmetic mean of the static environmental temperature sequence as the static temperature mean, and extract the overall thickness shrinkage rate of the premium sheet sample at the end of static stage as the thickness shrinkage rate; combine and store the static temperature mean and thickness shrinkage rate of all premium sheet samples within the same initial residual amount range to form a premium shrinkage rate mapping parameter set indexed by the initial residual amount range.

5. The method for traceability and control of the skin quality of a supercritical foamed silicone robot according to claim 1, characterized in that, In step S6, the specific process for obtaining the target static ambient temperature is as follows: Collect the real-time initial residual amount of the new batch of supercritical foamed silicone sheets, compare the real-time initial residual amount with the upper and lower boundaries of each initial residual amount interval, determine the initial residual amount interval to which the real-time initial residual amount belongs, and extract the interval identifier of the initial residual amount interval. Access the joint parameter set of contraction dynamics based on the interval identifier, and extract the fitting parameter set corresponding to the interval identifier; input the fitting parameter set and the preset normalized decay coefficient threshold into the feedforward control strategy model, perform the model inverse solution operation, and obtain the recommended average static temperature that satisfies the preset normalized decay coefficient threshold; calibrate the recommended average static temperature as the target static environment temperature.

6. The method for traceability and control of supercritical foamed silicone robotic skin quality according to claim 1, characterized in that, S6 further includes obtaining the actual static environment temperature sequence and actual overall thickness shrinkage rate of the new batch of sheets during the actual static stage, and obtaining the actual initial tactile sensitivity, actual service sensitivity sequence and actual contact stress integral value of the new batch of sheets during the actual service stage, and calculating the actual normalized attenuation coefficient. After associating the actual static ambient temperature sequence, actual overall thickness shrinkage rate, actual initial tactile sensitivity, actual service sensitivity sequence, actual contact stress integral value, and actual normalized attenuation coefficient with the unique identifier of the new batch of sheets, the data is written back to the silicone skin full life cycle database to obtain the updated silicone skin full life cycle database. Based on the updated silicone skin full life cycle database, the premium shrinkage rate mapping parameter set, shrinkage dynamics joint parameter set, and feedforward control strategy model are updated sequentially.

Images