Moxibustion incubator control method and system based on multi-sensor data fusion

Abstract

The present application relates to the technical field of temperature control, and more particularly to a moxibustion incubator control method and system based on multi-sensor data fusion. The method comprises: collecting real-time temperature data of multiple regions through a sensor array, constructing an initial temperature field mapping through preprocessing and interpolation, and dividing unevenly heat-conducting partitions; extracting temperature gradient values from the partitions to identify abnormal regions, generating error quantization indicators and their spatio-temporal characteristics; inputting the characteristics into a pre-trained model to predict material aging trends, obtaining an aging influence coefficient, constructing an optimization objective function in combination with current temperature data, and solving to generate an adjusted heating strategy; updating the temperature field distribution using the strategy, calculating uniformity indicators and iteratively optimizing the strategy accordingly; and finally continuously monitoring the operating state through the optimized strategy to generate a stable temperature field record. The present application solves the problem of temperature field instability caused by heat conduction differences and material aging in the incubator, and realizes long-term uniform and stable control and adaptive compensation of the temperature field.

Description

Technical Field

[0001] This invention relates to the field of temperature control technology, and in particular to a control method and system for a moxibustion warming box based on multi-sensor data fusion. Background Technology

[0002] Moxibustion therapy utilizes the heat radiation generated by burning moxa sticks to act on acupoints on the human body. Its therapeutic effect is closely related to the stability and uniformity of the moxibustion temperature. Moxibustion warming boxes achieve their thermotherapy function by integrating electric heating elements within a closed or semi-closed enclosure. However, in practical applications, uneven temperature distribution is a prominent problem inside these boxes. The root cause lies in the fact that the interior of the box is composed of various materials; for example, the metal mesh at the bottom conducts heat rapidly, while the insulation cotton on the side walls conducts heat slowly. The inherent differences in thermal conductivity across different areas lead to a significant temperature gradient between the central and peripheral areas near the heat source, directly affecting the consistency of the user's moxibustion sensation and the therapeutic effect.

[0003] Traditional control methods typically rely on feedback from a few discrete temperature sensors for constant temperature regulation, which has significant technical limitations. First, the limited number of sensor nodes cannot accurately construct a continuous temperature field distribution across the entire treatment space, making it difficult to identify areas of uneven thermal conductivity caused by structural and material differences. Second, with long-term operation, insulation and thermal conductive materials inevitably age, increasing thermal resistance and decreasing thermal conductivity, leading to a slow drift in temperature field distribution characteristics. Traditional PID control or fixed threshold control methods lack the ability to perceive and model material aging trends, resulting in a continuous deterioration in control accuracy over time. A deeper problem lies in the fact that existing temperature controls generally lack a closed-loop mechanism that seamlessly integrates temperature field anomaly diagnosis, material aging prediction, and heating strategy optimization. This makes it impossible to dynamically compensate for the dual effects of structural differences and material aging, resulting in difficulties in maintaining temperature field uniformity over the long term.

[0004] To address the aforementioned issues, there is an urgent need for a closed-loop control scheme that can integrate multi-sensor data to achieve high-precision temperature field reconstruction, intelligent anomaly diagnosis, material aging prediction, and adaptive optimization of heating strategies. This would fundamentally solve the problem of temperature field instability caused by differences in thermal conductivity and material aging in moxibustion heat boxes, ensuring the stability and safety of therapeutic effects during long-term operation. Summary of the Invention

[0005] This invention provides a method and system for controlling a moxibustion chamber based on multi-sensor data fusion. It is used to construct a temperature field mapping, identify areas of thermal conductivity difference, quantify temperature anomalies, and predict material aging trends by fusing multi-sensor data, thereby dynamically optimizing the heating strategy and achieving long-term uniform and stable control of the temperature field inside the chamber.

[0006] In a first aspect, the present invention provides a method for controlling a moxibustion chamber based on multi-sensor data fusion, the method comprising: S1. Real-time temperature data of multiple areas inside the moxibustion heat box are collected through a sensor array. After preprocessing the real-time temperature data, an initial temperature field mapping is constructed. Based on the initial temperature field mapping, regions with different thermal conductivity are divided to determine the uneven thermal conductivity zones. S2. Extract temperature gradient values ​​from the non-uniform thermal conductivity partition, determine the temperature anomaly region, and generate error quantification index and corresponding spatial distribution characteristics and temporal evolution characteristics. S3. Input the spatial distribution characteristics and temporal evolution characteristics of the error quantification index into the pre-trained prediction model to predict the material aging trend and determine the aging influence coefficient; construct an optimization objective function based on the aging influence coefficient and the current temperature data, and solve for the adjusted heating strategy. S4. Apply the adjusted heating strategy, collect the temperature data after execution through the sensor array, update the temperature field distribution data, calculate the uniformity index of the updated temperature field distribution data, and iteratively optimize the heating strategy based on the uniformity index. S5. Continuously monitor the operation status of the moxibustion chamber through the optimized heating strategy and generate a stable temperature field record.

[0007] Secondly, the present invention also provides a moxibustion chamber control system based on multi-sensor data fusion, for implementing the above-mentioned method, the system comprising: The temperature field partitioning unit is used to collect real-time temperature data of multiple areas inside the moxibustion heat box through a sensor array, preprocess the real-time temperature data to construct an initial temperature field mapping, and divide the thermal conductivity difference area according to the initial temperature field mapping to determine the thermal conductivity non-uniform partition. The error feature quantization unit is used to extract temperature gradient values ​​from the thermally uneven partition, determine the temperature abnormal area, and generate error quantization index and corresponding spatial distribution characteristics and temporal evolution characteristics. The heating strategy solving unit is used to input the spatial distribution characteristics and temporal evolution characteristics of the error quantification index into a pre-trained prediction model to predict the material aging trend and determine the aging influence coefficient; and to construct an optimization objective function based on the aging influence coefficient and the current temperature data to solve for the adjusted heating strategy. The strategy evaluation and iteration unit is used to apply the adjusted heating strategy, collect temperature data after execution through the sensor array, update the temperature field distribution data, calculate the uniformity index of the updated temperature field distribution data, and iteratively optimize the heating strategy based on the uniformity index. The temperature field monitoring unit is used to continuously monitor the operating status of the moxibustion chamber through an optimized heating strategy and generate a stable temperature field record.

[0008] Thirdly, the present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the above-described method.

[0009] The beneficial effects of this invention are as follows: This invention constructs an initial temperature field mapping through multi-point acquisition by a sensor array and Kriging interpolation, fusing discrete sensor data into a continuous three-dimensional temperature distribution map, providing a high-precision spatial perception foundation for subsequent analysis. Based on this, a clustering algorithm is used to actively divide the thermally non-uniform partitions, spatially locating the structural thermal conductivity differences inside the temperature chamber caused by variations in materials such as the metal insulation mesh and thermal insulation cotton. This elevates the system from sensing temperature levels to understanding differences in thermal conductivity characteristics. Subsequently, temperature gradient values ​​are extracted from the thermally non-uniform partitions to generate error quantification indicators, simultaneously extracting their spatial distribution characteristics and temporal evolution characteristics. This transforms ambiguous temperature field anomalies into precise spatiotemporal feature vectors, providing structured input for the quantitative prediction of material aging. A multilayer perceptron neural network utilizes these spatiotemporal features to predict the thermal conductivity decay rate and thermal resistance increase, generating an aging influence coefficient after weighting and environmental calibration. This coefficient quantifies the previously invisible material aging state. After obtaining the aging impact coefficient, it is embedded as a weighting factor into an optimization objective function with the dual goals of minimizing temperature deviation and energy consumption. The optimal heating power and time parameters are solved using a particle swarm optimization algorithm, generating an adjusted heating strategy. This allows severely aging areas to receive higher compensation priority during the optimization process, achieving adaptive adjustment of the control strategy to material aging. Before application, the strategy undergoes safety verification through finite element heat conduction simulation to avoid the risk of localized overheating. After updating the temperature field distribution using the adjusted strategy, a uniformity index is calculated, triggering parameter iteration of the particle swarm algorithm. Finally, a stable temperature field record containing timestamps and temperature field data is generated through continuous cyclic monitoring. Through the synergy of these technical solutions, a leap from static temperature control to dynamic adaptive temperature control is achieved, ensuring that the moxibustion chamber maintains a uniform and stable temperature field throughout its entire lifespan, fundamentally guaranteeing the consistency of moxibustion efficacy and the safety of equipment operation. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments 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.

[0011] Figure 1 This is a flowchart of the moxibustion chamber control method based on multi-sensor data fusion in the embodiment; Figure 2This is a schematic diagram illustrating the trend of uniformity index changing with the number of iterations in the example; Figure 3 This is a structural diagram of the moxibustion warming box control system based on multi-sensor data fusion in the embodiment; Detailed Implementation This invention provides a method and system for controlling an moxibustion chamber based on multi-sensor data fusion. The terms "first," "second," "third," "fourth," etc. (if present)," in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0012] For ease of understanding, the specific process of the embodiments of the present invention will be described below, such as... Figure 1 As shown, the moxibustion chamber control method based on multi-sensor data fusion in this embodiment of the invention includes: S1. Real-time temperature data of multiple areas inside the moxibustion heat box are collected through a sensor array. After preprocessing the real-time temperature data, an initial temperature field mapping is constructed. Based on the initial temperature field mapping, regions with different thermal conductivity are divided to determine the uneven thermal conductivity zones. In S1, the initial temperature field mapping is constructed, including: Inside the moxibustion chamber, temperature sensing nodes are arranged at a preset grid spacing to form a sensor array. The temperature values ​​of each temperature sensing node are read synchronously according to a preset sampling frequency. The spatial coordinates of each temperature sensing node are bound to the corresponding temperature value to generate real-time temperature data. The real-time temperature data was filtered in the time domain using the Kalman filter algorithm to remove missing data and obtain a clean temperature dataset. The spatial coordinates and temperature data of each temperature sensing node in the clean temperature dataset are extracted. The Kriging interpolation algorithm is used to perform spatial continuity calculation. A spherical semivariogram function is fitted to determine the spatial variation parameters. A three-dimensional temperature distribution map is constructed to generate an initial temperature field mapping.

[0013] Specifically, due to the influence of the burning state of the moxa stick and the differences in the thermal conductivity of various materials in the box, the internal temperature field of the moxibustion box exhibits significant non-uniformity and time-varying characteristics. In order to accurately identify the areas of thermal conductivity difference, it is necessary to first establish an initial temperature field mapping that can finely characterize the spatial distribution of temperature inside the box.

[0014] Multiple temperature sensing nodes are arranged inside the moxibustion chamber according to a preset grid spacing. For example, thermocouple sensing nodes are arranged at equal intervals on the inner wall of the chamber and on the support frame, so that each node covers the main space inside the chamber, thus forming a sensor array. This sensor array is a distributed temperature measurement network composed of multiple temperature sensing nodes arranged according to a preset spatial topology. The temperature values ​​output by each temperature sensing node are read synchronously according to a preset sampling frequency, and the spatial coordinates of each node are bound and associated with the temperature value at the corresponding time to generate real-time temperature data containing spatial location information and temperature measurement value.

[0015] After obtaining the real-time temperature data, the Kalman filter algorithm is used to perform time-domain filtering on the data stream of each temperature sensing node. The Kalman filter algorithm is a recursive optimal estimation algorithm based on a state-space model. It uses a prediction step to deduce a prior estimate of the current state based on the optimal estimate of the previous time step, and then uses an update step to introduce the measured temperature data of the current time step to correct the prior estimate. Finally, it outputs the optimal estimate of the temperature at the current time, thereby effectively filtering out high-frequency random noise introduced by electromagnetic interference and internal airflow disturbances during the heating process. The optimal estimates obtained by each node after Kalman filtering are summarized, and missing data caused by hardware communication delays are removed to obtain a clean temperature dataset.

[0016] Subsequently, the spatial coordinates of each temperature sensor node and its corresponding optimal estimate were extracted from the clean temperature dataset to construct a discrete spatial temperature point set. The Kriging interpolation algorithm was used to perform spatial continuity calculations on this discrete point set. Kriging interpolation is an unbiased optimal estimation method based on spatial autocorrelation. It first calculates the spatial distance between any two nodes in the discrete spatial temperature point set and the semivariance of their optimal temperature estimates, generating an experimental semivariance scatter plot. The semivariance is used to measure the degree of difference in temperature values ​​between two locations separated by a certain distance. Then, a spherical semivariance function was used to fit the experimental semivariance scatter plot to determine three spatial variability parameters: nugget constant, sill value, and range. The nugget constant reflects the random variance caused by measurement error or microvariation, the sill value represents the maximum variance reached by spatial variability, and the range represents the maximum distance at which the temperature data still has spatial correlation. Based on the fitted spherical semivariance function, a spatial covariance matrix was constructed. For any unplaced temperature sensor node inside the chamber... The spatial coordinates of the sensing nodes to be estimated are determined, and the spatial covariance vector between these coordinates and all known temperature sensing nodes is calculated. The Kriging weight coefficient equations are solved using the weight coefficients of each known node to the unknown point as variables to be solved. The equations require that the sum of all weight coefficients equals 1 to satisfy the unbiasedness constraint, and that the sum of the products of each weight coefficient and the covariance between known nodes equals the covariance between the unknown point and the corresponding known node. The weight values ​​of each known temperature sensing node to the unknown coordinates are obtained by simultaneously solving these linear equations. The optimal temperature estimate of each known temperature sensing node is weighted and summed with its corresponding weight value to obtain the continuous temperature prediction value of the spatial coordinates to be estimated. All spatial coordinates inside the moxibustion chamber are traversed according to a preset spatial resolution, and the above weighting and summing process is repeated. All continuous temperature prediction values ​​are mapped to their corresponding spatial coordinates to generate a three-dimensional temperature distribution map, which is then defined as the initial temperature field mapping.

[0017] The above technical solution achieves a precise mapping from sparse discrete point measurements to a continuous spatial temperature distribution, providing a high-fidelity data foundation for subsequent thermal conductivity difference region division and temperature anomaly detection, and effectively avoiding temperature field distortion caused by sparse sensor arrangement and noise interference.

[0018] Further, in S1, the thermal conductivity difference region is divided according to the initial temperature field mapping to determine the thermal conductivity non-uniformity partition, including: A density-based spatial clustering algorithm is used to jointly cluster the temperature data and three-dimensional spatial coordinates in the initial temperature field mapping, dividing the temperature distribution region into multiple temperature distribution regions. The heating rate and heat flux density per unit area of ​​each temperature distribution region are extracted, and the thermal conductivity parameters of each region are calculated. The absolute value of the difference between the thermal conductivity parameters of adjacent regions is calculated as the thermal conductivity difference value. The boundary of adjacent regions and the preset range on both sides of the thermal conductivity difference value are marked as thermally non-uniform partitions. A unique region identifier is generated for each thermally non-uniform partition and its location and range are recorded. The relevant data of the thermally non-uniform partitions are stored.

[0019] Specifically, because moxibustion chambers typically contain multiple heterogeneous materials such as metal mesh and thermal insulation cotton, these materials have inherent differences in thermal conductivity and heat capacity, resulting in significant non-uniformity in heat transfer within the chamber. If the entire temperature field is directly treated as a homogeneous medium for subsequent anomaly detection, normal temperature jumps caused by material differences may be misjudged as abnormalities, or actual areas of thermal conductivity defects may be missed.

[0020] To address the aforementioned technical challenges, a density-based spatial clustering algorithm is employed to jointly cluster the temperature data and 3D spatial coordinates in the initial temperature field mapping. Density-based spatial clustering is an unsupervised learning method that does not require pre-specifying the number of clusters. Its core idea is to group spatially adjacent points with similar attributes into the same cluster by calculating the density connectivity between data points, while effectively identifying and filtering out isolated noise points. Specifically, the temperature value and corresponding 3D spatial coordinates of each sampling point in the initial temperature field mapping are first obtained, forming a multidimensional feature vector to be clustered. Then, a spatial distance metric and a temperature difference metric are defined between any two sampling points. When the spatial distance between two sampling points is less than a preset distance threshold and the temperature difference is less than a preset temperature difference threshold, they are determined to be density-connected and grouped into the same density-connected cluster. By traversing all sampling points, multiple independent clusters are ultimately formed, each corresponding to a temperature distribution region. The sampling points within each region exhibit high similarity in both spatial location and temperature value.

[0021] After dividing the temperature distribution areas, it is necessary to further extract the thermal conductivity parameters of each area to quantify the differences in thermal conductivity between areas. Thermal conductivity parameters are physical quantities reflecting the sensitivity of a region to heat input; they are defined as the rate of temperature rise under a unit heat flux density. A larger parameter value indicates faster heat conduction in that region. Specifically, for each divided temperature distribution area, the temperature change over multiple consecutive heating cycles is obtained, and the ratio of the temperature change to the time cycle is calculated to obtain the temperature rise rate of that region. Simultaneously, based on the current heating power data of the moxibustion heat box and the surface area of ​​the region, the heat flux density per unit area is calculated. Dividing the temperature rise rate by the heat flux density yields the thermal conductivity parameter for that region. By iterating through all adjacent temperature distribution areas, the absolute value of the difference in thermal conductivity parameters between adjacent areas is calculated. This absolute value is used as the thermal conductivity difference value to quantify the degree of deviation in thermal response characteristics between adjacent areas.

[0022] Based on the thermal conductivity difference value, areas with significant thermal conductivity differences are marked as thermally uneven partitions. Specifically, a thermal conductivity difference threshold is preset, which is determined based on the thermophysical characteristics of each component material in the moxibustion chamber. The calculated thermal conductivity difference value between each adjacent region is compared with the thermal conductivity difference threshold one by one. If the thermal conductivity difference value is greater than the threshold, it indicates that there is a significant abrupt change in thermal response characteristics at the boundary between the adjacent regions, which may be caused by structural features such as the interface between the metal parts and the insulation material. At this time, the boundary between the two adjacent temperature distribution regions and the range extending to both sides by a preset width are marked as thermally uneven partitions. For each generated thermally uneven partition, a unique digital code is assigned as a region identifier according to the order of its generation, and the three-dimensional spatial coordinate set of all sampling points on the edge contour of the thermally uneven partition is extracted, recording its precise location and range information. Finally, the region identifier, location range information, and corresponding thermal conductivity difference value of each thermally uneven partition are combined into structured partition data and persistently saved in the non-volatile memory of the moxibustion chamber control system.

[0023] The above technical solution uses a density-based spatial clustering algorithm to adaptively divide the temperature field data into regions without pre-setting the number of partitions, and can truly reflect the complex thermal conductivity distribution inside the temperature chamber. By defining and calculating thermal conductivity parameters and thermal conductivity difference values, the traditional static temperature difference analysis is upgraded to dynamic thermal response characteristic analysis, which can more fundamentally reveal the structural differences in heat transfer efficiency in different regions, providing a reliable spatial division basis for subsequent anomaly location and accurate compensation.

[0024] S2. Extract temperature gradient values ​​from the thermally uneven partition, determine temperature anomaly regions, and generate error quantification indicators and corresponding spatial distribution characteristics and temporal evolution characteristics; specifically including: Calculate the maximum temperature gradient value of each sampling point within the non-uniform thermal conductivity zone, compare the maximum temperature gradient value with a preset normal thermal conductivity limit threshold, and mark the corresponding area where the maximum temperature gradient value exceeds the preset threshold as a physical error area. The error amplitude is obtained by multiplying the basic error value of each physical error region by the corresponding spatial weight. The error amplitude of all physical error regions is summed to obtain the error quantification index. The error quantification index is then subjected to maximum and minimum normalization to obtain the normalized error quantification index. The spatial distribution characteristics and temporal evolution characteristics of the error quantification index are then generated.

[0025] Specifically, the non-uniform thermal conductivity zones identify boundary areas within the heat chamber where significant abrupt changes in thermal response characteristics occur due to differences in materials or structure. However, simply knowing the location and extent of these zones is insufficient to quantify their actual impact on temperature field uniformity. In the actual operation of moxibustion heat chambers, localized heat accumulation or abnormal heat loss often occurs within the non-uniform thermal conductivity zones, manifesting as drastic temperature changes between adjacent locations. This abnormal temperature gradient is both an external manifestation of material aging or structural defects and a key quantitative basis for subsequent error compensation and strategy optimization. Therefore, this step further extracts temperature gradient values ​​from the defined non-uniform thermal conductivity zones, identifies physical error regions, and generates error quantification indicators that include spatial distribution characteristics and temporal evolution characteristics.

[0026] During implementation, the maximum temperature gradient value of each sampling point within the non-uniform thermal conductivity zone is first calculated. Specifically, for each sampling point within the non-uniform thermal conductivity zone, its spatial coordinates are used as the center, and a preset Euclidean distance is used as the radius to determine the neighborhood range of that sampling point. The spatial coordinates and real-time temperature values ​​of all other sampling points within this neighborhood are then obtained. The temperature gradient value is defined as the ratio of the absolute value of the temperature difference between two sampling points to the straight-line distance between them, characterizing the degree of temperature change per unit distance. For the central sampling point, the temperature gradient values ​​between it and each sampling point in its neighborhood are calculated in different directions, and the largest value is selected as the maximum temperature gradient value for that central sampling point. This calculation process traverses all sampling points within the entire non-uniform thermal conductivity zone, obtaining the maximum temperature gradient value corresponding to each sampling point, thus constituting the temperature gradient distribution of the zone.

[0027] Subsequently, the maximum temperature gradient value at each sampling point is compared with the preset normal thermal conductivity limit threshold to identify areas of temperature abnormality. The normal thermal conductivity limit threshold is a pre-calibrated physical limit value based on the thermal conductivity characteristics of the heat-conducting material used in the moxibustion heat box under normal working conditions, determined by testing the maximum temperature gradient capability of a standard material sample under defect-free conditions. When the maximum temperature gradient value at a certain location exceeds this threshold, it indicates that the heat transfer at that location has exceeded the range allowed by the material's normal thermal conductivity, potentially indicating physical defects such as poor contact, local aging, or abnormal thermal resistance. When the maximum temperature gradient value is less than or equal to this threshold, the temperature change at that location is considered to be within the normal thermal conductivity range and is not marked. Accordingly, for sampling points with a maximum temperature gradient value greater than the normal thermal conductivity limit threshold, this sampling point and its surrounding area are collectively marked as a physical error region, and the boundary coordinates of this physical error region are recorded. Multiple physical error regions may be marked within a thermally uneven zone, or no physical error regions may be marked due to a relatively flat temperature field.

[0028] Based on the marked physical error regions, a quantitative error index is further generated. This quantitative error index is a comprehensive value reflecting the overall degree of thermal conductivity anomaly caused by physical errors in the current moxibustion chamber. Its calculation process considers both the degree of anomaly and the scope of influence of each physical error region. Specifically, for each physical error region, the maximum temperature gradient value of all sampling points within that region is extracted. The difference between the maximum temperature gradient value of each sampling point and the normal thermal conductivity limit threshold is calculated, and the arithmetic mean of these differences is taken as the basic error value for that physical error region. Simultaneously, the area of ​​this physical error region within the moxibustion chamber is obtained, and a corresponding spatial weight is assigned according to a preset area-weight correspondence. This correspondence can be determined based on experimental calibration or empirical formulas, ensuring that a larger physical error region has a more significant impact on the overall temperature field uniformity, and thus a larger spatial weight. Multiplying the basic error value by the spatial weight yields the error amplitude of that physical error region. By iterating through all physical error regions within the moxibustion chamber and summing the error amplitudes of each region, the resulting sum is the quantitative error index reflecting the overall degree of thermal conductivity anomaly in the current chamber.

[0029] Because the error quantification indicators generated under different operating conditions may have significant numerical range differences, to facilitate subsequent model processing, it is necessary to perform maximum-minimum normalization on the error quantification indicators based on the maximum and minimum error quantification indicator values ​​that have appeared in the historical operation records of the moxibustion heat chamber. Based on the normalization process, the spatial distribution characteristics and temporal evolution characteristics of the error quantification indicators are further generated. The internal space of the heat chamber is pre-divided into a fixed number of grid cells. The error amplitude of each physical error region is mapped to the corresponding grid cell according to spatial coordinates. If a physical error region exists within a grid cell, the corresponding error amplitude value is used; otherwise, zero is used. Each grid cell is numbered, thereby generating a sequence of error amplitude values ​​with fixed dimensions arranged according to the grid cell numbers, which serves as the spatial distribution feature vector, i.e., the spatial distribution characteristic. The temporal evolution characteristic consists of a time series data vector composed of normalized error quantification indicators from multiple consecutive heating cycles arranged in chronological order, reflecting the evolution trend of the degree of heat conduction anomaly with the operating time of the equipment.

[0030] The above technical solution can not only accurately locate the physical error area that exceeds the normal thermal conductivity of the material, avoiding misjudging the normal temperature field gradient as anomaly, but also extract two types of features: spatial distribution and temporal evolution. This provides rich and structured input for the subsequent aging prediction model, which is conducive to the model capturing the spatiotemporal evolution law of material performance degradation more accurately, thereby improving the stability and adaptability of long-term temperature field control.

[0031] S3. Input the spatial distribution characteristics and temporal evolution characteristics of the error quantification index into the pre-trained prediction model to predict the material aging trend and determine the aging influence coefficient; construct an optimization objective function based on the aging influence coefficient and the current temperature data, and solve for the adjusted heating strategy. In S3, predicting the material aging trend and determining the aging influence coefficient includes: The spatial distribution characteristics and temporal evolution characteristics of the error quantification index are input into a pre-trained multilayer perceptron neural network to analyze the correlation between material aging and temperature anomalies. The thermal conductivity decay rate and thermal resistance increase value are output by the multilayer perceptron neural network as material aging trend data. The initial aging influence coefficient is obtained by weighted summation of the thermal conductivity decay rate and the increase in thermal resistance. The corresponding environmental compensation coefficient is obtained based on the current ambient temperature and humidity data. The initial aging influence coefficient is multiplied by the environmental compensation coefficient to obtain the calibrated aging influence coefficient. The calibrated aging influence coefficient is stored in the control register.

[0032] Specifically, the generated error quantification indicators and their spatial distribution and temporal evolution characteristics depict the distribution pattern and evolution trend of thermal conductivity anomalies inside the moxibustion chamber from both spatial and temporal dimensions. However, these characteristics only reflect the abnormal performance of the temperature field and have not yet established a direct correlation with the degradation of the physical properties of the materials themselves. During the long-term operation of the moxibustion chamber, the insulation and heat-conducting materials gradually age, manifested as a decrease in thermal conductivity and an increase in thermal resistance. This degradation of material properties is the fundamental reason for the continuous deterioration of temperature field uniformity. Therefore, it is necessary to establish a quantitative mapping relationship between temperature anomaly characteristics and the degree of material aging, so as to predict the material aging trend in advance before significant instability occurs in the temperature field, and thus proactively adjust the control strategy for compensation.

[0033] In implementation, a multilayer perceptron neural network (MLPN) was used as the prediction model. An MPN is a feedforward artificial neural network consisting of an input layer, multiple hidden layers, and an output layer. The layers are fully connected, and each connection has trainable weight parameters. This network transforms and abstracts the input features layer by layer through nonlinear activation functions, enabling it to learn the complex nonlinear mapping relationship between input features and the output target. The aforementioned MPN was used to establish a prediction model for data ranging from the spatiotemporal characteristics of error quantification indicators to material aging trend data.

[0034] The training process of the multilayer perceptron neural network was completed offline, and the training data was obtained through a combination of accelerated aging experiments and simulation. Specifically, samples of thermally conductive and insulating materials identical to those used in the actual moxibustion chamber were selected and subjected to accelerated thermal aging experiments in an aging test chamber with a controlled temperature and humidity and the same structure as the moxibustion chamber. Multiple aging time nodes were set, and at each time node, the thermal resistance and thermal conductivity of the samples were measured using a heat flux meter and a laser thermal conductivity meter, respectively. The thermal conductivity decay rate and thermal resistance increase value at each time node were recorded as training labels. Simultaneously, each sample was assembled in a moxibustion chamber simulation experimental platform, and a standard heating program was executed. Temperature field data was collected through a sensor array, and the spatial distribution characteristics and temporal evolution characteristics of the error quantification index corresponding to each aging time node were calculated as training inputs according to the above steps. The above multiple sets of training inputs and training labels were paired to form a training dataset. The multilayer perceptron neural network was trained under supervision using the backpropagation algorithm and the mean squared error loss function. The weight parameters of the network were iteratively updated using a gradient descent optimizer until the loss function converged to below a preset threshold, resulting in the trained multilayer perceptron neural network. The trained network was deployed to the inference module of the moxibustion chamber control system for online prediction.

[0035] During the inference phase, the spatial distribution characteristics and temporal evolution characteristics of the error quantification indicators generated in the above implementation steps are used as input data and fed into the trained multilayer perceptron neural network. The input layer of the multilayer perceptron neural network receives a concatenated vector of spatial distribution feature vector and temporal evolution feature vector. The spatial distribution feature vector consists of a sequence of error amplitude values ​​arranged according to grid cell numbers, and the temporal evolution feature vector consists of normalized error quantification indicators from multiple consecutive heating cycles arranged in chronological order. After the input data undergoes linear transformation and nonlinear activation in the first hidden layer, shallow feature representations are extracted. The second hidden layer further performs nonlinear mapping on the shallow features to obtain deep feature representations. These deep feature representations are passed to the output layer, which contains two output nodes. One output node outputs the thermal conductivity decay rate, representing the percentage decrease in the heat transfer capacity of the thermally conductive material, and the other output node outputs the thermal resistance increase value, representing the increase in the thermal insulation material's ability to prevent heat loss. The thermal conductivity decay rate and thermal resistance increase value together constitute the material aging trend data, intuitively reflecting the degree of physical change in material performance.

[0036] After obtaining the material aging trend data, the aging influence coefficient is further calculated. This aging influence coefficient is a dimensionless value used to comprehensively quantify the impact of material aging on the temperature control performance of the moxibustion chamber. The specific calculation process is as follows: The thermal conductivity decay rate is assigned a first weight value, and the increase in thermal resistance is assigned a second weight value. The sum of the first and second weight values ​​is 1. The thermal conductivity decay rate is multiplied by the first weight value to obtain a first product, and the increase in thermal resistance is multiplied by the second weight value to obtain a second product. The first and second products are then added together, and the sum is the initial aging influence coefficient. The specific values ​​of the first and second weight values ​​are determined through experimental calibration based on the relative importance of the thermally conductive and insulating materials in the overall thermal balance of the moxibustion chamber.

[0037] Because the external environment significantly affects the heat dissipation rate and moisture absorption of the moxibustion chamber material when it operates under varying temperature and humidity conditions, thus altering the material's actual thermal conductivity, directly using the initial aging influence coefficient for control parameter optimization may introduce deviations caused by environmental factors. Therefore, environmental calibration of the initial aging influence coefficient is necessary. Specifically, ambient temperature and humidity data are acquired using sensors placed outside the moxibustion chamber. The corresponding environmental compensation coefficient is then found in a pre-set compensation coefficient table based on this data. This table was obtained through standard thermal response tests at different temperature and humidity ranges, recording the compensation coefficient values ​​required to eliminate environmental interference. Multiplying the initial aging influence coefficient by the environmental compensation coefficient yields the calibrated aging influence coefficient. This calibrated coefficient eliminates the interference of external environmental fluctuations on material aging assessment, accurately reflecting the material's physical aging state.

[0038] Finally, the calibrated aging influence coefficient is stored in a designated control register of the moxibustion chamber's main control chip. In subsequent heating cycles, the main control chip directly reads the calibrated aging influence coefficient from this control register and uses it to apply weight correction to the temperature deviation term when constructing the optimization objective function. This results in a higher penalty weight for temperature deviations in severely aged areas during the optimization solution, thereby driving the optimization algorithm to prioritize temperature control accuracy in severely aged areas and achieving adaptive compensation of control parameters for material aging.

[0039] The above technical solution establishes a nonlinear mapping from the spatiotemporal characteristics of temperature anomalies to the degree of material aging through a pre-trained multilayer perceptron neural network, realizing indirect perception and quantitative prediction of the aging of invisible materials. Through weighted summation and dual calibration of environmental compensation coefficients, the interference of differences in the influence of different materials and external environmental fluctuations on aging assessment is eliminated, making the aging influence coefficient accurate and reliable. The aging influence coefficient is written into the control register as a reference for optimized control, providing a direct quantitative input for the adaptive adjustment of subsequent heating strategies, thereby realizing a closed-loop connection from anomaly diagnosis to control compensation.

[0040] Furthermore, in S3, the adjusted heating strategy is generated, including: Obtain the current temperature data of each area of ​​the moxibustion heat box and the aging influence coefficient, construct a dual optimization objective function to minimize temperature deviation and heating energy consumption, and use the aging influence coefficient as the weighting factor of the temperature deviation term. The particle swarm optimization algorithm is used to iteratively solve the optimization objective function to obtain the optimal heating power and heating time parameters. The optimal heating power and heating time parameters are then encoded into control commands to generate an adjusted heating strategy. The adjusted heating strategy is input into the finite element heat conduction simulation environment for feasibility verification. When it is confirmed that the highest temperature value of all areas does not exceed the safe temperature threshold and the temperature deviation of each area is less than the preset allowable deviation threshold, the adjusted heating strategy is output to the moxibustion chamber control system.

[0041] Specifically, after obtaining the calibrated aging influence coefficient, although this coefficient quantifies the degree of impact of material aging on thermal conductivity in each region, it has not yet been translated into directly executable control commands. The heating control of the moxibustion chamber needs to simultaneously coordinate two mutually constraining objectives: temperature accuracy and energy consumption. Furthermore, the response characteristics of severely aged regions to heat input have changed, and these must be given priority in parameter optimization. Therefore, it is necessary to combine the current temperature data with the aging influence coefficient to construct a multi-objective optimization problem and solve it through an intelligent optimization algorithm, generating an adjusted heating strategy that balances temperature uniformity, energy consumption control, and aging compensation.

[0042] During implementation, the current temperature data and aging influence coefficient of each area of ​​the moxibustion heat box are first acquired, and a dual optimization objective function is constructed to minimize temperature deviation and heating energy consumption. The current temperature data is obtained in real-time from the sensor array and after preprocessing and smoothing, including the temperature values ​​of each divided area. The aging influence coefficient is read from the control register. For areas with a larger aging influence coefficient, the material performance degradation is more severe, and the tolerance for temperature deviation should be lower, requiring a higher penalty weight in the objective function. The specific construction method of the optimization objective function is as follows: the sum of the squares of the differences between the actual temperature value of each area and the preset target temperature is used as the temperature deviation term, and the sum of the products of the heating power and heating duration of each heating element is used as the energy consumption term. The temperature deviation term and the energy consumption term are combined through weighting coefficients to form a comprehensive optimization objective function. Specifically, the squares of the temperature differences of each area in the temperature deviation term are multiplied by the aging influence coefficient of the corresponding area before summing, so that the temperature deviation of severely aging areas receives a greater weight in the optimization objective, and is thus prioritized for reduction during the optimization process.

[0043] Subsequently, the particle swarm optimization algorithm is used to iteratively solve the above-mentioned objective function. In this embodiment, the position vector of each particle in the particle swarm optimization algorithm includes parameters in two dimensions: the heating power of each heating element in the moxibustion heat box and the continuous heating time. In the initialization phase, the particle swarm size is set, and the initial positions of each particle are randomly generated within the hardware-allowed range of heating power and the reasonable range of heating time, and the initial velocities of each particle are randomly initialized. In the iteration phase, for each particle, the heating power and time parameters represented by its position vector are substituted into the objective function to calculate the fitness value of the particle. The current fitness value of each particle is compared with its historical best fitness value. If the current fitness value is better, the individual optimal position is updated. The current fitness values ​​of all particles in the swarm are compared with the global historical best fitness value. If any particle is better, the global optimal position is updated. Each particle updates its own velocity according to its individual optimal position and global optimal position. An inertia weight is introduced into the velocity update formula to balance the global search and local search capabilities, and an acceleration constant is introduced to adjust the step size for approaching the individual optimal and global optimal positions. The particle adjusts its position according to the updated velocity. Repeat the above iterative process until the preset maximum number of iterations is reached or the change in the global optimal fitness value is less than the preset convergence threshold multiple times in a row. The algorithm is then considered to have converged, and the heating power and heating time parameters corresponding to the global optimal position are output as the optimal solution.

[0044] After obtaining the optimal heating power and heating time parameters, they are encoded into control instructions that can be executed by the moxibustion chamber control system. Specifically, the optimal heating power of each heating element is converted into the duty cycle value of the pulse width modulation signal, and the optimal heating time is converted into the timer count value. The duty cycle value and the timer count value are encapsulated into a data frame according to the communication protocol of the control system to generate the adjusted heating strategy.

[0045] Before implementing the adjusted heating strategy on the actual hardware, its feasibility is verified to avoid the risk of local overheating or temperature field runaway due to improper control parameters. The verification process uses the finite element method for heat conduction simulation. A three-dimensional finite element mesh model of the moxibustion chamber is established. Mesh elements are divided according to the actual physical structure dimensions of the chamber, and each mesh element is assigned thermophysical properties such as thermal conductivity, specific heat capacity, and density corresponding to the material. The heating power corresponding to each heating element in the adjusted heating strategy is applied as a heat source boundary condition to the corresponding location on the model. The simulation time step and total simulation duration are set to be consistent with the heating time parameters in the heating strategy. By solving the transient heat conduction differential equation, the temperature value of each mesh node at each time step is calculated, obtaining the simulated temperature distribution data of each region inside the chamber during the entire heating process. After the simulation, the highest temperature value of each region during the simulation is extracted and compared with the preset safe temperature threshold. Simultaneously, the deviation between the simulated temperature value of each region and the preset target temperature value is calculated. If the highest temperature value of all regions does not exceed the safe temperature threshold, and the temperature deviation of each region is less than the preset allowable deviation threshold, then the adjusted heating strategy is deemed to have passed the feasibility verification. If the temperature in a region exceeds the safety threshold or the deviation is too large, the constraints of the particle swarm optimization algorithm or the weighting coefficients of the objective function are adjusted, and the solution and verification are repeated until a verified heating strategy is obtained.

[0046] The adjusted heating strategy, which has passed feasibility verification, is output to the underlying microcontroller of the moxibustion chamber control system via a serial communication bus. This updates the currently executed heating control parameters in the microcontroller's memory. The microcontroller then drives each heating element according to the new pulse width modulation duty cycle and timer settings, completing the real-time application of the heating strategy. Through this technical solution, the aging influence coefficient is embedded as a weighting factor in the temperature deviation term within the optimization objective function, seamlessly integrating material aging compensation with control parameter optimization, prioritizing temperature accuracy in severely aging areas. The particle swarm optimization algorithm efficiently solves multi-objective optimization problems in a complex multi-dimensional parameter space, balancing temperature accuracy and energy consumption. Finite element heat conduction simulation performs safety pre-verification before strategy application, avoiding the risk of local overheating and ensuring the safe and reliable adjustment of the control strategy.

[0047] S4. Apply the adjusted heating strategy, collect temperature data after execution through a sensor array, update the temperature field distribution data, calculate the uniformity index of the updated temperature field distribution data, and iteratively optimize the heating strategy based on the uniformity index; specifically including: The adjusted heating strategy is transmitted to the moxibustion chamber control system in real time. The duty cycle of the corresponding heating element is adjusted by pulse width modulation to perform the heating operation. The real-time temperature data of each area after heating is collected by the sensor array. The real-time temperature data is mapped to the preset three-dimensional spatial coordinate system to update the temperature field distribution information. The updated temperature field distribution data is smoothed by the weighted moving average algorithm to eliminate short-term fluctuations. The variance of temperature in each region is calculated based on the updated temperature field distribution data. The uniformity index is obtained by dividing the preset benchmark constant by the sum of the variance and the smoothing constant. If the uniformity index is lower than the preset qualified threshold, the inertia weight and acceleration constant of the particle swarm optimization algorithm are adjusted, and the particle swarm optimization algorithm is rerun to iteratively generate a new heating strategy.

[0048] Specifically, the adjusted heating strategy has been feasible and ready for implementation, but its actual improvement in temperature field uniformity still needs to be quantitatively evaluated using real temperature data. The raw temperature data collected by sensors is subject to high-frequency random noise due to uneven air convection and thermal radiation, making direct use for evaluation prone to misjudgment. Furthermore, whether the temperature field meets the standards requires clear quantitative measurement criteria; if it does not, parameter optimization must be triggered, otherwise, maintaining temperature field uniformity will be difficult.

[0049] During implementation, the adjusted heating strategy is first transmitted in real time to the moxibustion chamber control system for heating operation. The adjusted heating strategy was encoded in the previous step, encapsulating the pulse width modulation duty cycle value and timer count value corresponding to each heating element in the form of a data frame. After receiving the data frame, the control system parses the control parameters of each heating element and adjusts the duty cycle of the corresponding heating element through pulse width modulation. The higher the duty cycle, the more heat the heating element outputs per unit time. Each heating element synchronously executes the heating operation according to its corresponding duty cycle and timer setting value. After the timer reaches the set time, it automatically triggers an interrupt signal to stop heating or switch to the heat preservation stage.

[0050] During and after the heating operation, real-time temperature data for each area is continuously collected via a sensor array. This array consists of multiple thermocouple sensors evenly distributed on the inner wall of the moxibustion chamber and the support frame. Each sensor synchronously acquires the temperature value at its corresponding spatial location at a preset sampling frequency. The collected sensor temperature values ​​are then bound to their coordinates in a preset three-dimensional spatial coordinate system, replacing the original temperature values ​​at the corresponding coordinates in the temperature field distribution information, thus updating the temperature field distribution data. The updated temperature field distribution data reflects the actual temperature distribution under the adjusted heating strategy.

[0051] Due to random convection disturbances in the internal air during the heating process of the moxibustion heat box, as well as local non-uniformity of thermal radiation from the heating element, the temperature data collected by the sensor will be superimposed with high-frequency random noise, manifesting as drastic jumps in temperature values ​​between adjacent sampling times. To eliminate the interference of such short-term fluctuations on the assessment of temperature field uniformity, a weighted moving average algorithm is used to smooth the updated temperature field distribution data. Specifically, for each spatial coordinate point in the temperature field distribution data, its temperature values ​​at the current sampling time and multiple consecutive historical sampling times are extracted to construct a time series data window; according to a preset weighting scheme, the temperature value closest to the current sampling time is assigned the largest weight, and the temperature value farthest from the current sampling time is assigned the smallest weight; the temperature values ​​in the time series data window are multiplied by their corresponding weights and summed, and the resulting weighted average is used as the smoothed temperature value of the spatial coordinate point at the current sampling time, replacing the original temperature value. This process is repeated for all spatial coordinate points in the temperature field distribution data to complete the smoothing of all sampling points, resulting in smoothed temperature field distribution data after eliminating short-term fluctuations.

[0052] Based on the smoothed temperature field distribution data, a uniformity index is calculated to quantitatively evaluate the uniformity of the temperature field distribution. The uniformity index is a value inversely proportional to the dispersion of the temperature distribution; a larger index value indicates that the temperature in each region is more consistent, and the temperature field uniformity is better. Specifically, the current temperature value of each region in the smoothed temperature field distribution data is obtained, and the arithmetic mean of the temperature values ​​of all regions is calculated to obtain the global average temperature. The difference between the temperature value of each region and the global average temperature is calculated, and the square of the difference is calculated. The sum of the squared values ​​for all regions is then divided by the total number of regions to obtain the temperature variance value. This variance value reflects the dispersion of the temperature in each region relative to the global average temperature. A preset baseline constant is divided by the sum of the temperature variance value and the smoothing constant; the quotient obtained is the uniformity index. The smoothing constant is a preset positive number used to prevent singularities in the division operation when the temperature variance value is zero, and also to adjust the sensitivity of the uniformity index to small temperature field fluctuations.

[0053] The calculated uniformity index is compared with a preset uniformity threshold. The uniformity threshold is a minimum acceptable uniformity level preset according to the therapeutic standards of the moxibustion heat box. When the uniformity index is greater than or equal to the threshold, it indicates that the current temperature field uniformity meets the requirements, and the current heating strategy continues to operate. When the uniformity index is lower than the threshold, it indicates that there is still a large degree of non-uniformity in the current temperature field distribution, and the heating strategy needs to be further iteratively optimized.

[0054] The iterative optimization process re-runs the particle swarm optimization algorithm and adjusts its key parameters to enhance its optimization capability. Key parameters of the particle swarm optimization algorithm include inertia weight and acceleration constant. Inertia weight controls the ability of particles to maintain their original velocity; a larger inertia weight is beneficial for global search, while a smaller inertia weight is beneficial for local fine-grained search. The acceleration constant is divided into individual acceleration constant and global acceleration constant, which adjust the step size of particles approaching their individual and global optimal positions, respectively. Based on the ratio of the difference between the uniformity index and the acceptable threshold, parameter adjustment coefficients are calculated: a larger difference ratio indicates a more severe deviation of the current temperature field from the target, requiring enhanced global search capability to escape possible local optima; in this case, the inertia weight is increased and the global acceleration constant is appropriately decreased. A smaller difference ratio indicates that the current temperature field is close to the target state, requiring enhanced local fine-grained search capability; in this case, the inertia weight is decreased and the individual acceleration constant is appropriately increased. Using the adjusted inertia weights and acceleration constants, the positions and velocities of each particle in the particle swarm are reinitialized. The particle swarm optimization algorithm is then run iteratively to solve the objective function. Each particle's position vector still contains parameters for both the heating power and duration of each heating element. The fitness value remains a weighted sum of the temperature deviation and energy consumption terms. After iterating until the convergence condition is met, the heating power and duration parameters corresponding to the new global optimal position are output. Following the same coding and finite element simulation verification as described above, a new heating strategy is generated and applied again to actual heating control until the uniformity index reaches a satisfactory threshold. Figure 2 As shown, the curve of uniformity index changing with the number of iterations is given. The index value rises rapidly from the initial value of about 0.60, and stabilizes above 0.90 after about 12 iterations, exceeding the preset qualified threshold (0.85 dashed line). This verifies that adaptive iterative optimization can effectively converge to a uniform and stable temperature field.

[0055] The above technical solution uses a weighted moving average algorithm to smooth the updated temperature field data in the time domain, effectively filtering out high-frequency noise caused by random convection and uneven thermal radiation, thus avoiding misjudgments in uniformity assessment. The uniformity index is expressed as the sum of the baseline constant and the variance and the smoothing constant, and is inversely proportional to the degree of temperature dispersion, intuitively reflecting the level of temperature field uniformity. When the uniformity index is below the threshold, the particle swarm optimization algorithm is called again and the inertia weight and acceleration constant are adjusted according to the degree of deviation, which enhances the algorithm's adaptability to optimization under different degrees of deviation and ensures that the heating strategy continuously approaches the optimum.

[0056] S5. Continuously monitor the operation status of the moxibustion chamber using the optimized heating strategy to generate a stable temperature field record, specifically including: The uniformity improvement effect of the new heating strategy was verified in a virtual simulation environment. The change process of the uniformity index and the adjustment of algorithm parameters were recorded to form an optimization log. The verified effective heating strategy was used as the basic strategy. The operating temperature data of the temperature chamber was continuously collected and the temperature field update and uniformity evaluation were performed in a loop to generate a stable temperature field record containing timestamps and temperature field data.

[0057] Specifically, based on the new heating strategy generated through the above iterations, the heating power and time parameters have been re-optimized according to feedback from the uniformity index. However, the effect of the new heating strategy on improving temperature field uniformity has not yet been practically verified, and there is uncertainty in directly putting it into long-term operation. As a physiotherapy device that needs to work continuously for a long time, the operating status of the moxibustion heat box needs to be continuously monitored, and the temperature field distribution data needs to be systematically recorded to form a traceable operating record, providing a basis for equipment maintenance, performance evaluation, and subsequent optimization.

[0058] In this embodiment, the uniformity improvement effect of the new heating strategy is verified in a virtual simulation environment. The virtual simulation environment is the finite element heat conduction simulation model used in the previous steps. This model is established based on the actual physical structure dimensions of the moxibustion heat box and the thermal physical properties parameters such as thermal conductivity, specific heat capacity, and density of each component material. It can simulate the temperature field change process after the heating strategy is executed with high fidelity. The heating power corresponding to each heating element in the new heating strategy is loaded as the heat source boundary condition to the corresponding position of the simulation model. The simulation time step and total simulation duration are set to be consistent with the heating time parameters in the heating strategy. The temperature values ​​of each grid node at each time step are obtained by solving the transient heat conduction differential equation, forming simulated temperature field distribution data. The expected uniformity index is calculated based on the simulated temperature field distribution data according to the same uniformity index calculation method as in the previous steps. The expected uniformity index is compared with the uniformity index before adjustment. If the expected uniformity index is greater than the uniformity index before adjustment, it is determined that the new heating strategy has an effect on improving uniformity; if the value of the expected uniformity index reaches or exceeds the preset uniformity qualification threshold, the strategy is confirmed to be effective and can be put into actual operation.

[0059] During the verification process, the changes in uniformity index and the adjustments to algorithm parameters are recorded synchronously, forming an optimization log. The optimization log is a structured data record used to trace the entire temperature field optimization process. It includes the uniformity index values ​​before and after adjustment, the corresponding timestamps, and the changes in algorithm parameters triggered by the adjustment. Specifically, the uniformity index values ​​before adjustment, the expected uniformity index values ​​after this adjustment, the current system timestamp, the inertia weights and acceleration constants of the particle swarm optimization algorithm before adjustment, and the inertia weights and acceleration constants after adjustment are correlated and concatenated to form a complete optimization record entry. This optimization record entry is appended to the non-volatile memory of the moxibustion chamber control system in chronological order, while existing historical optimization record entries remain unchanged, thus forming a continuously growing optimization log text arranged in rows. The continuous accumulation of the optimization log provides a complete audit trail for subsequent analysis of temperature field evolution patterns, evaluation of the effectiveness of algorithm parameter adjustment strategies, and equipment fault diagnosis.

[0060] The heating strategy, validated through virtual simulation, is set as the current default execution scheme, i.e., the basic strategy. The control system continuously drives the heating hardware according to the power and time parameters of each heating element specified in this basic strategy, and enters a continuous operation status monitoring loop based on this. In the monitoring loop, the sensor array continuously collects real-time temperature data of each area inside the chamber at a preset sampling frequency. The collected temperature data is mapped to a preset three-dimensional spatial coordinate system to update the temperature field distribution information, and the current uniformity index is calculated after weighted moving average smoothing. If the uniformity index remains at a level greater than or equal to the uniformity qualification threshold, the basic strategy remains unchanged, and the cycle of collection and evaluation continues. If the uniformity index falls below the uniformity qualification threshold again, a new round of iteration is triggered, generating an updated heating strategy, which is then validated through virtual simulation again before replacing the current basic strategy. This cycle repeats continuously, forming a continuous closed loop for the control of the moxibustion chamber.

[0061] During the aforementioned continuous monitoring process, a stable temperature field record is generated synchronously. This stable temperature field record is a set of temperature field state data arranged in a time series, used to comprehensively record the temperature distribution history of the moxibustion heat box during operation. Specifically, in each sampling period, the current system timestamp, the temperature values ​​of each region in the updated and smoothed temperature field distribution data, the corresponding uniformity index values, and the currently effective basic strategy version identifier are structurally combined to form a temperature field record entry. This temperature field record entry is written to non-volatile memory in chronological order, forming a stable temperature field record containing consecutive timestamps and corresponding temperature field data. This stable temperature field record can be used for equipment operation status retrospective analysis, therapeutic effect correlation analysis, and long-term temperature field stability assessment.

[0062] The above technical solution uses a virtual simulation environment to pre-verify the uniformity improvement effect of the new heating strategy, ensuring the effectiveness of the strategy before putting it into actual operation, thus avoiding interference from ineffective strategies on the stability of the temperature field. The optimization log fully records the evolution trajectory of the uniformity index and the process of algorithm parameter adjustment, providing traceable data support for system maintenance and performance analysis. Through continuous cyclical monitoring, evaluation and optimization, as well as the generation of stable temperature field records, the moxibustion heat box can maintain the uniformity of the temperature field for a long time, and provides a complete operational archive for efficacy evaluation and equipment management.

[0063] This invention also provides a moxibustion chamber control system based on multi-sensor data fusion, used to implement the above-mentioned method, such as... Figure 3 As shown, the system includes: The temperature field partitioning unit is used to collect real-time temperature data of multiple areas inside the moxibustion heat box through a sensor array, preprocess the real-time temperature data to construct an initial temperature field mapping, and divide the thermal conductivity difference area according to the initial temperature field mapping to determine the thermal conductivity non-uniform partition. The error feature quantization unit is used to extract temperature gradient values ​​from the thermally uneven partition, determine the temperature abnormal area, and generate error quantization index and corresponding spatial distribution characteristics and temporal evolution characteristics. The heating strategy solving unit is used to input the spatial distribution characteristics and temporal evolution characteristics of the error quantification index into a pre-trained prediction model to predict the material aging trend and determine the aging influence coefficient; and to construct an optimization objective function based on the aging influence coefficient and the current temperature data to solve for the adjusted heating strategy. The strategy evaluation and iteration unit is used to apply the adjusted heating strategy, collect temperature data after execution through the sensor array, update the temperature field distribution data, calculate the uniformity index of the updated temperature field distribution data, and iteratively optimize the heating strategy based on the uniformity index. The temperature field monitoring unit is used to continuously monitor the operating status of the moxibustion chamber through an optimized heating strategy and generate a stable temperature field record.

[0064] The present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the above-described method.

[0065] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0066] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0067] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling a moxibustion chamber based on multi-sensor data fusion, characterized in that, The method includes: S1. Real-time temperature data of multiple areas inside the moxibustion heat box are collected through a sensor array. After preprocessing the real-time temperature data, an initial temperature field mapping is constructed. Based on the initial temperature field mapping, regions with different thermal conductivity are divided to determine the uneven thermal conductivity zones. S2. Extract temperature gradient values ​​from the non-uniform thermal conductivity partition, determine the temperature anomaly region, and generate error quantification index and corresponding spatial distribution characteristics and temporal evolution characteristics. S3. Input the spatial distribution characteristics and temporal evolution characteristics of the error quantification index into the pre-trained prediction model to predict the material aging trend and determine the aging influence coefficient; construct an optimization objective function based on the aging influence coefficient and the current temperature data, and solve for the adjusted heating strategy. S4. Apply the adjusted heating strategy, collect the temperature data after execution through the sensor array, update the temperature field distribution data, calculate the uniformity index of the updated temperature field distribution data, and iteratively optimize the heating strategy based on the uniformity index. S5. Continuously monitor the operation status of the moxibustion chamber through the optimized heating strategy and generate a stable temperature field record.

2. The method according to claim 1, characterized in that, In S1, the initial temperature field mapping is constructed, including: Inside the moxibustion chamber, temperature sensing nodes are arranged at a preset grid spacing to form a sensor array. The temperature values ​​of each temperature sensing node are read synchronously according to a preset sampling frequency. The spatial coordinates of each temperature sensing node are bound to the corresponding temperature value to generate real-time temperature data. The real-time temperature data was filtered in the time domain using the Kalman filter algorithm to remove missing data and obtain a clean temperature dataset. The spatial coordinates and temperature data of each temperature sensing node in the clean temperature dataset are extracted. The Kriging interpolation algorithm is used to perform spatial continuity calculation. A spherical semivariogram function is fitted to determine the spatial variation parameters. A three-dimensional temperature distribution map is constructed to generate an initial temperature field mapping.

3. The method according to claim 2, characterized in that, In S1, the non-uniform thermal conductivity zones are identified, including: A density-based spatial clustering algorithm is used to jointly cluster the temperature data and three-dimensional spatial coordinates in the initial temperature field mapping, dividing the temperature distribution region into multiple temperature distribution regions. The heating rate and heat flux density per unit area of ​​each temperature distribution region are extracted, and the thermal conductivity parameters of each region are calculated. The absolute value of the difference in thermal conductivity parameters between adjacent regions is calculated as the thermal conductivity difference value. The boundary and the preset range on both sides of the adjacent regions where the thermal conductivity difference value is greater than a preset threshold are marked as thermally non-uniform partitions. A unique region identifier is generated for each thermally non-uniform partition and its location and range are recorded. The relevant data of the thermally non-uniform partitions are stored.

4. The method according to claim 1, characterized in that, In S2, error quantification indicators and their corresponding spatial distribution and temporal evolution characteristics are generated, including: Calculate the maximum temperature gradient value of each sampling point within the non-uniform thermal conductivity zone, compare the maximum temperature gradient value with a preset normal thermal conductivity limit threshold, and mark the corresponding area where the maximum temperature gradient value exceeds the preset threshold as a physical error area. The error amplitude is obtained by multiplying the basic error value of each physical error region by the corresponding spatial weight. The error amplitude of all physical error regions is summed to obtain the error quantification index. The error quantification index is then subjected to maximum and minimum normalization to obtain the normalized error quantification index. The spatial distribution characteristics and temporal evolution characteristics of the error quantification index are then generated.

5. The method according to claim 1, characterized in that, In S3, the material aging trend is predicted and the aging effect coefficient is determined, including: The spatial distribution characteristics and temporal evolution characteristics of the error quantification index are input into a pre-trained multilayer perceptron neural network to analyze the correlation between material aging and temperature anomalies. The thermal conductivity decay rate and thermal resistance increase value are output by the multilayer perceptron neural network as material aging trend data. The initial aging influence coefficient is obtained by weighted summation of the thermal conductivity decay rate and the increase in thermal resistance. The corresponding environmental compensation coefficient is obtained based on the current ambient temperature and humidity data. The initial aging influence coefficient is multiplied by the environmental compensation coefficient to obtain the calibrated aging influence coefficient. The calibrated aging influence coefficient is stored in the control register.

6. The method according to claim 5, characterized in that, In S3, the adjusted heating strategy is generated, including: Obtain the current temperature data of each area of ​​the moxibustion heat box and the aging influence coefficient, construct a dual optimization objective function to minimize temperature deviation and heating energy consumption, and use the aging influence coefficient as the weighting factor of the temperature deviation term. The particle swarm optimization algorithm is used to iteratively solve the optimization objective function to obtain the optimal heating power and heating time parameters. The optimal heating power and heating time parameters are then encoded into control commands to generate an adjusted heating strategy. The adjusted heating strategy is input into the finite element heat conduction simulation environment for feasibility verification. When it is confirmed that the highest temperature value of all areas does not exceed the safe temperature threshold and the temperature deviation of each area is less than the preset allowable deviation threshold, the adjusted heating strategy is output to the moxibustion chamber control system.

7. The method according to claim 1, characterized in that, In S4, the temperature field distribution data is updated, and a new heating strategy is iteratively generated, including: The adjusted heating strategy is transmitted to the moxibustion chamber control system in real time. The duty cycle of the corresponding heating element is adjusted by pulse width modulation to perform the heating operation. The real-time temperature data of each area after heating is collected by the sensor array. The real-time temperature data is mapped to the preset three-dimensional spatial coordinate system to update the temperature field distribution information. The weighted moving average algorithm is used to smooth the updated temperature field distribution data to eliminate short-term fluctuations. The variance of temperature in each region is calculated based on the updated temperature field distribution data. The uniformity index is obtained by dividing the preset benchmark constant by the sum of the variance and the smoothing constant. If the uniformity index is lower than the preset qualified threshold, the inertia weight and acceleration constant of the particle swarm optimization algorithm are adjusted, and the particle swarm optimization algorithm is rerun to iteratively generate a new heating strategy.

8. The method according to claim 1, characterized in that, The heating strategy is iteratively optimized based on the uniformity index, and a stable temperature field record is generated, including: The uniformity improvement effect of the new heating strategy is verified in a virtual simulation environment. The change process of uniformity index and the adjustment amount of algorithm parameters are recorded to form an optimization log. The verified effective heating strategy is used as the basic strategy. The operating temperature data of the temperature chamber is continuously collected and the temperature field update and uniformity evaluation are performed in a loop to generate a stable temperature field record containing timestamps and temperature field data.

9. A moxibustion chamber control system based on multi-sensor data fusion, used to implement the method as described in any one of claims 1-8, characterized in that, The system includes: The temperature field partitioning unit is used to collect real-time temperature data of multiple areas inside the moxibustion heat box through a sensor array, preprocess the real-time temperature data to construct an initial temperature field mapping, and divide the thermal conductivity difference area according to the initial temperature field mapping to determine the thermal conductivity non-uniform partition. The error feature quantization unit is used to extract temperature gradient values ​​from the thermally uneven partition, determine the temperature abnormal area, and generate error quantization index and corresponding spatial distribution characteristics and temporal evolution characteristics. The heating strategy solving unit is used to input the spatial distribution characteristics and temporal evolution characteristics of the error quantification index into a pre-trained prediction model to predict the material aging trend and determine the aging influence coefficient; and to construct an optimization objective function based on the aging influence coefficient and the current temperature data to solve for the adjusted heating strategy. The strategy evaluation and iteration unit is used to apply the adjusted heating strategy, collect temperature data after execution through the sensor array, update the temperature field distribution data, calculate the uniformity index of the updated temperature field distribution data, and iteratively optimize the heating strategy based on the uniformity index. The temperature field monitoring unit is used to continuously monitor the operating status of the moxibustion chamber through an optimized heating strategy and generate a stable temperature field record.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the method as described in any one of claims 1-8.

Images