Sacrococcygeal shear force decoupling and microenvironment evolution prediction and control method and system

Abstract

This invention discloses a method and system for decoupling shear force and predicting and controlling the evolution of the microenvironment in the sacrococcygeal region, belonging to the field of medical diagnostic monitoring technology. It includes collecting multidimensional raw signals of pressure, shear force, and microenvironment in the sacrococcygeal region; obtaining real-time decoupling sensor data through cross-decoupling calculations; performing bidirectional driven inference analysis by combining data feature mapping and prior rule matching to obtain tissue damage risk prediction instructions and extract the intervention range and intensity, generating dynamic intervention instructions; controlling an active offsetting dynamic intervention actuator to perform physical intervention actions, collecting the reflected sensor change signals after execution, and generating tissue response data; optimizing the bidirectional driven inference analysis and updating the inference parameters. This invention employs a combination of multidimensional sensing decoupling, bidirectional driven prediction, and closed-loop active offsetting intervention, enabling accurate prediction and targeted active control of sacrococcygeal tissue damage risk, achieving personalized and adaptive risk management effects.

Description

Technical Field

[0001] This invention relates to the field of medical diagnostic monitoring technology, and in particular to a method and system for decoupling shear force in the sacrococcygeal region and predicting and controlling the evolution of the microenvironment. Background Technology

[0002] Sacrococcygeal tissue injury, especially pressure sores, is a common and serious complication in people who are bedridden or sedentary for extended periods. Its development is the result of multiple factors, among which vertical pressure, horizontal shear force, and the local skin microenvironment—namely, temperature and humidity—are recognized as the three core pathogenic factors. Therefore, real-time and accurate monitoring of these key physiological and environmental parameters is a prerequisite and crucial for the effective prevention and management of sacrococcygeal tissue injury risks.

[0003] In related technologies, Chinese invention patent application CN120611568A discloses a finite element modeling and analysis method for the stress on the sacrococcygeal soft tissue. The method includes acquiring three-dimensional image data of normal sacrococcygeal soft tissue, accurately extracting the soft tissue contour using image segmentation technology, constructing a three-dimensional surface dynamic model based on the extracted soft tissue contour, simulating the deformation properties of the three-dimensional surface dynamic model under stress, measuring the deformation properties of the soft tissue under test under stress, and comparing the deformation properties of the soft tissue under test under stress with the deformation properties of the three-dimensional surface dynamic model under stress to evaluate the support force of the soft tissue under test.

[0004] However, while the above-mentioned methods can assess tissue deformation and support force under specific boundary conditions through simulation calculations, they still have limitations in actual clinical long-term monitoring and pressure ulcer prevention scenarios: First, these methods heavily rely on large-scale static offline medical imaging data, making it impossible to achieve continuous, real-time, in-situ monitoring of patients in their daily bedridden or sedentary states; second, existing mechanical deduction models often treat external loads as overall macroscopic stress, making it difficult to cross-decouple and extract the "horizontal shear force" and "vertical pressure" that are highly likely to cause subcutaneous microcirculation ischemia and necrosis; in addition, the assessment dimensions are mostly limited to pure mechanical physical fields, ignoring the key "microenvironment" factors that are highly likely to catalyze tissue damage and the coupling effects of the patient's individual dynamic physiological rhythms; finally, existing technical systems are all limited to passive, stage-based risk diagnosis, lacking a closed-loop active management mechanism based on real-time situational awareness, and unable to translate predicted risks into quantitative physical intervention actions to block the evolution of the pathological microenvironment, making it difficult to substantially prevent tissue damage in the early stages. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method and system for sacrococcygeal shear force decoupling and microenvironment evolution prediction and control. By employing a combination of multidimensional perception decoupling, bidirectional driving prediction, and closed-loop active offsetting intervention, it can achieve accurate prediction and targeted active control of sacrococcygeal tissue damage risk, resulting in personalized and adaptive risk management.

[0006] The above objectives can be achieved through the following approach: A method for decoupling shear force and predicting and controlling the evolution of the microenvironment in the sacrococcygeal region includes: collecting multidimensional raw signals of pressure, shear force, and microenvironment from a biomimetic multilayer flexible sensor array deployed in the sacrococcygeal region; performing cross-decoupling calculations to obtain real-time decoupled sensor data; constructing target prediction data based on the real-time decoupled sensor data; performing bidirectional driven deductive analysis of data feature mapping and prior rule matching to obtain tissue damage risk prediction instructions including risk level and main cause classification; extracting the intervention range and intensity based on the tissue damage risk prediction instructions to generate dynamic intervention instructions; sending the dynamic intervention instructions to an active offset dynamic intervention actuator for instruction parsing and control conversion to generate physical intervention actions matching the main cause classification; collecting the sensor change signals reflected by the biomimetic multilayer flexible sensor array after the active offset dynamic intervention actuator performs the physical intervention actions to generate tissue response data; and optimizing the bidirectional driven deductive analysis based on the tissue response data to update the deductive parameters of the output tissue damage risk prediction instructions.

[0007] Optionally, obtaining real-time decoupled sensing data includes: simultaneously acquiring a first raw signal from the sacrococcygeal monitoring point through the pressure sensing layer of the biomimetic multilayer flexible sensing array; simultaneously acquiring a second raw signal from the sacrococcygeal monitoring point through the shear force sensing layer of the biomimetic multilayer flexible sensing array; simultaneously acquiring a third raw signal from the sacrococcygeal monitoring point through the microenvironment sensing layer of the biomimetic multilayer flexible sensing array; performing signal decoupling and feature extraction on the first, second, and third raw signals respectively to obtain mutually independent pressure feature vectors, shear force feature vectors, and microenvironment feature vectors; and fusing the pressure feature vectors, shear force feature vectors, and microenvironment feature vectors to generate real-time decoupled sensing data.

[0008] Optionally, constructing target prediction data based on real-time decoupled sensing data includes: obtaining individual physiological rhythm parameters corresponding to the sacrococcygeal region, fusing them with real-time decoupled sensing data to generate enhanced real-time sensing data, which is then used as target prediction data.

[0009] Optionally, obtaining individual physiological rhythm parameters corresponding to the sacrococcygeal region includes: acquiring historical time-series sensing data through long-term monitoring using a biomimetic multilayer flexible sensing array; performing periodic analysis on the historical time-series sensing data to extract periodic fluctuation patterns in pressure distribution and microenvironment parameters of the sacrococcygeal region; constructing a physiological rhythm model characterizing the user's individual habits based on the periodic fluctuation patterns; and calculating individual physiological rhythm parameters based on the current time.

[0010] Optionally, obtaining the tissue damage risk prediction instruction containing risk level and main cause classification includes: performing feature mapping operation on the target prediction data in a data-driven dimension to obtain a first risk assessment result; performing rule matching operation on the target prediction data in a knowledge-driven dimension to obtain a second risk assessment result; weighting and fusing the first risk assessment result and the second risk assessment result to generate a comprehensive risk score; mapping the comprehensive risk score to a risk threshold range for numerical comparison to determine the risk level; analyzing the contribution of each causative factor in the first risk assessment result and the second risk assessment result, determining the main cause classification based on the contribution, and concatenating the risk level to obtain the tissue damage risk prediction instruction.

[0011] Optionally, generating dynamic intervention instructions includes: performing numerical mapping extraction on the risk level to obtain basic intervention intensity parameters; performing control logic lookup table matching based on the main cause classification to obtain target intervention mode parameters; combining the basic intervention intensity parameters and the target intervention mode parameters in a time sequence to generate an initial intervention instruction sequence; and performing spatial registration on the initial intervention instruction sequence based on the spatial distribution information in the real-time decoupled sensor data to generate dynamic intervention instructions that include execution location, magnitude and direction of force, and timing of action.

[0012] Optionally, generating physical intervention actions that match the primary cause classification includes: reducing the order and format of the dynamic intervention command to obtain the execution unit control command; parsing the execution unit control command to extract the execution position parameters, action timing parameters, and force parameters; when the primary cause classification is shear force dominant, configuring the force parameter as a counteracting force vector opposite to the shear force direction; and using the execution position parameters, action timing parameters, and counteracting force vector to drive the physical execution unit in the active counteracting dynamic intervention execution mechanism to perform the intervention operation and generate the physical intervention action.

[0013] Optionally, generating tissue response data includes: acquiring sensing sequences continuously collected by a biomimetic multilayer flexible sensing array within a feedback time window to obtain post-intervention sensing data; performing mean filtering and variance extraction on the post-intervention sensing data to calculate steady-state feature values ​​after the intervention takes effect; extracting the difference vector between the steady-state feature values ​​and the corresponding feature values ​​before the intervention and performing standardized mapping to generate quantitative indicators of intervention effect; and concatenating and associating the quantitative indicators of intervention effect with dynamic intervention command execution data segments to generate tissue response data.

[0014] Optionally, the optimization of the bidirectional driven inference analysis includes: extracting the actual risk change trend from the organizational response data, comparing it with the organizational damage risk prediction instruction, and generating a prediction error signal; using the prediction error signal, adjusting the connection weights of the neural network used for feature mapping operations in the data-driven dimension using a backpropagation algorithm; and correcting the confidence parameters of the prior rules used for rule matching operations in the knowledge-driven dimension based on the correlation between the prediction error signal and the main cause classification.

[0015] Based on the same inventive concept, this invention also provides a sacrococcygeal shear force decoupling and microenvironment evolution prediction and control system, comprising: a multidimensional signal acquisition and decoupling module, used to acquire multidimensional raw signals of pressure, shear force, and microenvironment output by a biomimetic multilayer flexible sensor array deployed in the sacrococcygeal region, perform cross-decoupling calculations, and obtain real-time decoupling sensor data; a risk extrapolation and prediction module, used to construct target prediction data based on real-time decoupling sensor data, perform bidirectional driven extrapolation analysis of data feature mapping and prior rule matching, and obtain tissue damage risk prediction instructions including risk level and main cause classification; and a dynamic intervention instruction generation module, used to generate instructions based on tissue... The damage risk prediction instruction extracts the intervention range and intensity to generate dynamic intervention instructions; the intervention execution and control module sends the dynamic intervention instructions to the active offset dynamic intervention execution mechanism for instruction parsing and control conversion, generating physical intervention actions that match the main cause classification; the tissue response perception feedback module collects the sensing change signals reflected by the biomimetic multilayer flexible sensor array after the active offset dynamic intervention execution mechanism performs physical intervention actions, generating tissue response data; the inference parameter closed-loop optimization module optimizes the bidirectional drive inference analysis based on the tissue response data, updating the inference parameters of the output tissue damage risk prediction instruction.

[0016] Compared with the prior art, the present invention has the following advantages: This invention achieves comprehensive perception of the sacrococcygeal tissue state by decoupling and comprehensively analyzing multidimensional raw signals of pressure, shear force, and microenvironment. Compared with traditional techniques that only monitor a single pressure, this invention can simultaneously acquire and distinguish between two major categories of key injury-causing factors: mechanics and the microenvironment, thus establishing a data foundation.

[0017] This invention constructs a bidirectional driven inductive analysis model that combines data feature mapping and prior rule matching, and integrates individual physiological rhythm parameters, thereby achieving forward-looking and causal prediction of tissue damage risk. This method not only provides early warning of risks but also clearly identifies the dominant factors of those risks, making risk assessment no longer a simple threshold alarm but an instruction with in-depth diagnostic significance, thus improving the intelligence level of risk management and the reliability of decision-making basis.

[0018] This invention proposes an active, offsetting dynamic intervention mechanism that can generate and execute precisely matched physical intervention actions based on the main causative factors clearly defined in the risk prediction instructions. Particularly effective against highly destructive shear forces, the system can generate opposing offsetting forces for active neutralization, resolving the key mechanical conditions for tissue damage at their source. Its targeted nature and effectiveness far surpass traditional passive decompression strategies.

[0019] This invention designs a complete adaptive learning loop from organizational response perception feedback to closed-loop optimization of inference parameters. The system can evaluate and quantify the accuracy of the prediction model by monitoring the actual effects after intervention, and use the prediction error to adjust the model parameters in reverse, enabling the entire control system to have the ability to continuously improve itself and adapt to individual needs. It can continuously optimize its risk management strategies for specific users, and the effectiveness and accuracy of long-term application are continuously improved.

[0020] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic flowchart of a method for decoupling sacrococcygeal shear force and predicting and controlling microenvironment evolution according to an embodiment of the present invention.

[0023] Figure 2 This is a real-time decoupled multimodal feature space grid mapping diagram according to an embodiment of the present invention.

[0024] Figure 3 This is an embodiment of the present invention showing the individual physiological rhythm cycle fluctuation pattern and model calculation diagram.

[0025] Figure 4 This is a schematic diagram of the structure of a sacrococcygeal shear force decoupling and microenvironment evolution prediction and control system according to an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0027] Reference Figure 1 One embodiment of the present invention proposes a method and system for decoupling shear force and predicting and controlling the evolution of the microenvironment in the sacrococcygeal region. By combining multidimensional perception decoupling, bidirectional driving prediction and closed-loop active cancellation intervention, it is possible to accurately predict and actively control the risk of tissue damage in the sacrococcygeal region, and achieve personalized and adaptive risk management.

[0028] The method in this embodiment specifically includes: S1. Collect multidimensional raw signals of pressure, shear force and microenvironment output by the biomimetic multilayer flexible sensor array deployed in the sacral and coccygeal region, perform cross-decoupling calculations, and obtain real-time decoupled sensor data. Optionally, obtaining real-time decoupled sensing data includes: The first raw signal of the sacral and coccygeal monitoring points is collected synchronously through the pressure sensing layer of the biomimetic multilayer flexible sensor array. The second raw signal of the sacral and coccygeal monitoring point is collected simultaneously through the shear force sensing layer of the biomimetic multilayer flexible sensor array. The third original signal of the sacral and coccygeal monitoring points is collected simultaneously through the microenvironment sensing layer of the biomimetic multilayer flexible sensing array. Signal decoupling and feature extraction were performed on the first, second, and third original signals respectively to obtain independent pressure feature vectors, shear force feature vectors, and microenvironment feature vectors. By fusing pressure feature vectors, shear force feature vectors, and microenvironment feature vectors, real-time decoupled sensing data is generated.

[0029] Specifically, a biomimetic multilayer flexible sensor array deployed on the surface of a mattress or cushion integrates a pressure-sensing layer, a shear force-sensing layer, and a microenvironment-sensing layer, which operate in a synchronous triggering mode. At a set sampling frequency, typically 50-100Hz, a multi-channel synchronous analog-to-digital converter (ADC) is used to acquire in parallel the first raw signal reflecting pressure, the second raw signal reflecting shear force, and the third raw signal reflecting temperature and humidity at various monitoring points in the sacral and coccygeal region.

[0030] Due to the coupling effect in the physical structure of the sensors, there is cross-sensitivity between the original signals, meaning that pressure changes can interfere with shear force readings. To achieve signal decoupling, a pre-calibrated cross-sensitivity correction matrix is ​​used to linearly transform the first and second original signals. This correction matrix is ​​obtained by applying pure pressure or pure shear force loads under controlled experimental conditions, measuring and modeling the response relationships of each sensing channel, thereby eliminating or reducing mutual interference between signals. The decoupled signals undergo digital filtering, for example, using a Kalman filter to filter out noise introduced by minor body vibrations or electromagnetic interference. Subsequently, feature extraction is performed on the three purified signals. For example, pressure peak value, average pressure, and pressure gradient are calculated from the pressure signal to form a pressure feature vector; shear force amplitude and shear force direction are calculated from the shear force signal to form a shear force feature vector; and current temperature, humidity values, and their rate of change per unit time are calculated from the microenvironment signal to form a microenvironment feature vector.

[0031] The independent pressure feature vectors, shear force feature vectors, and microenvironment feature vectors extracted from all monitoring points at the same sampling time are structurally fused. This fusion operation involves concatenating vectors in a predetermined order and adding timestamps and spatial coordinate indices of each monitoring point to form a high-dimensional, real-time, decoupled sensor data frame. The fusion function introduces a flattened cascading mechanism with spatial mapping. The mathematical expression of this data frame can be defined as: , in, represent Real-time decoupled sensor data frames at any given moment; The total number of effective monitoring points in the biomimetic multilayer flexible sensor array; Representing the Fixed two-dimensional spatial coordinate index of each monitoring point; , and Representing respectively in the The sublayer pressure feature vector, shear force feature vector, and microenvironment feature vector were independently extracted from each monitoring point; Represents feature cascading operations based on spatial grid order; The representative will use the global timestamp The merge character placed at the beginning of the input for concatenation; superscript. This represents the matrix transpose, used to generate a standardized one-dimensional column vector suitable for subsequent data processing. If the dimensions of single-point pressure, shear force, and microenvironment features are respectively... , and Then the generated For one dimension fixed as A high-dimensional feature array. This data frame comprehensively describes the overall mechanical and microenvironmental state of the sacrococcygeal tissues at a specific moment, forming the basis for subsequent predictive analysis. For example... Figure 2 As shown in the figure, this figure intuitively reflects the physical state of the clean data after eliminating cross-sensitivity correction in a two-dimensional spatial coordinate index.

[0032] For example, taking the real-time monitoring process of a long-term bedridden disabled patient in the intensive care unit as an example, the system is set to a sampling frequency of 100Hz, and the timestamp of the current Unix system is... At any given moment, the biomimetic multilayer flexible sensor array synchronously collects data from the sacral and coccygeal regions. The signals from each valid monitoring point are analyzed. Taking the first monitoring point with coordinate indices (10, 20) as an example, after filtering out high-frequency and environmental electromagnetic noise using a Kalman filter, the dimension is extracted based on a peak-to-average algorithm. Pressure feature vector Dimensions are extracted based on amplitude and angle algorithms. shear force eigenvector Dimensions are extracted based on instantaneous values ​​and rates of change. microenvironment vector According to the cascade formula First, the local column vector is obtained by sequentially stitching together the grid of the first point. , dimension Dimension. After performing the same extraction and concatenation on the remaining 3 points, the total length is obtained as follows: A 3D spatial feature array is used, with the global timestamp placed at the top, to calculate the... For one dimension, strictly fixed as A high-dimensional, real-time decoupled sensor data frame column vector. Through hardware synchronization and software-level cross-decoupling correction, coupling interference caused by physical structure is eliminated. Furthermore, by employing a matrix concatenation method with structured spatial mapping, the spatiotemporal topological relationships of multimodal data are fully preserved, providing a reliable source for representing physical reality from the underlying data.

[0033] S2. Based on real-time decoupled sensor data, construct target prediction data, perform bidirectional driven deduction analysis of data feature mapping and prior rule matching, and obtain tissue damage risk prediction instructions that include risk level and main cause classification. Optionally, constructing target prediction data based on real-time decoupled sensing data includes: Individual physiological rhythm parameters corresponding to the sacrococcygeal region are obtained and fused with real-time decoupled sensor data to generate enhanced real-time sensor data, which is used as target prediction data.

[0034] Specifically, individual physiological rhythm parameters corresponding to the sacrococcygeal region are obtained. These parameters are not measured in real time, but rather extracted from a pre-established and continuously updated individual physiological rhythm model based on the current system timestamp. These individual physiological rhythm parameters are structured numerical vectors containing forward-looking or periodic indicators derived from analysis of historical user data, such as the expected probability of sacrococcygeal pressure distribution center shift at the current time point, the typical stable duration in lying or sitting positions, and the normal fluctuation range of microenvironmental temperature and humidity. To accommodate computer processing and model calculation requirements, this individual physiological rhythm model is updated at the current time... Truncate the output parameter vector It is formatted as a one-dimensional column vector as follows: ; in, for The expected probability of pressure center shift at any given time. The expected remaining settling time for the current posture. and These are the normal steady-state baseline values ​​of temperature and humidity in the microenvironment calculated based on the rhythm model at the current moment. After obtaining these parameters, a fusion operation is performed, which is represented by the expansion or concatenation of feature vectors in the data processing flow. The real-time decoupled sensor data frame is merged with the queried individual physiological rhythm parameter vector to generate enhanced real-time sensor data with higher dimensions and richer information. This data not only describes the current physical and environmental state but also incorporates expected state information that conforms to the user's individual habits. This fusion process can be represented by the following formula: , in, The enhanced real-time sensor data, representing the cascaded and reassembled data, is a dimension equal to Dimensions and A single high-dimensional column vector whose dimensions are summed can be directly used as the input tensor for subsequent data-driven neural networks or knowledge-driven rule bases. for The real-time decoupled sensor data frame at any given moment is a high-dimensional vector containing multi-dimensional features of pressure, shear force, microenvironment, and spatial index. To be based on the current time Parameter vectors extracted from individual circadian rhythm models; and These are pre-defined dimensionless weighting coefficients, which are applied to the vectors via a scalar broadcasting mechanism. and The internal elements of the model, with values ​​ranging from 0 to 1, are used to adjust the relative importance of instantaneous data and rhythm parameters in the prediction model in more complex fusion algorithms. Their values ​​can be determined through offline model training optimization.

[0035] For example, query the current individual physiological rhythm model. Extract the output parameter vector from the rhythm parameters at each time point. The corresponding meanings are: 80% probability of pressure center shift, expected remaining stable time in the current supine position is 45 minutes, and normal reference temperature. Humidity 55%. Enter the fusion formula. In this phase, default values ​​for the weight coefficients, determined through offline grid search optimization, are set. , Using a scalar broadcast mechanism, the real-time shear force amplitude of the first monitoring point is extracted. implement Similarly, execute the expected settling time. Minutes. After traversing all elements, the 37-dimensional weighted real-time data is vertically concatenated with the 4-dimensional weighted rhythm parameters to generate a single high-dimensional column vector of 41 dimensions. This method uses real-time monitoring indicators as target prediction data. It organically combines the transient characteristics of real-time monitoring indicators with the steady-state expectations of individual historical rhythms, enabling downstream prediction systems to not only see "what is happening now" but also incorporate prior information on "what will happen based on individual patient habits," thereby improving the robustness of the prediction benchmark.

[0036] Optionally, obtaining individual physiological rhythm parameters corresponding to the sacrococcygeal region includes: Historical time-series sensing data is acquired through long-term monitoring using a biomimetic multilayer flexible sensing array. Periodic analysis of historical time-series sensor data is performed to extract periodic fluctuation patterns in pressure distribution and microenvironment parameters in the sacrococcygeal region. Based on the periodic fluctuation pattern, a physiological rhythm model representing individual user habits is constructed, and individual physiological rhythm parameters are calculated based on the current time.

[0037] Specifically, historical time-series sensing data is acquired by storing real-time decoupled sensing data streams collected by a biomimetic multilayer flexible sensing array at a preset sampling frequency over a continuous period of at least 7 to 14 days, forming a large dataset containing multi-dimensional features that change over time.

[0038] Periodic analysis was performed on the historical time-series sensor data to extract periodic fluctuation patterns during periods when the user was bedridden or sedentary. This analysis primarily focused on the time series of key features such as the displacement trajectory of the pressure center, the area of ​​the high-pressure region, and the average temperature. By applying signal processing algorithms such as Fourier transform or autocorrelation function analysis, the dominant periods of change in these features were identified, such as long periods related to the user turning over at night and adjusting sitting posture during the day, and short periods related to breathing and heart rate, thereby quantifying the frequency, amplitude, and phase information of the periodic fluctuation patterns.

[0039] Based on the extracted periodic fluctuation patterns, a physiological rhythm model representing individual user habits is constructed. This model is expressed as a function composed of the superposition of multiple harmonic components, with the following structure: , in, At the current time The calculated individual physiological rhythm parameters, such as the expected location of the pressure center or temperature value; This is the baseline value of the parameter, obtained by averaging the historical time-series sensor data over time. , , They are the first The amplitude, angular frequency, and phase of each harmonic component are all directly determined by the spectral peak analysis results of the Fourier transform in the periodic analysis. It is a periodic function, usually a sine or cosine function. For example... Figure 3 As shown in the figure, this diagram illustrates the process of constructing an individual's circadian rhythm model based on long-term historical monitoring scatter data with noise.

[0040] For example, the system retrieved the sacrococcygeal skin temperature sequence stream from the patient's historical time-series sensor data over the past 14 days. To extract the dominant period, the algorithm applied a Fast Fourier Transform to convert the time-domain temperature signal into a frequency-domain signal to find the energy spectrum peaks and determine the dominant fluctuation angular frequency of the patient's temperature periodic fluctuation pattern. Simultaneously, the baseline value is calculated from the 14-day global time mean. Extract the amplitude of this frequency band from the complex output of the FFT. and phase Applying the rhythm formula When the time point advances to 6:00 AM, the corresponding rhythm parameters are calculated by substituting the sinusoidal periodic function: This value serves as the baseline calibration value for the patient's temperature channel at the current moment. This method uses rigorous frequency domain transformation to condense massive amounts of long-historical time-series data into a low-computing-power model composed of simple harmonic superposition. It can reproduce highly personalized long-cycle physiological evolution patterns in real-time queries without consuming a large amount of edge computing resources.

[0041] Optionally, obtaining tissue damage risk prediction instructions that include risk level and primary cause classification includes: The first risk assessment result is obtained by performing data-driven feature mapping operations on the target prediction data. The second risk assessment result is obtained by performing knowledge-driven rule matching operations on the target prediction data. The results of the first risk assessment and the second risk assessment are weighted and integrated to generate a comprehensive risk score. The comprehensive risk score is mapped to a risk threshold range for numerical comparison to determine the risk level. The contribution of each causative factor in the first and second risk assessment results is analyzed. Based on the contribution, the main causative factors are classified and the risk levels are combined to obtain the tissue damage risk prediction instruction.

[0042] Specifically, the target prediction data is processed in parallel across two paths. The first path involves feature mapping operations driven by the data dimension. A pre-trained deep neural network, such as a multilayer perceptron with 3-5 hidden layers, performs a nonlinear transformation on the target prediction data. This network is trained by learning from a large number of historical cases containing sensor data and corresponding clinical tissue damage results. Its output, the first risk assessment result, is a continuous probability value between 0 and 1, directly representing the probability of tissue damage occurring.

[0043] The second approach involves rule matching calculations driven by knowledge. The same target prediction data is input into an expert rule base based on fuzzy logic. This rule base contains multiple IF-THEN rules developed by clinical medical experts, such as "if the pressure peak exceeds 4.5 kPa and lasts for more than 30 minutes, the pressure risk factor is set to high" or "if the shear force amplitude is greater than 1.5 N / cm² and the local humidity exceeds 85%, the combined shear force and microenvironment risk factor is set to extremely high." The activation intensity of each rule is calculated and inferences are performed to output a comprehensive second risk assessment result. This result is normalized to the range of 0 to 1, reflecting risk judgment based on medical knowledge.

[0044] The two results are then weighted and fused to generate a comprehensive risk score. This fusion process follows the formula below: , in, For comprehensive risk scoring; This is the result of the first risk assessment; This is the result of the second risk assessment; and The preset weighting coefficients, and This reflects the confidence allocation between data-driven models and knowledge-driven rules in the final decision-making process. Typically, during the system debugging phase, the configuration is optimized based on the performance of the two models on the validation set. It can be set to 0.6. Set it to 0.4.

[0045] Obtain a comprehensive risk score Then, the values ​​are mapped to a preset risk threshold range for numerical comparison to determine the risk level. For example, an S value between 0 and 0.3 is defined as Level 1, i.e., low risk; 0.3 to 0.6 is Level 2, i.e., medium risk; and above 0.6 is Level 3, i.e., high risk.

[0046] The retrospective analysis assesses the contribution of each causative factor in generating the first and second risk assessment results. For neural networks, sensitivity analysis or SHAP methods can be used to calculate the contribution of each input feature, such as pressure, shear force, temperature, and humidity, to the output. For rule bases, the main factor types involved in high-risk contribution rules are directly statistically analyzed. Based on the contribution ranking, the main causative categories are determined, such as "pressure-dominated," "shear force-dominated," or "microenvironment deterioration-type." Finally, the determined risk levels and main causative categories are concatenated into strings or constructed into a data structure to form the final tissue damage risk prediction instruction.

[0047] For example, the 41-dimensional data tensor is first fed into a multilayer perceptron with four hidden layers. After nonlinear activation mapping, it outputs the first risk assessment probability value of the tissue damage. Simultaneously, this data was input into a fuzzy logic rule base of clinical medical experts, and the rule base matched the patient's condition to "local pressure at the specified location reached..." And the humidity of the microenvironment The antecedent of "exceeding the standard" triggers a second risk assessment probability value after intensity normalization. Apply the preset default confidence weights and select... and Substitute into the fusion formula Compare risk threshold ranges: Set a level 1 Level 2 Level 3 Since 0.792 > 0.6, the risk level was locked at "Level 3" (high risk). The SHAP attribution algorithm was used to calculate the contribution percentage of each feature in the judgment process. The stress feature vector showed a contribution of 68%, ranking first. Therefore, the causative category "stress-dominant" was extracted, and finally, a "Level 3 - Stress-dominant" tissue damage risk prediction instruction was generated. This method bridges the gap between the lack of interpretability in purely data-driven models and the inability of purely knowledge-based rules to handle nonlinear, minute changes, ensuring that the output high-risk instruction is both sensitive and forward-looking, and strictly adheres to clinical medical pathology standards.

[0048] S3. Extract the scope and intensity of intervention based on tissue damage risk prediction instructions, and generate dynamic intervention instructions; Optionally, generating dynamic intervention instructions includes: Numerical mapping extraction is performed on the risk level to obtain the basic intervention intensity parameter; Based on the classification of the main causes, control logic lookup table matching is performed to obtain the parameters of the target intervention mode; The basic intervention intensity parameters and the target intervention mode parameters are combined in a time sequence to generate an initial intervention instruction sequence; Based on the spatial distribution information in the real-time decoupled sensor data, the initial intervention command sequence is spatially registered to generate dynamic intervention commands that include execution location, force magnitude and direction, and timing.

[0049] Specifically, the process begins with parsing the risk level in the tissue damage risk prediction instruction. For each risk level, such as "Level 1," "Level 2," and "Level 3," a numerical mapping extraction is performed to convert it into a quantified baseline intervention intensity parameter. This is a table lookup operation; for example, Level 1 risk corresponds to a baseline intervention intensity parameter of 0.2, Level 2 to 0.5, and Level 3 to 0.9.

[0050] The control logic performs a lookup table matching based on the primary causative classification in the predictive instructions to determine the target intervention mode parameters. These parameters define the type and strategy of the intervention action. For example, if the primary causative classification is "pressure-dominated," the matched target intervention mode parameter is defined as "regional airbag group alternating decompression mode"; if it is "shear force-dominated," it is matched as "small translational cancellation mode opposite to shear force"; and if it is "microenvironment deterioration type," it is matched as "local ventilation or temperature control system activation mode."

[0051] After obtaining the baseline intervention intensity parameters and the target intervention mode parameters, the two are combined temporally to generate an initial intervention instruction sequence. This sequence defines the macroscopic rhythm and intensity of the intervention actions, such as "perform alternating decompression of regional airbag groups at an intensity of 0.5, repeating once every 15 minutes."

[0052] Based on the spatial distribution information in the real-time decoupled sensor data, the initial intervention command sequence is spatially registered to generate the final dynamic intervention command. This process utilizes the pressure and shear force distribution map provided by the sensor array to locate the coordinates of the highest-risk areas. For example, the sensor unit indexes of pressure peak points or shear force maximum values ​​are identified. Subsequently, the coordinate information of these high-risk areas is embedded into the intervention command, and the specific physical quantities acting on that area are calculated in conjunction with the basic intervention intensity parameters, such as the target air pressure value of the airbag to be inflated, or the magnitude and direction vector of the force required by the actuator to perform the counteracting action. The final generated dynamic intervention command is a highly structured data packet that defines the location, the force, and the time pattern for performing the physical intervention.

[0053] For example, after intercepting the "Level 3 - Pressure-Dominated" risk instruction, the system executes a lookup logic mapping. It retrieves the basic intervention intensity comparison table and finds that the numerical basic intervention intensity parameter directly corresponding to the "Level 3" risk is set to 0.9. It then retrieves the main cause matching table and matches "Pressure-Dominated" to the target intervention mode parameter "Regional Airbag Group Alternating Decompression Mode". After temporal combination, a macro-rhythm is generated: "Execute airbag decompression at an intensity of 0.9, with an alternating execution cycle of 10 minutes." Subsequently, spatial registration is performed. The system retrieves spatial distribution information from preceding data frames, locates the source of this high risk as the sensor physical unit at coordinate index (10, 20), and calculates the target physical force of the corresponding airbag: the target inflation pressure equals the system's rated pressure. Multiply by the intensity reversal factor, since pressure ulcers require decompression, the factor is... ,Right now After considering the above parameters, the final output dynamic intervention command is: "Lock the mattress plane coordinates (10,20) area, and control the bottom airbag release pressure to the target value." "and maintain a dynamic, alternating emptying rhythm for 10 minutes." This method transforms abstract qualitative conclusions about medical risks into readable and executable parameter control logic for underlying physical systems such as mechanical microfluidics or servo motors, thus bridging the key gap from risk discovery to automated processing.

[0054] S4. Send the dynamic intervention command to the active offset dynamic intervention execution mechanism for command parsing and control conversion, and generate physical intervention actions that match the main cause classification; Optionally, generating physical intervention actions that match the primary causative category includes: The dynamic intervention commands are downgraded and formatted to obtain the execution unit control commands. The execution unit control commands are parsed and the execution position parameters, action timing parameters, and action force parameters are extracted. When the main cause is classified as shear force, the action force parameters are configured as a counteracting force vector opposite to the direction of shear force. By utilizing the execution position parameters, action timing parameters, and counteracting force vector, the physical execution unit in the active counteracting dynamic intervention actuator is driven to perform intervention operations and generate physical intervention actions.

[0055] Specifically, the received dynamic intervention commands undergo downsizing and format conversion. Dynamic intervention commands are data structures containing complex spatiotemporal information, while the underlying physical execution units, such as micro-pumps, solenoid valves, or linear actuators, typically only accept simple switching signals, pulse-width modulation (PWM) signals, or serial communication protocols. Therefore, the magnitude of the force in the command needs to be converted into the corresponding physical execution unit's drive voltage, PWM duty cycle, or target position encoding, and the timing sequence needs to be decomposed into a series of timestamped discrete control events to generate execution unit control commands.

[0056] The control system parses the control commands from the execution units, extracting key execution parameters: execution position parameters, action timing parameters, and action force parameters. Execution position parameters are used for addressing, activating specific physical execution units located below the risk area; action timing parameters control the start, duration, and end times of these unit actions. When parsing the action force parameters, conditional judgments are made based on the primary cause classification inherited from the dynamic intervention commands. When the primary cause classification is shear force dominant, special configuration logic is triggered. At this time, the action force parameter is configured as a counteracting force vector. The magnitude of this counteracting force vector is proportional to the shear force amplitude monitored in the real-time decoupled sensor data, with the specific proportion determined by the basic intervention intensity parameter, and its direction is strictly opposite to it. Its calculation formula can be expressed as: , in, The force vector is the force vector that cancels out the force. It is the shear force vector monitored in real time, which contains information on magnitude and direction; It is a dimensionless intervention intensity coefficient, whose value is derived from the baseline intervention intensity parameter, and its range is between 0 and 1. The negative sign indicates the opposite direction.

[0057] By utilizing the analyzed and configured execution position parameters, timing parameters, and force parameters, especially the counteracting force vector configured for shear force-dominated risks, electrical signals are sent to the corresponding physical execution units in the active counteracting dynamic intervention actuator via a drive circuit. Upon receiving the signals, these execution units perform intervention operations, such as micro-inflation / deflation of the airbag to adjust the pressure distribution, or micro-actuators applying a reverse shear force through minute displacement in a specific direction. This generates physical intervention actions that match the cause of the risk, directly acting on the sacrococcygeal tissue interface to achieve source intervention of the risk.

[0058] For example, taking the underlying control in actual mattress intervention as an example, when the primary cause at a certain moment is classified as "shear force dominant", and the extracted intervention intensity parameter coefficients are... The configuration is set to 0.6. The underlying microcontroller interprets the instruction as being applied to the high-risk shear zone on the right side of the mattress. Real-time decoupled sensor data indicates that the shear force vector at this point is at its maximum peak value. The force is applied in a longitudinal direction parallel to the body, tilting towards the feet at an angle of [angle missing]. Triggering force vector configuration formula The computational logic calculates the absolute value of the size that the underlying mechanism needs to generate. The thrust. Through the inversion of the negative sign in the formula, the direction of the intervention force vector is set as... It stretches towards the head. The control module will... The physical force threshold is converted into a PWM pulse signal with a 60% duty cycle for the 48V DC brushless motor on the actuation side, which then activates the corresponding electromagnetic push-pull rod. The execution unit then applies a mechanical action at a given spatial position that precisely matches the set inverse compensation vector. For skin tearing "shear force" injuries that traditional anti-pressure sore air mattresses struggle to address, this method creatively neutralizes external forces precisely through dynamic inverse vector synthesis and active mechanical displacement, severing the risk of subcutaneous microvascular deformation and rupture at the physical source.

[0059] S5. Collect the sensor change signals reflected by the biomimetic multilayer flexible sensor array after the active cancellation dynamic intervention actuator performs physical intervention actions, and generate tissue response data. Optionally, generating organizational response data includes: The sensing sequence continuously collected by the biomimetic multilayer flexible sensing array within the feedback time window is obtained to obtain the sensing data after intervention. Mean filtering and variance extraction were performed on the sensor data after the intervention to calculate the steady-state characteristic values ​​after the intervention took effect. Extract the difference vector between the steady-state feature values ​​and the corresponding feature values ​​before intervention, and perform standardized mapping to generate quantitative indicators of intervention effect; By concatenating and linking quantitative indicators of intervention effectiveness with dynamic intervention command execution data segments, organizational response data is generated.

[0060] Specifically, from the continuous data stream of the biomimetic multilayer flexible sensor array, a sensing sequence within a preset feedback time window is extracted. The length of this time window is typically set to 30 seconds to 2 minutes to ensure that the immediate and stable responses of the tissue to the intervention can be captured. This extracted sequence is the post-intervention sensing data.

[0061] To eliminate the interference of transient fluctuations and noise on the evaluation results, the sensor data after intervention were filtered and statistically processed. A mean filtering algorithm was applied to key feature values ​​such as pressure, shear force, temperature, and humidity in the sequence to smooth out high-frequency noise, and then the variance of the filtered sequence was calculated. When the variance is less than a preset stability threshold, for example, less than 1% of the mean, the tissue state is considered to have reached a new steady state under intervention, and the mean of this period is extracted as the steady-state feature value after the intervention takes effect.

[0062] The corresponding feature value at the moment immediately before the intervention is executed is extracted as a baseline. This baseline is then subtracted from the calculated steady-state feature value to form a difference vector. To facilitate comparisons between different physical quantities and subsequent model processing, this difference vector is standardized, for example, using a max-min normalization method to map it to the interval -1 to 1, thereby generating a dimensionless quantitative index of the intervention effect. This index can be calculated as follows: , in, Quantify the effectiveness of the intervention; The steady-state characteristic value after the intervention takes effect; The corresponding feature value before intervention; and These are the maximum and minimum values ​​of the feature in historical data, used for normalization.

[0063] The generated quantitative indicators of intervention effectiveness are concatenated and linked with the execution data segment of the dynamic intervention command that triggered the intervention. This data segment contains all contextual information, including the type, intensity, location, and time of the intervention. In this way, a complete organizational response data record is generated. This record not only quantifies "how effective" the intervention was, but also indicates "under what intervention the effect was achieved."

[0064] For example, after the cancellation action is performed, the sensor initiates a 60-second feedback time window to apply an arithmetic mean filtering algorithm to the acquired 100Hz high-frequency sensor data after intervention. Statistical calculations show that the sliding variance of the shear force sequence has been reduced to its mean after filtering. of If the steady-state condition of less than 1% is met, then... Extract steady-state characteristic values ​​after the intervention takes effect. Retrieve characteristic values ​​corresponding to the severely stressed baseline measured in the last second before the intervention. By reviewing the patient's recent historical data, the maximum peak value of the shear wave was determined. Minimum value Normalization was performed using the max-min standardized mapping formula: Quantitative indicators of intervention effect. Since the dimensionless scale is negative, it proves that the harmful injury force is reduced by a staggering 46.7% compared to the full range. This nominal parameter -0.467 is encrypted and structured with the previously triggered underlying PWM drive command log execution space address and stored in the library. This method utilizes strict statistical thresholds to define biological homeostasis and outputs normalized quantitative physical correction indicators, ending the dilemma of clinical nursing relying on subjective judgment of intervention effectiveness and enabling medical auxiliary equipment to adaptively adjust and form a closed loop of quantitative evidence.

[0065] S6. Based on the tissue response data, optimize the bidirectional driving inference analysis and update the inference parameters of the output tissue damage risk prediction command.

[0066] Optionally, the optimization of the bidirectional drive deduction analysis includes: The actual risk change trend is extracted from the organizational response data and compared with the organizational damage risk prediction instructions to generate a prediction error signal; Using the prediction error signal, the backpropagation algorithm is employed to adjust the connection weights of the neural network used for feature mapping operations in the data-driven dimension; Based on the correlation between the prediction error signal and the main cause classification, the confidence parameter of the prior rule used for rule matching operation in the knowledge-driven dimension is corrected.

[0067] Specifically, quantitative indicators of intervention effectiveness, representing the actual risk change trend after intervention, are extracted from organizational response data, and the actual risk score is recalculated based on the steady-state characteristic values ​​after intervention. The actual risk score is then compared with the predicted risk score in the pre-intervention organizational damage risk prediction instruction to generate a quantified prediction error signal. The formula for calculating this signal can be expressed as: , in, This is the prediction error signal; The comprehensive risk score is generated by the two-way driven extrapolation and analysis before intervention; The actual risk score is calculated using the same risk assessment algorithm based on the steady-state sensor data collected after the intervention.

[0068] Using this prediction error signal This involves optimizing the feature mapping operations for the data-driven dimension. This is accomplished through iterative execution of the backpropagation algorithm. The prediction error signal is then processed. As the target value of the loss function, the loss gradient is calculated and backpropagated to each layer of the neural network. Depending on the preset learning rate, this value is typically set between 0.001 and 0.01 to fine-tune the connection weights of all neurons in the network. This process allows the neural network to more closely approximate the observed risk state when it encounters similar input data in the future.

[0069] In parallel, based on the prediction error signal The correlation with the primary causal classification in the tissue damage risk prediction instruction is used to correct the confidence parameters of the prior rules used for rule matching operations in the knowledge-driven dimension. The main activation rules leading to this prediction are analyzed. If the prediction error... If the absolute value of E is large, it indicates that the prediction is inaccurate, and the primary cause at that time was classified as "shear force dominant." In this case, the confidence parameter of the expert rule associated with high shear force risk should be selectively reduced, for example, by multiplying the parameter by a decay factor less than 1, such as 0.98. Conversely, if the prediction is accurate, the absolute value of E is close to 0. In this case, the confidence parameter of the relevant rule that contributed to the correct judgment should be moderately increased, for example, by multiplying it by an enhancement factor slightly greater than 1, such as 1.02, but ensuring that its value does not exceed the preset upper limit of 1.0.

[0070] For example, for the system's intervention response in the first half of the night, the same fusion evaluation algorithm is invoked again to perform deep verification and computing power validation of the baseline steady-state characteristics before the intervention, resulting in an actual risk score with extremely high accuracy at the baseline time. ; Review the comprehensive score of the original tissue damage risk prediction instruction calculated by the preceding feedforward. The prediction error signal between the two is obtained based on the formula. Since the error value deviated to positive, it indicates that the neural network's initial prediction slightly overestimated the onset threshold, and the system immediately... The loss function gradient is converted into a forward propagation loss function, supplemented by a preset adaptive learning rate parameter. Perform backpropagation to update the connection weights of the activated synaptic network of the first hidden layer neurons in the multilayer perceptron that are connected to the high-frequency feature channel of pressure, such as by Simultaneously reduced to Parallel analytical analysis revealed that in the expert rule base of the knowledge-driven dimension, the prior rule regarding "humidity exceeding 60%" made a positive contribution with precise direction in this response. To solidify the priority of this rule, a confidence-based adjustment logic was used to assign it an enhancement parameter of 1.02, increasing its weight from 0.80 to [previous value]. This method fully utilizes the error gradient sources accumulated during long-term system operation and the objective interaction generated by its own actions, enabling the core computing layer of the intelligent monitoring system to achieve autonomous evolution of personalized features under unattended conditions. It becomes more accurate with use and consistently matches the dynamic deterioration and evolution of complications and signs in patients over a long period of time.

[0071] Based on the same inventive concept, such as Figure 4 As shown, the present invention also provides a sacrococcygeal shear force decoupling and microenvironment evolution prediction and control system, comprising: The multidimensional signal acquisition and decoupling module is used to acquire multidimensional raw signals of pressure, shear force and microenvironment output by the biomimetic multilayer flexible sensor array deployed in the sacral and coccygeal region, perform cross-decoupling calculations, and obtain real-time decoupling sensor data. The risk extrapolation and prediction module is used to construct target prediction data based on real-time decoupled sensor data, perform bidirectional driven extrapolation analysis of data feature mapping and prior rule matching, and obtain tissue damage risk prediction instructions that include risk level and main cause classification. The dynamic intervention instruction generation module is used to extract the intervention range and intensity based on the tissue damage risk prediction instruction and generate dynamic intervention instructions. The intervention execution and control module is used to send dynamic intervention instructions to the active counteracting dynamic intervention execution mechanism for instruction parsing and control conversion, and generate physical intervention actions that match the main cause classification; The tissue response perception feedback module is used to collect the sensor change signals reflected by the biomimetic multilayer flexible sensor array after the active cancellation dynamic intervention actuator performs physical intervention actions, and generate tissue response data. The inference parameter closed-loop optimization module is used to optimize the bidirectional driven inference analysis based on tissue response data and update the inference parameters of the output tissue damage risk prediction command.

[0072] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0073] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for decoupling shear force in the sacrococcygeal region and predicting and controlling the evolution of the microenvironment, characterized in that, The method includes: The pressure, shear force and microenvironment multidimensional raw signals output by the biomimetic multilayer flexible sensor array deployed in the sacral and coccygeal region are collected, and cross-decoupling calculations are performed to obtain real-time decoupled sensor data. Based on the real-time decoupled sensor data, target prediction data is constructed, and bidirectional driven deductive analysis of data feature mapping and prior rule matching is performed to obtain tissue damage risk prediction instructions that include risk level and main cause classification. Based on the tissue damage risk prediction instructions, the intervention range and intensity are extracted to generate dynamic intervention instructions; The dynamic intervention command is sent to the active offset dynamic intervention execution mechanism for command parsing and control conversion, generating a physical intervention action that matches the main cause classification; The biomimetic multilayer flexible sensor array collects the sensor change signals reflected after the active cancellation dynamic intervention actuator performs the physical intervention action, and generates tissue response data. Based on the tissue response data, the bidirectional driving inference analysis is optimized, and the inference parameters for outputting the tissue damage risk prediction command are updated.

2. The method for decoupling sacrococcygeal shear force and predicting and controlling microenvironment evolution according to claim 1, characterized in that, The obtained real-time decoupled sensing data includes: The first raw signal of the sacral and coccygeal monitoring point is simultaneously acquired through the pressure sensing layer of the biomimetic multilayer flexible sensor array. The second raw signal of the sacral and coccygeal monitoring point is simultaneously acquired through the shear force sensing layer of the biomimetic multilayer flexible sensing array. The third original signal of the sacral and coccygeal monitoring point is simultaneously acquired through the microenvironment sensing layer of the biomimetic multilayer flexible sensing array. Signal decoupling and feature extraction are performed on the first original signal, the second original signal and the third original signal respectively to obtain independent pressure feature vector, shear force feature vector and microenvironment feature vector; The pressure feature vector, the shear force feature vector, and the microenvironment feature vector are fused to generate real-time decoupled sensing data.

3. The method for decoupling sacrococcygeal shear force and predicting and controlling microenvironment evolution according to claim 1, characterized in that, The construction of target prediction data based on the real-time decoupled sensing data includes: The individual physiological rhythm parameters corresponding to the sacrococcygeal region are obtained and fused with the real-time decoupled sensing data to generate enhanced real-time sensing data, which is used as target prediction data.

4. The method for decoupling sacrococcygeal shear force and predicting and controlling microenvironment evolution according to claim 3, characterized in that, The acquisition of individual physiological rhythm parameters corresponding to the sacrococcygeal region includes: Historical time-series sensing data is acquired through long-term monitoring using the aforementioned biomimetic multilayer flexible sensing array. Periodic analysis is performed on the historical time-series sensing data to extract the periodic fluctuation patterns of the sacrococcygeal region in terms of pressure distribution and microenvironment parameters; Based on the periodic fluctuation pattern, a physiological rhythm model representing individual user habits is constructed, and individual physiological rhythm parameters are calculated based on the current time.

5. The method for decoupling sacrococcygeal shear force and predicting and controlling microenvironment evolution according to claim 4, characterized in that, The instructions for obtaining tissue damage risk prediction, which include risk level and main cause classification, include: Perform feature mapping operations on the target prediction data in a data-driven dimension to obtain the first risk assessment result; The target prediction data is subjected to knowledge-driven rule matching operations to obtain a second risk assessment result; The first risk assessment result and the second risk assessment result are weighted and fused to generate a comprehensive risk score; The comprehensive risk score is mapped to a risk threshold range for numerical comparison to determine the risk level. The contribution of each causative factor in the first risk assessment result and the second risk assessment result is analyzed. The main causative factors are classified according to their contribution. The risk levels are then combined to obtain the tissue damage risk prediction instruction.

6. The method for decoupling sacrococcygeal shear force and predicting and controlling microenvironment evolution according to claim 1, characterized in that, The generation of dynamic intervention instructions includes: Numerical mapping extraction is performed on the risk level to obtain the basic intervention intensity parameter; Based on the classification of the main causes, a control logic lookup table matching is performed to obtain the target intervention mode parameters; The basic intervention intensity parameters and the target intervention mode parameters are combined in a time sequence to generate an initial intervention instruction sequence; Based on the spatial distribution information in the real-time decoupled sensor data, the initial intervention command sequence is spatially registered to generate dynamic intervention commands that include execution location, force magnitude and direction, and timing.

7. The method for decoupling shear force and predicting and controlling microenvironment evolution in the sacrococcygeal region according to claim 1, characterized in that, The generation of physical intervention actions that match the primary causative classification includes: The dynamic intervention command is downgraded and formatted to obtain the execution unit control command; The execution unit control command is parsed to extract the execution position parameter, action timing parameter and action force parameter. When the main cause is classified as shear force dominant, the action force parameter is configured as a counteracting force vector opposite to the direction of shear force. Using the execution position parameters, the action timing parameters, and the counteracting force vector, the physical execution unit in the active counteracting dynamic intervention execution mechanism is driven to perform intervention operations and generate physical intervention actions.

8. The method for decoupling shear force and predicting and controlling microenvironment evolution in the sacrococcygeal region according to claim 1, characterized in that, The generated organizational response data includes: The sensing sequence continuously collected by the biomimetic multilayer flexible sensing array within the feedback time window is obtained to obtain the sensing data after intervention. The mean filter and variance extract are performed on the sensor data after the intervention to calculate the steady-state characteristic value after the intervention takes effect; Extract the difference vector between the steady-state feature value and the corresponding feature value before intervention, and perform standardized mapping to generate a quantitative index of intervention effect; The quantitative indicators of the intervention effect are concatenated and associated with the data segment of the dynamic intervention instruction execution to generate organizational response data.

9. The method for decoupling shear force and predicting and controlling microenvironment evolution in the sacrococcygeal region according to claim 5, characterized in that, The optimization of the bidirectional drive deduction analysis includes: The actual risk change trend is extracted from the tissue response data and compared with the tissue damage risk prediction instruction to generate a prediction error signal; Using the prediction error signal, the backpropagation algorithm is employed to adjust the connection weights of the neural network used for feature mapping operations in the data-driven dimension; Based on the correlation between the prediction error signal and the main cause classification, the confidence parameter of the prior rule used for rule matching operation in the knowledge-driven dimension is corrected.

10. A sacrococcygeal shear force decoupling and microenvironment evolution prediction and control system, characterized in that, The system includes: The multidimensional signal acquisition and decoupling module is used to acquire multidimensional raw signals of pressure, shear force and microenvironment output by the biomimetic multilayer flexible sensor array deployed in the sacral and coccygeal region, perform cross-decoupling calculations, and obtain real-time decoupling sensor data. The risk extrapolation and prediction module is used to construct target prediction data based on the real-time decoupled sensor data, perform bidirectional driven extrapolation analysis of data feature mapping and prior rule matching, and obtain tissue damage risk prediction instructions that include risk level and main cause classification. The dynamic intervention instruction generation module is used to extract the intervention range and intensity based on the tissue damage risk prediction instruction and generate dynamic intervention instructions. The intervention execution and control module is used to send the dynamic intervention command to the active offset dynamic intervention execution mechanism for command parsing and control conversion, and generate physical intervention actions that match the main cause classification; The tissue response sensing feedback module is used to collect the sensing change signals reflected by the biomimetic multilayer flexible sensing array after the active cancellation dynamic intervention actuator performs the physical intervention action, and generate tissue response data. The inference parameter closed-loop optimization module is used to optimize the bidirectional driven inference analysis based on the tissue response data and update the inference parameters that output the tissue damage risk prediction command.

Images