Dual moxa and far infrared composite energy system of moxa chair and regulation method

Abstract

This invention belongs to the field of intelligent control technology for moxibustion chairs, specifically involving a dual-moxibustion and far-infrared composite energy system and control method for a moxibustion chair. It collects user physiological characteristics and energy transfer parameters to generate a fused feature vector, then calculates the remaining efficacy distribution map based on moxa cake state data using a pharmacokinetics release kinetic model; establishes a user energy-physiological response mapping relationship and predicts physiological state; and employs a collaborative control unit to fuse the remaining efficacy map, predict physiological state, and formulate conditioning target instructions. A pre-trained deep strategy network generates basic control strategies, which are then refined into spatiotemporal collaborative control instructions for regulating each independent execution unit. This invention achieves precise and adaptive collaborative output of the heat field and the pharmacokinetics field in space and time through supply-side efficacy perception, demand-side personalized modeling, and intelligent collaborative decision-making, significantly improving the depth, personalization, and safety of moxibustion therapeutic effects.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for moxibustion chairs, specifically relating to a dual-moxibustion and far-infrared composite energy system and control method for a moxibustion chair. Background Technology

[0002] Moxibustion, a traditional Chinese medicine therapy, stimulates acupoints and meridians through heat or medicinal properties to achieve the health benefits of dispelling cold and dampness, and regulating qi and blood. With the advancement of health technology, electronic moxibustion devices such as moxibustion chairs or moxibustion instruments have been widely used, aiming to improve ease of use, safety, and standardization while retaining the traditional efficacy of moxibustion.

[0003] Existing moxibustion chair applications still have significant limitations in their intelligent control level, restricting the depth of their therapeutic effects and personalized experience. This is manifested in the following ways: at the energy output level, most devices simply superimpose or switch modes between far-infrared heat radiation and the release of medicine from the moxa stick, lacking a complex control logic that can precisely coordinate across time and space. This fails to effectively construct a synergistic pathway where the "heat field" promotes the penetration of the "medicinal field." For example, while far-infrared radiation can promote local blood circulation and pore dilation, if it is not synchronized with the release rhythm of the effective components of moxa, it is difficult to achieve the enhanced effect of medicinal penetration, affecting the overall depth and efficiency of the moxibustion's therapeutic effect.

[0004] At the state perception level, existing devices typically use a control system to control the heating equipment to heat the moxa cake using a pre-set control program. As a core consumable, the moxa cake itself lacks information storage and interaction capabilities. The system cannot identify its identity, cumulative usage time, remaining efficacy, or other key states. The system can only heat the moxa cake according to the pre-set program and cannot grasp the usage status of the moxa cake. This leads to a disconnect between heating control and the actual efficacy of the moxa cake, making it impossible for the system to determine whether the remaining efficacy of the moxa cake meets the volatilization efficiency required for moxibustion, and thus unable to determine the moxibustion effect. On the other hand, the heating control of existing devices mostly relies on preset programs and cannot adaptively adjust according to individual user characteristics such as body shape differences and real-time vital sign feedback.

[0005] Therefore, in existing control systems, due to the lack of understanding and adaptive adjustment of the supply side (moxa cake) and demand side (user) status, there is a lack of adaptability between the system's energy output and the user's actual needs, which affects the optimization of moxibustion efficacy and the user's safety and comfort. Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, this invention provides a dual-moxibustion and far-infrared composite energy system and control method for a moxibustion chair. This solves the technical problems of existing moxibustion chairs, which suffer from weak sensing capabilities and rigid control logic, resulting in an inability to sense the state of the moxa cake and individual user needs, and an inability to effectively utilize the composite energy, making it difficult to achieve personalized dual-energy synergistic dynamic control, thus restricting the therapeutic effect and user experience.

[0007] The objective of this invention can be achieved through the following technical solution: a dual-moxibustion and far-infrared composite energy system for a moxibustion chair, the system comprising: The multimodal perception module is used to collect user physiological characteristic data and physical energy transfer parameters acting on the user. After data extraction and feature fusion of the collected data, a multimodal fusion feature vector is generated. The reused execution drive module includes a drug-driven unit and a far-infrared driven unit. The drug-driven unit and the far-infrared driven unit are used to receive control commands from the main control module and perform coordinated energy output to specific acupoints. The medicinal power driving unit includes moxa cakes that can be heated independently in separate zones, and each moxa cake has a readable unique identification code. The main control module is communicatively connected to the multiplexed execution drive module and the multimodal perception module, and the main control module includes a drug release calculation unit, a physical release calculation unit and a collaborative control unit; The drug release calculation unit is used to acquire information about the current moxa cake and its current state data, and to calculate the remaining drug effect distribution map and the overall release rate of the moxa cake based on the pre-trained drug release calculation model. The state data includes the spatial temperature field matrix of the moxa cake itself and the environmental tensor of the heating cavity. The physical release calculation unit is used to establish the energy-physiological response mapping relationship of the current user based on the user's physiological characteristic data and the physical energy transfer parameters acting on the user, and to predict the predicted physiological response state of the next control cycle. The collaborative control unit is used to acquire the remaining drug efficacy distribution map, the user's predicted physiological response state corresponding to the energy-physiological response mapping relationship, and the current conditioning target instruction. The acquired data is input into a pre-trained deep policy network to output a basic control strategy. Based on the basic control strategy, spatiotemporal collaborative control instructions for regulating each independent unit in the drug efficacy driving unit and the far-infrared driving unit are generated and issued.

[0008] Preferably, in the multimodal sensing module, the user physiological characteristic data includes gridded temperature data reflecting the local body surface temperature distribution of the user, acupoint bioelectrical impedance data reflecting changes in the state of meridian tissues, and blood perfusion data. The physical energy transfer parameters acting on the user include net heat flux density data and pressure distribution data.

[0009] Preferably, in the drug release unit, the heating cavity environment tensor includes a volatile organic compound concentration distribution matrix, an air flow velocity scalar, and a relative humidity scalar, where: The volatile organic compound concentration distribution matrix is used to quantify the relative concentration spatial distribution of the characteristic volatile substances of wormwood in the air directly above the wormwood cake and close to its volatilization surface; The air flow velocity scalar is used to quantify the average flow velocity of the air circulation in the heating cavity of the wormwood cake; The relative humidity scalar is used to quantify the average flow velocity of the air circulation in the heating cavity of the wormwood cake.

[0010] Preferably, in the drug release unit, calculating the remaining drug efficacy distribution map and the overall drug release rate of the current wormwood cake according to a pre-trained drug release calculation model includes the following steps: For each partition in the wormwood cake, apply the improved kinetic equation of the drug release kinetic model to calculate the current instantaneous drug release rate; Construct a state update equation through a discrete difference equation, and calculate the remaining effective ingredient concentration of each partition of the wormwood cake according to the state update equation; Arrange the calculated remaining effective ingredient concentrations of each partition according to the spatial position of the corresponding partition to obtain a remaining drug efficacy distribution map in the form of a heat map.

[0011] Preferably, in the drug release unit, the calculation formula for the current instantaneous drug release rate is: ; In the formula, is the instantaneous drug release rate of the partition (i,j) of the wormwood cake at time t, is the ideal gas constant, is the real-time temperature of the partition (i,j) of the wormwood cake at time t, 、 are the intrinsic release parameters of the wormwood cake, is the pre-factor, is the apparent activation energy of the wormwood cake, is the environmental correction function, is the volatile organic compound concentration distribution matrix of the partition (i,j) of the wormwood cake at time t, is the air flow velocity scalar at time t, is the relative humidity scalar at time t, is the previous calculation moment of the partition (i,j) of the real-time relative remaining effective ingredient concentration.

[0012] Preferably, the calculation formula for the environmental correction function is: In the formula, Concentration inhibition term, For reference concentration calibration value, It is the concentration inhibition coefficient. This is an airflow enhancement term. It is the gas enhancement coefficient. This is a humidity suppression term. This is the humidity suppression coefficient.

[0013] Preferably, the process of determining the intrinsic release parameters of the mugwort cake includes: For each batch of mugwort cakes, the stable release rate of the mugwort cakes at multiple temperature points was measured through standard experiments. At each temperature point, the change in drug release rate R over time was precisely measured and recorded. The average value of R during the steady-state phase was taken, and the corresponding absolute temperature T was recorded simultaneously. Obtain multiple sets of (R,T) and fit the linearized form of the Arrhenius equation. The fitting formula is as follows: ; by The x-axis is... Using the vertical axis as the ordinate, the experimental data points are plotted on a two-dimensional coordinate system, and the fitted line is obtained through linear regression. The intercept of the fitted line is obtained by calculating... Calculate the slope of the fitted line to obtain .

[0014] Preferably, in the collaborative control unit, the deep policy network adopts an encoder-decoder architecture based on an Actor-Critic framework, specifically including: Encoders are constructed using multilayer perceptrons or transformers that incorporate attention mechanisms to gain a deep understanding of the complex relationships that jointly represent states. The strategy header (Actor) is used to output the basic control strategy; The value head Critic is used to output the state value estimate.

[0015] A method for regulating the dual-moxibustion and far-infrared composite energy system of a moxibustion chair, the method comprising the following steps: S1: Real-time acquisition of user physiological characteristic data and physical energy transfer parameters acting on the user; data extraction and feature fusion of the two acquired data to generate a multimodal fusion feature vector. S2: Read the current moxa cake information, the spatial temperature field matrix of the moxa cake itself, and the heating cavity environment tensor. Based on the pre-trained drug release kinetic model, calculate and output the remaining drug effect distribution map and overall release rate of the current moxa cake. S3: Based on the multimodal fusion feature vector, establish and update the energy-physiological response mapping relationship of the current user, and output the predicted physiological response state for the next control cycle; S4: Receive the remaining drug efficacy distribution map, predicted physiological response state and current conditioning target instruction, fuse the received data into a unified joint state representation vector, input it into a pre-trained deep policy network, and output the basic control policy. S5: Based on the basic control strategy, refine the instructions and verify safety and compliance, and generate spatiotemporal coordinated control instructions for independently regulating each execution unit in the drug-driven unit and the far-infrared driven unit.

[0016] The beneficial effects of this invention are as follows: This invention achieves precise synergy and dynamic compensation of thermal energy and medicinal energy in time and space. The output control commands control the power timing and phase of each independent execution unit (far-infrared emission point, moxa cake heating zone), realizing multiple dynamic synergy modes such as first preheating with far-infrared to open pores, and then precisely releasing the medicinal power of moxa or alternating heat and medicine at the same frequency. The heat field promotes the synergistic pathway of medicinal field penetration, giving full play to the combined energy synergistic effect of heat field and medicinal field, and significantly enhancing the warming and unblocking of meridians and the depth and efficiency of medicinal penetration of moxibustion. By reading the identity of the moxa cake to obtain its intrinsic release parameters, and combining it with real-time temperature field and cavity environment data, the system dynamically calculates and displays the remaining medicinal effect distribution map of the moxa cake. The system quantifies the real-time medicinal power of the moxa cake through the remaining medicinal effect distribution map, thereby intelligently adjusting the heating strategy according to the medicinal effect status, and realizing the precise matching between heating control and the actual efficacy of the moxa cake. It also adaptively adjusts based on each user's real-time physiological feedback, ensuring the individualized optimal treatment dosage and stimulation intensity, greatly improving user comfort and acceptance. Based on real-time safety monitoring and predictive compliance verification of heat flux density, moxa cake temperature, and user body surface temperature, it can proactively prevent risks such as burns. At the same time, precise management of the moxa cake status avoids ineffective or overheating, improving the economy and safety of the equipment. Attached Figure Description

[0017] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0018] Figure 1 This is a schematic diagram of the system structure of the present invention; Figure 2A schematic diagram illustrating the steps of calculating the remaining efficacy distribution map and the overall release rate of the moxa cake in the drug release unit of the present invention; Figure 3 This is a schematic diagram of the method steps of the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0020] Please see Figures 1-2 This embodiment provides a dual-moxibustion and far-infrared composite energy system for a moxibustion chair, characterized in that: the system includes: The multimodal perception module is used to collect user physiological characteristic data and physical energy transfer parameters acting on the user. It performs data extraction and feature fusion on the collected data to generate a multimodal fusion feature vector. The multimodal perception module includes a data acquisition unit and a data fusion and feature extraction unit connected in sequence. The specific functions of the two units are as follows: The data acquisition unit includes: an infrared temperature sensor array for gridded temperature data reflecting the local surface temperature distribution of the user; flexible bioelectrical impedance measurement electrodes at preset acupoint locations for collecting acupoint bioelectrical impedance data reflecting changes in the state of meridian tissues; a photoelectric blood flow sensor for collecting blood perfusion data reflecting the local microcirculation state; and a skin conductivity sensor for collecting skin conductivity signals reflecting the local sweating state. The flexible bioelectrical impedance measurement electrodes are integrated in an array within the contact pad of the moxibustion chair, and their positions are pre-calibrated and matched with the anatomical locations of the main acupoints on the back or abdomen of the human body. A heat flux density sensor is used to collect net heat flux density data transmitted vertically to the user's body surface. A contact pressure sensor array is used to collect pressure distribution data reflecting the degree of contact with the user's body. The heat flux density sensor is a precision sensing element used to directly measure the heat power passing through a unit area. In this system, sensors based on the principles of thin-film thermopile or heat flux plate are typically used. These sensors are flexibly encapsulated and integrated into the contact pad surface of the moxibustion chair, directly contacting the user's skin or through a thin fabric. When heat flux (including far-infrared radiation heat and heat conduction / convection from moxa cake) passes vertically through the sensor, a thermoelectric potential (thermopile type) or voltage signal (heat flux plate type) proportional to the heat flux density is generated inside, thereby enabling real-time measurement. The measured value is the net heat flux density, which represents the net heat energy actually transferred to the user's skin tissue per unit time and per unit area. The real-time change in heat flux density can immediately reflect changes in the user's skin blood flow (increased blood flow will take away more heat, resulting in a decrease in the measured net heat flux) or sweating (evaporative heat dissipation leads to a decrease in net heat flux), providing the system with direct dynamic feedback on energy interaction.

[0021] In addition, all the sensing elements of the data acquisition unit are integrated into the same multi-layer composite fabric structure in a flexible and stretchable package form, forming the intelligent sensing pad of the moxibustion chair.

[0022] The data fusion and feature extraction unit receives user physiological characteristic data and various raw data from the physical energy transfer parameters acting on the user. After extracting the corresponding target features, it preprocesses the features of each single mode, including outlier removal and smoothing filtering, to ensure the stability of the single-mode features. The extracted target features include: The average value, gradient and non-uniformity statistics of body surface temperature distribution; the amount and rate of change of acupoint bioelectrical impedance relative to the user's baseline value; the time domain amplitude, frequency domain characteristics or trends of blood perfusion signal; the spatial distribution, instantaneous value and cumulative amount of net heat flux density; the entropy value, center of gravity position and time change of pressure distribution; and the activity level or specific events of skin conductance signal. Since the feature values ​​of different modalities vary greatly (for example, body surface temperature is in the range of 30-40℃, while the electrodermal signal is only at the micro Siemens level), all single-modal features need to be normalized or standardized. Normalization includes scaling to the [0,1] range using Min-Max, or standardizing to a mean of 0 and a variance of 1 using Z-score, so that all features are on the same numerical scale and large numerical features dominate subsequent calculations. All standardized single-modal features are directly concatenated in sequence to form a high-dimensional vector. For example, temperature features (mean value, gradient, non-uniformity), impedance features (change amount, rate of change), and blood flow features (time domain amplitude, frequency domain features, trend) are connected in sequence to finally integrate into a one-dimensional vector containing all modal information. If the dimension of the concatenated vector is too high, dimensionality reduction methods such as principal component analysis (PCA) and t-SNE can be used to reduce the vector dimension while retaining the core information, thereby reducing computational complexity and avoiding overfitting. The resulting high-dimensional vector is a multimodal fusion feature vector used to characterize the user's real-time tolerance state, therapeutic response level, and comfort. The calculated multimodal fusion feature vector is then transmitted to the main control module.

[0023] The two types of data collected by the multimodal sensing module directly cover the two forms of energy at the input end and the physiological feedback data of the user at the output end. The body surface temperature intuitively reflects the effect of the thermal field, which is used to prevent burns and assess the basic thermal sensation. The change of acupoint impedance is an objective electrophysiological indicator of the "deqi" and meridian response in traditional Chinese medicine, which provides data input for subsequent research on the depth of therapeutic effect. The net heat flux density accurately measures the energy actually delivered to the user, realizing precise control of the system output dose and ensuring that the user's safety status, comfort level and therapeutic response are simultaneously perceived.

[0024] The reused execution drive module includes a pharmacodynamic driving unit and a far-infrared driving unit. These two units receive control commands from the main control module and perform coordinated energy output to specific acupoints. The composition and functions of the two units include: The medicinal power driving unit includes a moxa cake that can be heated independently in sections and a heating base that matches the shape of the moxa cake. These sections are divided into multiple independent heating blocks (such as a 3x3 or more grid). Each block consists of an independent heating wire (or PTC element) and a temperature sensor (such as a thermocouple) forming a closed-loop temperature control circuit. Each heating block can be independently set and precisely maintained at a different target temperature. Combined with adjustable dampers or airflow control, it can achieve differentiated heating of different areas of the moxa cake, thereby controlling the spatial distribution of medicinal power release. For example, the moxa cake can be heated only in the center to concentrate the medicinal effect, or the edge areas can be heated alternately to extend its overall lifespan. The moxa cake is embedded with a high-temperature resistant readable storage chip, which stores a unique identification code that the main control module can read, as well as the interaction data between the moxa cake and the main control module during use. This allows the main control module to monitor the actual usage of the moxa cake. The moxa cake is made of moxa wool and the chip. The readable storage chip has traceability and anti-counterfeiting functions. When using it, the user must scan the code to identify the moxa cake as a genuine product from the manufacturer before placing it in the device. Only then can the user start the subsequent control process through the buttons on the moxibustion chair. Furthermore, the moxa cake is formed into a sheet shape through a compression process, unlike ordinary moxa wool on the market, which is easy to scatter. This product is easy to take out and put in without leaving any residue.

[0025] The far-infrared driving unit comprises an array of independently controllable far-infrared emitting units, consisting of dozens to hundreds of small far-infrared emitting sources (such as ceramic heating elements or carbon fiber heating elements of specific wavelengths) arranged in a grid pattern. Each unit has independent power control and driving circuitry, and each unit can be individually programmed to control its switching, emission power (intensity), operating frequency (pulse modulation), and emission angle. This allows the system to create complex and varied "thermal fields," such as achieving localized focused heating, scanning heating along meridian pathways, or gradient heat therapy with alternating strengths.

[0026] The two drive units are arranged vertically in physical space, with the far-infrared emitting array at the top, responsible for projecting a heat field onto the human body; and the moxa cake and its heating array at the bottom, responsible for generating a medicinal field, so as to respond to the control of the main control module and work together.

[0027] The main control module is communicatively connected to the multiplexed execution drive module and the multimodal perception module, and includes a drug release calculation unit, a physical release calculation unit and a collaborative control unit. The drug release calculation unit is used to read the information and status data of the current moxa cake obtained from the storage chip, and to calculate the remaining drug effect distribution map and the overall release rate of the moxa cake in real time. The implementation steps include the following: Step 1: Read the unique identification code of the corresponding moxa cake, and read the moxa cake usage data (including whether it has been used and the cumulative usage time), spatial temperature field matrix, and heating cavity environment tensor stored in the storage chip. The various types of data are collected through the following methods: The spatial temperature field matrix is ​​used to quantify the spatial temperature distribution of the moxa cake itself, including the real-time contact surface temperature of each independent section of the heating base that is in direct contact with the moxa cake. Each element in the matrix corresponds to the current temperature of the previous section of the moxa cake. By integrating a temperature sensor, such as a miniature patch thermocouple or a negative temperature coefficient thermistor, into each independent moxa cake heater section (such as a ceramic heating plate or PTC module), the temperature control circuit of each section reads the resistance or voltage signal of the built-in sensor in real time. After analog-to-digital conversion, the controller of the drug release unit summarizes the data and organizes these temperature readings into a two-dimensional array, namely the spatial temperature field matrix of the current moxa cake, according to a predetermined spatial grid order. This array is then sent to the drug release calculation unit via a communication bus (such as CAN or SPI). The heating chamber environment tensor is used to quantify the microenvironmental state inside the heating chamber containing the moxa cake. This tensor includes a volatile organic compound (VOC) concentration distribution matrix, an airflow velocity scalar, and a relative humidity scalar. The VOC concentration distribution matrix quantifies the spatial distribution of the relative concentration of characteristic volatiles of mugwort (such as eucalyptol and camphor) in the air directly above and adjacent to the volatile surface of the moxa cake. This data is collected by an array of metal oxide semiconductor gas sensors. Each sensor outputs a conductivity change signal related to the concentration. After processing by a conditioning circuit, the signal is mapped onto the corresponding moxa cake zone coordinates, forming a concentration distribution matrix, i.e., the VOC concentration distribution matrix. The airflow velocity scalar quantifies the average flow velocity of the air circulation within the heating chamber, and is used to quantify the overall intensity of medicinal gas diffusion and convective heat transfer. This is measured by installing a miniature hot-wire or impeller anemometer at the air inlet, outlet, or key flow channels of the chamber. The measuring sensor outputs an analog or digital signal proportional to the flow velocity, which is converted into a scalar value for reading, i.e., the gas velocity scalar. Since humidity affects the moisture absorption state and evaporation rate of the moxa wool, a relative humidity scalar is used to measure the relative humidity of the air inside the heating chamber. A capacitive or resistive humidity sensor is placed inside the chamber, outputting an electrical signal corresponding to the relative humidity. After calibration, a percentage scalar value is obtained. During the data acquisition process, all sensor data is synchronously sampled and packaged by a central data acquisition unit to ensure that the temperature matrix, concentration matrix, flow rate, and humidity values ​​all have a unified timestamp, thus forming a heating chamber environment tensor that is temporally aligned and spatially correlated.

[0028] Step 2: Construct a pharmacokinetic release model based on the extended Arrhenius equation. The pharmacokinetic release model treats the artemisia cake as a heterogeneous reactant for calculation. The implementation process includes: Step 21: If Step 1 identifies it as a new Ai Bing (a type of herbal medicine), initialize a concentration of the remaining effective ingredients in all partitions; If it is an Ai Bing (a type of medicinal cake) in use, load it from the state saved in the previous cycle; Step 22: Calculate the instantaneous release rate of each partition. For each partition (i,j) in the moxa cake, calculate its current instantaneous release rate using the kinetic equations of the pharmacokinetics model. The calculation formula is as follows: ; In the formula, Let be the instantaneous rate of drug release of partition (i,j) of the mugwort cake at time t, which is the amount of effective drug released by this partition per unit time, in mol / s; is the ideal gas constant, with units of J / (mol·K); Let be the real-time temperature of partition (i,j) of the worm at time t, in K (Kelvin). , For the intrinsic release parameters of the agaric cake, among which, The pre-exponential factor represents the maximum theoretical rate of volatilization of the effective components of the moxa cake under combustion. It is a constant value with the unit mol / s. The larger the value, the greater the potential of the moxa cake to release medicinal power under ideal conditions. The apparent activation energy of the mugwort cake is expressed in J / mol. The higher the value, the more sensitive the release of medicinal power is to temperature. The environmental correction function, based on dimensionless correction terms for real-time volatile organic compound concentration, airflow velocity, and humidity, is used to mathematically model the coupled effects of three key environmental factors (concentration, airflow, and humidity) on pharmacological release. A calculated value greater than 1 indicates that the environmental factors (strong airflow, low concentration, and low humidity) promote release, less than 1 indicates that release is inhibited, and a value of 1 represents the standard state, which does not affect release. In the formula, The volatile organic compound concentration distribution matrix (scalar representation) of the Ai Bing region at time t in (i,j). Let be the airflow velocity scalar at time t. Let be the relative humidity scalar at time t. For partition (i,j) (grid coordinates) at the previous computation time... The real-time relative remaining active ingredient concentration is the ratio of the current residual active ingredient concentration in partition (i,j) to the highest concentration (which can be set to 1). It reflects the relative remaining amount of active ingredient, which decreases to near 0 over time. The calculation and control cycle for the preset time step.

[0029] It should be noted that, , The determination process includes: For each batch of Artemisia argyi, its stable release rate at multiple temperature points was measured using standard experiments (such as thermogravimetric analysis-mass spectrometry or a self-made isothermal release test bench). At each temperature point, the change in drug release rate R (by collecting and analyzing the mass of volatiles released over a certain period of time) over time is accurately measured and recorded, and the average value of the stable phase is taken, while the corresponding absolute temperature T is recorded simultaneously. Multiple sets of (R,T) values ​​were obtained from multiple temperature point tests. The linearized form of the Arrhenius equation was then fitted using the following formula: ,by The x-axis is... Using the ordinate as the vertical axis, the experimental data points are plotted on a coordinate system. A linear regression algorithm is then applied to these data points to obtain a fitted line. The line intercept of the fitted line is then calculated. Calculate the slope of the fitted line to obtain ; The fitted result of this batch of mugwort cakes , The values ​​are used as intrinsic release parameters for that batch. On the production line, these two parameter values ​​(or batch codes used for retrieval) are written into the storage chip embedded in each moxa cake in that batch. When the moxibustion chair system reads the identification code corresponding to a moxa cake for the first time, it will automatically download the corresponding batch code from the local database or the cloud. , Values.

[0030] Environment correction function The calculation formula is: ; In the formula, This is a concentration inhibition term used to quantify the local drug vapor saturation inhibition effect. This is a reference concentration calibration value, representing the threshold at which the concentration of the medicinal gas begins to have a significant impact on volatilization under this environment. It can be determined experimentally. It is the concentration inhibition coefficient, with a value of 0-1; This is the airflow enhancement term, used to quantify the forced convection-enhanced mass transfer effect, where... It refers to the airflow velocity calibration value, such as the typical velocity under natural convection. This is the gas enhancement factor, which is taken as 0-1; The humidity inhibition term is used to quantify the hindering effect of humidity competition on adsorption and diffusion. This is the humidity inhibition coefficient. In high humidity environments, water molecules compete with volatile molecules in mugwort for adsorption on the porous surface of the mugwort floss, occupying some active sites. Furthermore, water vapor fills the pores and cavities of the mugwort floss, potentially hindering the diffusion of volatile molecules. Therefore, the higher the humidity, the smaller this value (exponential decay). This indicates that at the same temperature, the actual evaporation rate is inhibited by humidity. The exponential form reflects that this inhibition may increase non-linearly with increasing humidity. Step 23: Construct a state update equation using discrete difference equations, and calculate the current concentration of the remaining effective ingredients in the mugwort cake based on the state update equation. The calculation formula is as follows: ; In the formula, For partition (i,j) (grid coordinates) at computation time The real-time remaining effective component concentration is a state vector for that moment and that partition, and the pixel values ​​are used to construct the subsequent map. The initial total mass of effective ingredients per unit area of ​​the mugwort cake; Arranging all partitions (i,j) of the moxa cake at time t according to their spatial location yields the real-time remaining efficacy distribution map at time t. This map visually displays the efficacy magnitude in the form of a heatmap. Here, the efficacy release kinetic model transforms the abstract efficacy consumption of the moxa cake into specific matrix numerical update operations, enabling the system to accurately track the remaining efficacy of every corner of the moxa cake, much like managing the inventory of a dynamic warehouse.

[0031] Step 24: Continuously compare the gas concentration predicted by the drug release kinetic model with the concentration measured by the actual sensor. If a systematic prediction deviation is found, the module will safely start the parameter fine-tuning program in the background to slightly optimize several key coefficients inside the model through the algorithm.

[0032] The drug release calculation unit, through a smart model (drug release kinetic model) that integrates the designed mechanism and data, provides real-time, quantitative, and spatial insight into the internal efficacy state of moxa cakes, a traditional Chinese medicine consumable. This enables the entire moxibustion system to shift from direct heating to quasi-energy management, flexibly adjusting the energy output strategy based on the real-time remaining efficacy of the moxa cake. This significantly improves the efficiency and safety of moxa cake usage while ensuring the stability of therapeutic effects, fundamentally solving the problems of blind control and uncertainty in therapeutic effects caused by the inability to perceive the state of the moxa cake.

[0033] The physical release calculation unit is used to establish the energy-physiological response mapping relationship of the current user based on the user's physiological characteristic data and the physical energy transfer parameters acting on the user, and to obtain the predicted physiological state of the user at the next control moment corresponding to the current energy. The implementation process includes the following: An energy action vector is constructed based on time-series data collected by the multimodal sensing module and physical energy transfer parameters that primarily affect the user. Encode the energy action vector as a feature, for example, calculate the net energy input rate corresponding to each body surface zone, for the contribution from the upper far-infrared radiation and the lower moxa cone conduction / radiation; Receive multiple target features corresponding to the multimodal fusion vector, calculate the cumulative physiological changes corresponding to each target feature from the start of the session to the present, such as the total decrease in impedance, and for a specific acupoint, calculate the energy features of the energy input unit directly above / below it, and establish energy-acupoint pairing data pairs; An energy-physiological response mapping relationship is constructed and updated in real time through an online adaptive system identification process. This mapping relationship allows for the calculation of the current equivalent energy input acting on each physiological monitoring point based on the currently collected energy. The construction and updating process includes: First, a hybrid model structure is constructed, employing a set of parallel recursive least squares (RLS) adaptive filters. Each filter corresponds to a key physiological feature data point, and a lightweight neural network (such as a multilayer perceptron MLP) is used to learn the nonlinear part of the RLS filter residuals and the influence of user static features. Based on the data collected by the multimodal acquisition module, the physical energy transfer parameters at the current moment are used to predict the user's predicted physiological state in the next control cycle according to the energy-physiological response mapping relationship. The predicted physiological state is a structured vector containing the predicted trajectory of each key physiological indicator in the short time domain in the future.

[0034] The collaborative control unit is used to acquire the remaining pharmacodynamic distribution map, the predicted physiological response state of the user corresponding to the energy-physiological response mapping relationship, and the current conditioning target instruction. It inputs the acquired data into a pre-trained deep policy network to output a basic control policy, and based on the basic control policy, generates and issues spatiotemporal collaborative control instructions for regulating each independent unit in the pharmacodynamic driving unit and the far-infrared driving unit. The collaborative control unit includes a data receiving and processing unit, a policy network construction subunit, a policy network inference subunit, a compliance verification subunit, and a metadata storage unit, all connected in sequence. The implementation process of each unit includes: The data receiving and processing subunit is used to receive the remaining efficacy distribution map of the current moxa cake calculated by the efficacy release calculation unit, and to perform dimensionality reduction and feature extraction on each pixel in the remaining efficacy distribution map, such as calculating spatial statistics (mean, standard deviation σ), spatial moments (centroid coordinates), and deep spatial feature vectors extracted by the convolutional neural network (CNN) encoder. The predicted physiological response state for the next control cycle is received from the physical release calculation unit. This can be achieved by extracting statistical features (mean, slope, peak) of the predicted trajectory or by using a recurrent neural network (RNN) encoder to process the time-series data and encode the received physiological response state into a fixed-dimensional state vector. The system receives the current treatment target instruction, such as "Warm Yang and Unblock Meridians - Governing Vessel (Intensity: Medium, Duration: 20min)," parses and encodes it into a multi-dimensional target vector. The multi-dimensional target vector includes the target meridian code, expected effect intensity, total duration, and preference pattern (such as emphasizing heat over medicine). The three feature vectors obtained from the above encoding are concatenated and fused to obtain the joint state representation vector.

[0035] The policy network construction subunit is used to build a deep policy network based on offline training, mapping the joint state representation vector to the probability distribution of basic control actions. The deep policy network is implemented in the form of an encoder-decoder using an Actor-Critic architecture. The implementation process includes: Encoders are constructed using multilayer perceptrons (MLPs) or transformers that incorporate attention mechanisms to gain a deep understanding of the complex relationships that jointly represent states. The strategy head Actor is used to output the basic control strategy. For continuous actions (such as power and temperature), it outputs the mean and diagonal covariance matrix of a Gaussian distribution. For discrete actions (such as operating mode selection), it outputs the probability of a multinomial distribution. The value head Critic is used to output the state value estimate.

[0036] The policy network inference subunit is used during runtime (deployment phase) to input the current joint state representation vector into the pre-trained deep policy network to output the basic control policy. The network forwards, outputting the parameters of the basic control actions through the policy head. For deterministic policies, the action values ​​corresponding to the basic control policy are directly output; for stochastic policies, the basic control policy is sampled from a parameterized distribution. Since the basic control strategy is an abstract and high-level vector, it needs to be refined into a sequence of specific control instructions with spatiotemporal information that can directly drive the independent units of the underlying actuators. The instruction refinement operation includes: Based on the spatial distribution pattern encoding in the basic control strategy (such as "linear gradient along the bladder meridian"), combined with the preset mapping relationship between "acupoint-drive", the independent set values ​​of each far-infrared emitting unit and each moxa cake heating zone are calculated. The independent set values ​​include power and temperature target values. Based on the coordination mode and phase information in the basic control strategy, spatiotemporal coordinated control instructions for the time series of the next few control cycles are generated for each independent unit, clarifying when and with what parameters each unit will work, and ensuring the precise coordination of the thermal field and the pharmacy field on the time axis. The compliance verification subunit is used to map the energy parameters corresponding to the generated spatiotemporal coordinated control commands through a mapping relationship. Simulation prediction is performed to obtain the user's physiological state data over several control cycles. The prediction results are compared with the safety threshold (burn prevention) and the effectiveness baseline (minimum stimulation intensity). If all prediction results are compliant, the instruction is approved. If any one of them is non-compliant, the instruction correction subroutine is triggered. For minor violations (such as local temperatures approaching but not exceeding a safety threshold), the correction subroutine may slightly reduce the power of the local unit. For serious violations or insufficient performance, it may request the cooperative control unit to reprogram or switch to a predefined safety fallback strategy. The specific spatiotemporal coordinated control instructions, after verification (or correction), are encapsulated into data packets with timestamps and unit addresses, and synchronously sent to the medicated force drive unit (artemisia cake heater array controller) and the far-infrared drive unit (emission array controller) via a real-time communication bus (such as CAN FD or real-time Ethernet).

[0037] Metadata storage units are used for: Record the joint state representation vector, the output basic control strategy, the refined spatiotemporal collaborative control instructions, and the actual physiological response triggered in each decision cycle. These data form a state-action-result triplet and are stored in the system log. During system idle or maintenance periods, newly accumulated state-action-result data can be used to fine-tune the pre-trained deep policy network online. By employing offline reinforcement learning or imitation learning techniques, the deep policy network can slowly adapt to the common feedback of all users or the long-term preferences of specific users, thus achieving continuous system evolution.

[0038] To address the contradiction between rigid control logic and lack of personalized energy output coordination in traditional moxibustion chair control systems, which cannot perceive the efficacy of moxa cakes and the user's physiological state and can only run preset programs, resulting in a simple superposition of the heat field and the medicine field with misaligned rhythms, the collaborative control unit integrates a deep strategy network to fuse the supply-side efficacy spectrum, the demand-side predicted physiological state, and the conditioning goals in real time. This dynamically generates non-preset, optimal collaborative control strategies, enabling precise coupling of heat energy and medicine energy release in time and space. For example, it can achieve the goal of first heating the meridians and then guiding the medicine deeper, or dynamically compensate for heat radiation based on the remaining medicinal power, thereby constructing a true synergistic effect pathway between the heat field and the medicine field.

[0039] Please see Figure 3 This embodiment provides a method for regulating the dual-moxibustion and far-infrared composite energy system of a moxibustion chair. The regulation method includes the following steps: S1: Real-time acquisition of user physiological characteristic data and physical energy transfer parameters acting on the user; after extracting and fusing the two sets of data, a multimodal fusion feature vector is generated. S2: Read the current moxa cake information, the spatial temperature field matrix of the moxa cake itself, and the heating cavity environment tensor. Based on the pre-trained drug release kinetic model, calculate and output the remaining drug effect distribution map and overall release rate of the current moxa cake. S3: Based on the multimodal fusion feature vector, establish and update the energy-physiological response mapping relationship of the current user, and output the predicted physiological response state for the next control cycle; S4: Receive the remaining drug efficacy distribution map, predicted physiological response state and current conditioning target instruction, fuse the received data into a unified joint state representation vector, input it into a pre-trained deep policy network, and output the basic control policy. S5: Based on the basic control strategy, refine the instructions and verify safety and compliance, and generate spatiotemporal coordinated control instructions for independently regulating each execution unit in the pharmacokinetics driving unit and the far-infrared driving unit. Spatiotemporal coordinated control commands are sent to the drug-driven unit and the far-infrared driven unit to drive them to perform coordinated energy output. During the energy output process, data perception and state estimation steps are continuously executed to form a closed-loop feedback. The energy-physiological response mapping relationship is dynamically updated based on real-time feedback data. The intelligent coordinated decision-making and command execution steps are executed cyclically until the moxibustion process ends.

[0040] This invention presents a dual-moxibustion and far-infrared composite energy system and control method for a moxibustion chair. It constructs a closed-loop adaptive control system encompassing supply-side state perception, demand-side dynamic modeling, and intelligent collaborative decision-making. Compared to control processes relying on heating combined with control mode switching or fixed programs, this invention introduces a pharmacokinetic release kinetic model and personalized physiological response modeling. Combined with a spatiotemporal collaborative mapping engine based on deep reinforcement learning, it achieves precise coordination and dynamic control of the heat field and pharmacokinetic field in the temporal and spatial dimensions during moxibustion. Specifically, by assigning a digital identity to the moxa cake and establishing a microscopic release model, the system can perceive and quantify the remaining pharmacokinetic distribution of the moxa cake in real time. Through multimodal physiological perception and online learning, a unique energy-physiological response relationship is constructed for each user. Finally, through a deep policy network, the system integrates the above multi-source information to generate non-preset, optimized collaborative energy output commands.

[0041] This invention achieves precise synergy and dynamic compensation of thermal and medicinal energy in time and space, fully leveraging the combined energy synergy of the thermal and medicinal fields. This significantly enhances the warming and unblocking of meridians and the depth and efficiency of medicinal penetration in moxibustion. Furthermore, it adaptively adjusts based on each user's real-time physiological feedback, ensuring individualized optimal treatment dosage and stimulation intensity, greatly improving user comfort and acceptance. Real-time safety monitoring and predictive compliance verification based on heat flux density, moxa cake temperature, and user body surface temperature proactively prevent risks such as burns. Simultaneously, precise management of the moxa cake's state avoids ineffective or excessive heating, improving the economy and safety of the equipment.

[0042] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A dual-moxibustion and far-infrared composite energy system for a moxibustion chair, characterized in that: The system includes: The multimodal perception module is used to collect user physiological characteristic data and physical energy transfer parameters acting on the user. After data extraction and feature fusion of the collected data, a multimodal fusion feature vector is generated. The reused execution drive module includes a drug-driven unit and a far-infrared driven unit. The drug-driven unit and the far-infrared driven unit are used to receive control commands from the main control module and perform coordinated energy output to specific acupoints. The medicinal power driving unit includes moxa cakes that can be heated independently in separate zones, and each moxa cake has a readable unique identification code. The main control module is communicatively connected to the multiplexed execution drive module and the multimodal perception module, and the main control module includes a drug release calculation unit, a physical release calculation unit and a collaborative control unit; The drug release calculation unit is used to acquire information about the current moxa cake and its current state data, and to calculate the remaining drug effect distribution map and the overall release rate of the moxa cake based on the pre-trained drug release calculation model. The state data includes the spatial temperature field matrix of the moxa cake itself and the environmental tensor of the heating cavity. The physical release calculation unit is used to establish the energy-physiological response mapping relationship of the current user based on the user's physiological characteristic data and the physical energy transfer parameters acting on the user, and to predict the predicted physiological response state of the next control cycle. The collaborative control unit is used to acquire the remaining drug efficacy distribution map, the user's predicted physiological response state corresponding to the energy-physiological response mapping relationship, and the current conditioning target instruction. The acquired data is input into a pre-trained deep policy network to output a basic control strategy. Based on the basic control strategy, spatiotemporal collaborative control instructions for regulating each independent unit in the drug efficacy driving unit and the far-infrared driving unit are generated and issued.

2. The moxa chair double moxa and far infrared composite energy system according to claim 1, characterized in that: In the multimodal sensing module, the user physiological characteristic data includes gridded temperature data reflecting the local body surface temperature distribution, acupoint bioelectrical impedance data reflecting changes in the state of meridian tissues, and blood perfusion data. The physical energy transfer parameters acting on the user include net heat flux density data and pressure distribution data.

3. The moxa chair double moxa and far infrared composite energy system according to claim 1, characterized in that: In the drug release unit, the environmental tensor of the heating chamber includes a volatile organic compound concentration distribution matrix, an airflow velocity scalar, and a relative humidity scalar, wherein: The volatile organic compound concentration distribution matrix is ​​used to quantify the spatial distribution of the relative concentration of characteristic volatile compounds of mugwort in the air directly above the mugwort cake and close to its volatile surface; The airflow velocity scalar is used to quantify the average flow rate of air circulating inside the mugwort cake heating chamber. The relative humidity scalar is used to quantify the average flow rate of air circulation within the heating chamber of the mugwort cake.

4. The moxa chair double moxa and far infrared composite energy system according to claim 3, characterized in that: The drug release unit includes calculating the remaining drug efficacy distribution map and the overall release rate of the moxa cake based on a pre-trained drug release calculation model, including the following steps: For each section of the moxa cake, the current instantaneous drug release rate is calculated by applying the improved kinetic equation of the drug release kinetic model; A state update equation is constructed using discrete difference equations, and the remaining effective ingredient concentration of each partition of the Ai Bing is calculated based on the state update equation. Based on the calculated concentration of the remaining active ingredient in each partition, the partitions are arranged according to their spatial location to obtain a heat map of the remaining efficacy distribution.

5. The moxa chair double moxa and far infrared composite energy system according to claim 1, characterized in that: In the drug release unit, the formula for calculating the current instantaneous drug release rate is as follows: ； In the formula, is the instantaneous drug release rate of the partition (i, j) of the mugwort cake at time t, is the ideal gas constant, is the real-time temperature of the partition (i, j) of the mugwort cake at time t, and are the intrinsic release parameters of the mugwort cake, is the pre-factor, is the apparent activation energy of the mugwort cake, is the environmental correction function, is the volatile organic compound concentration distribution matrix of the partition (i, j) of the mugwort cake at time t, is the scalar of the air flow velocity at time t, is the scalar of the relative humidity at time t, is the partition (i, j) at the previous calculation time of the real-time relative remaining effective ingredient concentration.

6. The dual-moxibustion and far-infrared composite energy system of a moxibustion chair according to claim 5, characterized in that: The formula for calculating the environmental correction function is as follows: ； In the formula, This is a concentration inhibition term. For reference concentration calibration value, It is the concentration inhibition coefficient. This is an airflow enhancement term. It is the gas enhancement coefficient. This is a humidity suppression term. This is the humidity suppression coefficient.

7. The dual-moxibustion and far-infrared composite energy system of a moxibustion chair according to claim 5, characterized in that: The process of determining the intrinsic release parameters of the mugwort cake includes: For each batch of mugwort cakes, the stable release rate of the mugwort cakes at multiple temperature points was measured through standard experiments. At each temperature point, the change in drug release rate R over time was precisely measured and recorded. The average value of R during the steady-state phase was taken, and the corresponding absolute temperature T was recorded simultaneously. Obtain multiple sets of (R,T) and fit the linearized form of the Arrhenius equation. The fitting formula is as follows: ； by The x-axis is... Using the vertical axis as the ordinate, the experimental data points are plotted on a two-dimensional coordinate system, and the fitted line is obtained through linear regression. The intercept of the fitted line is obtained by calculating... Calculate the slope of the fitted line to obtain .

8. The dual-moxibustion and far-infrared composite energy system of a moxibustion chair according to claim 1, characterized in that: In the collaborative control unit, the deep policy network adopts an encoder-decoder architecture based on an Actor-Critic framework, specifically including: Encoders are constructed using multilayer perceptrons or transformers that incorporate attention mechanisms to gain a deep understanding of the complex relationships that jointly represent states. The strategy header (Actor) is used to output the basic control strategy; The value head Critic is used to output the state value estimate.

9. A method for regulating the dual-moxibustion and far-infrared composite energy system of a moxibustion chair, applicable to the dual-moxibustion and far-infrared composite energy system of a moxibustion chair as described in any one of claims 1-8, characterized in that: The method includes the following steps: S1: Real-time acquisition of user physiological characteristic data and physical energy transfer parameters acting on the user; data extraction and feature fusion of the two acquired data to generate a multimodal fusion feature vector; S2: Read the current moxa cake information, the spatial temperature field matrix of the moxa cake itself, and the heating cavity environment tensor. Based on the pre-trained drug release kinetic model, calculate and output the remaining drug effect distribution map and overall release rate of the current moxa cake. S3: Based on the multimodal fusion feature vector, establish and update the energy-physiological response mapping relationship of the current user, and output the predicted physiological response state for the next control cycle; S4: Receive the remaining drug efficacy distribution map, predicted physiological response state and current conditioning target instruction, fuse the received data into a unified joint state representation vector, input it into a pre-trained deep policy network, and output the basic control policy. S5: Based on the basic control strategy, refine the instructions and verify safety and compliance, and generate spatiotemporal coordinated control instructions for independently regulating each execution unit in the drug-driven unit and the far-infrared driven unit.

Images