Fuzzy logic self-adaptive adjusting system of box-type temperature control equipment

Abstract

The invention discloses a fuzzy logic self-adaptive adjusting system of box-type temperature control equipment, and relates to the field of fuzzy logic self-adaptive adjustment, and the fuzzy logic self-adaptive adjusting system comprises a feature extraction module, a mapping module, an activation module, a control module and an adjusting module. The method comprises the steps of obtaining a temperature field, energy consumption and airflow velocity, establishing a thermodynamic vortex field model and performing feature extraction analysis, obtaining a curvature phase energy third-order feature tensor, performing high-order singular value decomposition, constructing a five-dimensional manifold space, and obtaining a final geometric invariant curvature membership degree and a learnable parameter tensor integral solution through a reaction diffusion equation and minimum projection. The method comprises the following steps: constructing a high-dimensional rule network, obtaining a rotation invariant rule activation tensor through quantum entropy screening and nonlinear activation, constructing a fractional order model, obtaining a final control signal through potential energy well resonance and coherence optimization analysis, constructing a parameter update vector through a key coefficient, and adjusting the final control signal in real time.

Description

Technical Field

[0001] The present invention relates to the field of fuzzy logic self-adaptive regulation, in particular to a fuzzy logic self-adaptive regulation system for box-type temperature control equipment. Background Art

[0002] Box-type temperature control equipment is widely used in industrial control, precision manufacturing and other scenarios. Its temperature control accuracy and energy efficiency directly affect the stability and operating costs of the equipment. Traditional control methods (such as PID control) are limited by the assumption of linear models and are difficult to effectively handle complex physical processes such as nonlinear coupling of thermal flow fields, dynamic interaction of airflow curl and temperature gradient. They have defects such as slow thermal equilibrium, large temperature fluctuations, parameter tuning and fixed rule base relying on manual experience. Especially under the dynamic coupling conditions of multiple physical fields, problems such as local overheating and energy consumption surge are prone to occur. Specific defects of the fuzzy logic adaptive adjustment system of existing box-type temperature control equipment:

[0003] 1. Insufficient multi-physics field coupling modeling capabilities: Traditional methods rely solely on single sensor data (such as single-point temperature), ignoring the nonlinear coupling effects of temperature fields, airflow fields, and energy consumption. They are unable to capture the rotational interaction between temperature gradients and airflow velocities, resulting in inaccurate quantification of heat exchange efficiency and difficulty in warning of local overheating risks.

[0004] 2. Lack of high-order feature extraction and rule adaptation: The lack of geometric analysis of the vortex topology makes it impossible to construct a manifold space to describe the thermal convection configuration; the fuzzy rule base is fixed and relies on manual presets, making it difficult to dynamically evolve according to real-time energy consumption fluctuations, and the control strategy lags behind changes in operating conditions.

[0005] 3. Limitations of linear control strategies: Linear control algorithms such as PID cannot cope with problems such as heat conduction lag and nonlinear response of fan speed. Thermal equilibrium time is long and temperature fluctuations are large. At the same time, there is a lack of nonlinear mechanisms such as resonant drive to optimize energy consumption, resulting in low power distribution efficiency.

[0006] 4. Parameter tuning relies on experience: Key parameters require manual debugging, and adaptive updates cannot be achieved based on performance indicators such as temperature non-uniformity and control offset peak. The system is not robust enough in scenarios such as equipment aging and changes in material thermal capacity.

[0007] In order to solve the above-mentioned defects, a technical solution is now provided. Summary of the Invention

[0008] In order to solve the technical problems raised by the above background technology, the present invention is proposed. An embodiment of the present invention provides a fuzzy logic adaptive adjustment system for a box-type temperature control device.

[0009] The purpose of the present invention can be achieved through the following technical solutions: A fuzzy logic adaptive adjustment system for box-type temperature control equipment, including a feature extraction module, a mapping module, an activation module, a control module and an adjustment module,

[0010] The feature extraction module measures the spatial distribution, energy consumption, and flow field dynamics of the box-type temperature control equipment to obtain the temperature field, energy consumption, and airflow velocity. Based on the temperature field, energy consumption, and airflow velocity, a thermodynamic vortex field model is established and feature extraction analysis is performed to obtain the third-order feature tensor of curvature phase energy.

[0011] The mapping module performs high-order singular value decomposition on the third-order eigentensor of curvature phase energy, constructs a five-dimensional manifold space, calculates geometric quantities, and obtains the final geometrically invariant curvature membership through reaction-diffusion equations and minimization projection;

[0012] The activation module constructs a high-dimensional regular network based on the final geometrically invariant curvature membership and the learnable parameter tensor integral decomposition. After quantum entropy screening, nonlinear activation and orthogonal fractal projection, a rotationally invariant regular activation tensor is obtained.

[0013] The control module constructs a fractional-order model based on the rotationally invariant rule activation tensor, superimposes heat conduction correction, and obtains the final control signal through potential well resonance and coherent optimization analysis;

[0014] The regulation module constructs a parameter update vector based on the key coefficients in the feature extraction module, mapping module and control module. It calculates the gradient of the performance index and the target deviation, combines the learning rate to dynamically optimize the parameters, and adjusts the final control signal in real time.

[0015] Furthermore, the steps of analyzing the third-order characteristic tensor of curvature phase energy are as follows:

[0016] Based on the dot product calculation of the vortex field vector and the temperature gradient, the directional correlation information of the temperature gradient is obtained. The scalar phase field of the vortex topology structure and the directional correlation information of the temperature gradient are combined and the characteristic wavelength parameter function is used to calculate the complex phase field.

[0017] The complex phase field is converted to frequency space through Fourier transform, and the real-time energy consumption is nonlinearly transformed using the hyperbolic tangent function. The spatial modulation field with integrated energy consumption characteristics is obtained by element-by-element multiplication and inverse Fourier transform reconstruction.

[0018] The real part of the spatial modulation field of the fusion energy consumption characteristics is used to capture the phase curvature characteristics, and the imaginary part is used to extract the phase gradient change, which is multiplied with the original vortex field modulus to obtain the third-order characteristic tensor of the curvature phase energy.

[0019] Furthermore, the scalar phase field analysis steps of the vortex topology are as follows:

[0020] The temperature field of the box-type temperature control device is obtained by performing spatially distributed measurements on the box-type temperature control device through a thermocouple array; the energy consumption of the box-type temperature control device is obtained through a smart meter, and the flow field of the box-type temperature control device is dynamically monitored through a hot wire anemometer to obtain the airflow velocity;

[0021] Based on the temperature field, energy consumption and air flow velocity of the temperature control equipment, a thermodynamic vortex field model is established by multi-physics field coupling and nonlinear energy consumption modulation to obtain the vortex field vector.

[0022] Based on the vortex field vector, the divergence component and curl modulus component of the vortex field are calculated and combined into a complex form to extract the phase angle, and the scalar phase field of the vortex topological structure is obtained.

[0023] Furthermore, the final geometrically invariant curvature membership analysis steps are as follows:

[0024] Calculate the gradient based on the coordinates of the five-dimensional manifold space, perform a generalized cross product on the gradient, and obtain the normal vector;

[0025] The dot product operation of the normal vector and the second-order derivative of the coordinate space of the five-dimensional manifold gives the second fundamental form that characterizes the extrinsic curvature characteristics.

[0026] Based on the fusion of the first basic form and the second basic form representing the external curvature characteristics, the Gaussian curvature and the mean curvature are calculated, and the product of the Gaussian curvature and the mean curvature is negatively exponentially processed to obtain the basic membership function;

[0027] The Gaussian curvature gradient is processed by the inverse of the modulus length to obtain the diffusion time scale. The reaction diffusion equation of curvature-driven diffusion term and nonlinear reaction term is established in the five-dimensional manifold space coordinate. The diffusion field after spatiotemporal smoothing is obtained by taking the basic membership function as the initial condition.

[0028] Based on the diffusion field after spatiotemporal smoothing, the difference between the differential form of the current membership and the differential form of the diffusion field after spatiotemporal smoothing is calculated by minimizing the independent variable operator, and the final geometrically invariant curvature membership is obtained by projection.

[0029] Furthermore, the first basic form analysis steps are as follows:

[0030] Performing high-order singular value decomposition on the third-order eigentensor of curvature phase energy to obtain five main eigencomponents and their corresponding spatial distribution vectors, and defining five-dimensional manifold space coordinates for the five main eigencomponents and their corresponding spatial distribution vectors;

[0031] Based on the five-dimensional manifold space coordinates, the rate of change in the physical space is calculated to obtain the manifold coordinate Jacobian matrix. The intrinsic geometric structure metric tensor of the manifold is obtained by summing the products of the third-order eigentensor of the curvature phase energy and the elements of the manifold coordinate Jacobian matrix, which is marked as the first fundamental form.

[0032] Furthermore, the steps of the rotation invariant rule activation tensor analysis are as follows:

[0033] Based on the entanglement entropy, the preset entropy threshold is compared and screened, and the rule components greater than the threshold are retained to obtain a simplified core rule network;

[0034] The activation function is used to compress the rule strength value range of the simplified core rule network, and the hyperbolic tangent function is used to calculate the interaction strength of the transposed product between rules. Then, the full permutation direction sign tensor is combined to introduce spatial sensitivity to obtain the activation tensor field of nonlinear interaction features.

[0035] For the activation tensor field of nonlinear interaction features, the rotation-invariant regular activation tensor is obtained by minimizing the projection transformation between the current tensor and the orthogonal shape while maintaining the orthogonality of the regular components.

[0036] Furthermore, the entanglement entropy analysis steps are as follows:

[0037] The final geometrically invariant curvature membership features are multiplied by tensors with learnable parameters, and then subjected to singular value decomposition to obtain three sets of basic rule core tensors. The three sets of basic rule core tensors are paired and connected using the Kronecker function to constrain the sum of indices to zero, resulting in a high-dimensional ring-interconnected initial rule network.

[0038] The quantum state information density matrix is ​​obtained by multiplying each regular component in the high-dimensional ring-interconnected initial regular network with its conjugate transpose. The entanglement entropy is obtained by combining the matrix trace operation and the natural logarithm function on the quantum state information density matrix.

[0039] Furthermore, the final control signal analysis steps are as follows:

[0040] Based on the calculation of the steady-state point position and the initial potential well area, the dynamic noise strategy is obtained, and the dynamic equation of adaptive noise driving is established based on the dynamic noise intensity and the steady-state point position;

[0041] Based on the initial state injection and the starting potential well state of the system, a discrete iterative formula is constructed, and the time step is adaptively adjusted according to the distance between the current state and the steady-state point. The dynamic equation driven by adaptive noise is numerically solved to obtain the numerical sequence of the system state.

[0042] Based on the previous and next states in the numerical sequence of the system state, the transition direction is determined and the moment of crossing the critical point of phase transition in the steady-state point is calculated through dynamic threshold and linear interpolation to obtain the time point of resonance occurrence;

[0043] The transition starting point and end point are determined based on the three steady-state points of the system temperature control. The original state sequence is extracted based on the time point when the resonance occurs. The steady-state positions of the transition starting point and end point are used as the benchmark, and energy calibration is performed in combination with the potential energy difference. The resonance control signal is obtained by superimposing the target control requirements.

[0044] The final control signal is obtained by coherently optimizing the function based on the resonance control signal.

[0045] Furthermore, the initial potential well region analysis steps are as follows:

[0046] A fractional-order history aggregation model is constructed based on the rotation-invariant rule activation tensor. The trace information of the rotation-invariant rule activation tensor is time-weighted accumulated through the Riemann-Liouville integral formula. The initial control signal of the long-term memory characteristic is obtained by using the fractional-order decay fusion of the gamma function history control.

[0047] Applying fractional time differential operation to the initial control signal of the long-term memory characteristic and multiplying it by the Laplace operator modulus length of the current temperature field to obtain the hysteresis effect correction value, and adding the hysteresis effect correction value to the initial control value to obtain the modified heat conduction compensation control;

[0048] The energy state potential energy function of the box-type temperature control system is defined, and the three steady-state points of the system temperature control are obtained by solving it. The modified heat conduction compensation control is injected as the initial state of the system, and the position relationship with the three steady-state points of the system temperature control is determined to obtain the initial potential well area.

[0049] Furthermore, the final control signal analysis steps are as follows:

[0050] Based on the dynamic heat-fluid coupling coefficient in the thermodynamic vortex field model, the nonlinear coupling coefficient in the reaction-diffusion equation of the curvature-driven diffusion term and the nonlinear reaction term, and the memory decay exponent in the fractional-order history aggregation model, the adjustment values ​​of the three parameters are defined and the parameter update vector is obtained.

[0051] The square of the cumulative temperature gradient is calculated based on the temperature field of the temperature control device. Based on the final control signal, the maximum absolute value of the deviation between the final control signal and the standard control signal is calculated to obtain the control signal offset peak value. The number of resonance events per unit time of the resonance control signal is obtained and sequentially integrated to obtain the system performance vector.

[0052] Set the target performance vector. The first two items are the cumulative temperature non-uniformity and the control signal offset peak, which are set to zero target values. The third item is to obtain the number of resonance events corresponding to the optimal resonance frequency determined by the heat capacity of the material in the database. Apply a small disturbance to each parameter update vector, calculate the difference between the performance vector before and after the disturbance and divide it by the disturbance amplitude, and obtain the derivative of the performance with respect to the parameter update vector. Use the unit matrix and the projection term constructed by the parameter update vector to multiply the derivative with the deviation of the target performance vector and the system performance vector, the projection term and the learning rate to obtain the parameter update vector, and dynamically adjust the final control signal in real time.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] 1. The present invention uses a feature extraction module to perform spatial distribution, energy consumption, and flow field dynamic measurements on the box-type temperature control equipment to obtain the temperature field, energy consumption, and airflow velocity. Based on the temperature field, energy consumption, and airflow velocity, a thermodynamic vortex field model and feature extraction analysis are established to obtain the third-order characteristic tensor of the curvature phase energy. The third-order characteristic tensor of the curvature phase energy is subjected to high-order singular value decomposition to construct a five-dimensional manifold space, calculate the geometric quantity, and obtain the final geometrically invariant curvature membership through the reaction-diffusion equation and the minimization projection. Based on the final geometrically invariant curvature membership and the learnable parameter tensor integral decomposition, a high-dimensional rule network is constructed. After quantum entropy screening, nonlinear activation, and orthogonal manifold projection, the high-dimensional rule network is obtained. Shadow, obtain the rotation invariant rule activation tensor, build a fractional order model based on the rotation invariant rule activation tensor, superimpose heat conduction correction, and obtain the final control signal through potential well resonance and coherent optimization analysis. Comprehensively consider the nonlinear coupling effect of temperature field, airflow field and energy consumption, and capture the rotational interaction between temperature gradient and airflow velocity, improve the heat exchange efficiency quantification, and warn of local overheating risks. It can perform geometric analysis of vortex topological structures and construct a manifold space to describe the thermal convection configuration. The fuzzy rule base can dynamically evolve according to real-time energy consumption fluctuations, can deal with problems such as heat conduction lag and nonlinear response of fan speed, and can also optimize energy consumption through nonlinear mechanisms such as resonance drive, and has high power distribution efficiency.

[0055] 2. The present invention constructs a parameter update vector based on feature extraction of key coefficients, calculates the gradient of performance indicators and target deviations, and dynamically optimizes parameters in combination with the learning rate to adjust the final control signal in real time. Key parameters can be adaptively updated based on performance indicators such as temperature non-uniformity and control offset peak value, and the system's robustness is improved in scenarios such as equipment aging and material heat capacity changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. The following drawings are not intentionally scaled to the actual size, and the focus is on illustrating the main purpose of the present invention.

[0057] Figure 1 is a system block diagram of the present invention;

[0058] Figure 2 It is the vortex field vector analysis flow chart of the present invention;

[0059] Figure 3 This is a flow chart of feature extraction module analysis of the present invention. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts also fall within the scope of protection of the present invention.

[0061] like Figure 1 As shown, a fuzzy logic adaptive adjustment system for box-type temperature control equipment includes a feature extraction module, a mapping module, an activation module, a control module and an adjustment module.

[0062] The feature extraction module measures the spatial distribution, energy consumption, and flow field dynamics of the box-type temperature control equipment to obtain the temperature field, energy consumption, and airflow velocity. Based on the temperature field, energy consumption, and airflow velocity, a thermodynamic vortex field model is established and feature extraction analysis is performed to obtain the third-order feature tensor of curvature phase energy.

[0063] The mapping module performs high-order singular value decomposition on the third-order eigentensor of curvature phase energy, constructs a five-dimensional manifold space, calculates geometric quantities, and obtains the final geometrically invariant curvature membership through reaction-diffusion equations and minimization projection;

[0064] The activation module constructs a high-dimensional regular network based on the final geometrically invariant curvature membership and the learnable parameter tensor integral decomposition. After quantum entropy screening, nonlinear activation and orthogonal fractal projection, a rotationally invariant regular activation tensor is obtained.

[0065] The control module constructs a fractional-order model based on the rotationally invariant rule activation tensor, superimposes heat conduction correction, and obtains the final control signal through potential well resonance and coherent optimization analysis;

[0066] The regulation module constructs a parameter update vector based on the key coefficients in the feature extraction module, mapping module and control module. It calculates the gradient of the performance index and the target deviation, combines the learning rate to dynamically optimize the parameters, and adjusts the final control signal in real time.

[0067] like Figure 3 As shown, specifically, the feature extraction module is analyzed as follows:

[0068] The temperature field of the box-type temperature control equipment is obtained by performing spatially distributed measurement of the box-type temperature control equipment through a thermocouple array. in represents the three-dimensional space coordinates, Where x, y and z represent the length, width and height coordinates of the equipment, and t represents the time coordinate. The energy consumption P(t) of the box-type temperature control equipment is obtained through the smart meter, and the flow field of the box-type temperature control equipment is dynamically monitored through the hot wire anemometer to obtain the airflow velocity. Specific represents the three-dimensional velocity vector;

[0069] like Figure 2 As shown, the temperature field based on the temperature control equipment Energy consumption P(t) and air flow velocity Multi-physics field coupling and nonlinear energy consumption modulation are performed to establish a thermodynamic vortex field model and obtain the vortex field vector. The specific thermodynamic vortex field model includes:

[0070] Where exp represents the natural exponential function, represents the vortex field vector, P no Indicates the rated power of the equipment, P ref represents the reference power, Calculate the vector field rotation, χ represents the dynamic heat-fluid coupling coefficient, a dynamically adjustable parameter, σP represents the standard deviation of energy consumption, represents the temperature gradient, P avg represents the historical average energy consumption, Γ(*) represents the real-time energy consumption modulation function, represents the time partial derivative operator, δ represents the dynamic response coefficient, transient process gain, and exp represents the exponential function with the natural constant e as the base;

[0071] Specifically, the real-time energy consumption modulation function Γ(*) is:

[0072] ξ represents the energy consumption deviation standardization parameter, u is the integral dummy variable, Gaussian distribution integral variable, is the Gaussian kernel function;

[0073] Specifically, the dynamic thermal-fluid coupling coefficient and the dynamic response coefficient are:

[0074]

[0075] Where tanh represents the hyperbolic tangent function, P crit represents the critical power, ρ represents the air density, k represents the safety factor, c ρrepresents the constant pressure specific heat capacity, V represents the constant pressure specific heat capacity, Δt represents the control period, Indicates the maximum airflow velocity, represents the Gaussian suppression function;

[0076] Specifically, by integrating the spatial temperature distribution measurement of thermocouple arrays, the real-time energy consumption monitoring of smart meters, and the three-dimensional airflow velocity acquisition of hot-wire anemometers, a multi-physics field coupling data foundation is constructed. Based on this, a thermodynamic vortex field model is established. With nonlinear differential equations as the mathematical core, the rotational coupling effect between the temperature gradient field and the airflow velocity field is accurately captured through the curl operator. The real-time energy consumption modulation function uses the standard normal cumulative distribution to probabilistically process energy consumption deviations. The dynamic heat flow coupling coefficient introduces a hyperbolic tangent function to achieve power saturation constraint, and the dynamic response coefficient uses a Gaussian suppression function to optimize the low-power state. This vortex field vector is input into the fuzzy logic adaptive adjustment system as a high-order physical feature. Its curl intensity quantifies the heat exchange efficiency, the divergence distribution reveals the risk of local overheating, and the kinetic energy distribution maps the thermal convection efficiency. It drives the rule base to dynamically evolve according to the real-time thermodynamic state and energy consumption level. Through a multi-objective decision-making mechanism, it collaboratively optimizes the temperature distribution uniformity, fan speed matching and power allocation efficiency, significantly improving the thermal balance speed and suppressing temperature fluctuations in closed-loop control. At the same time, overload protection is implemented based on the critical power mechanism, ultimately reducing the standard deviation of the internal temperature of the box to the set threshold, achieving intelligent coordination between temperature control accuracy and energy efficiency.

[0077] Based on the vortex field vector, the divergence component and curl modulus component of the vortex field are calculated and combined into a complex form to extract the phase angle and obtain the scalar phase field of the vortex topological structure;

[0078] Based on the dot product calculation of the vortex field vector and the temperature gradient, the directional correlation information of the temperature gradient is obtained. The scalar phase field of the vortex topology structure and the directional correlation information of the temperature gradient are combined and the characteristic wavelength parameter function is used to calculate the complex phase field.

[0079] The complex phase field is converted to frequency space through Fourier transform, and the real-time energy consumption is nonlinearly transformed using the hyperbolic tangent function. The spatial modulation field with integrated energy consumption characteristics is obtained by element-by-element multiplication and inverse Fourier transform reconstruction.

[0080] The phase curvature characteristics of the real part of the spatial modulation field of the fusion energy consumption characteristics are captured, and the phase gradient change is extracted from the imaginary part. The result is multiplied by the original vortex field modulus to obtain the third-order characteristic tensor of the curvature phase energy.

[0081] Specifically, the spiral phase encoding is performed through the vortex field vector vortex field, and the scalar phase field of the vortex topology structure is obtained by extracting the phase angle function. Where arg represents the complex argument (phase angle) function, Calculate the divergence characteristics of the vector field, i represents the imaginary unit, represents the modulus of the vortex field curl, Represents the divergence component of the vortex field. The scalar phase field of the vortex topology and the direction correlation information of the temperature gradient are combined through the characteristic wavelength parameter function to obtain the complex phase field, where the characteristic wavelength parameter function is

[0082] Where S0 represents the complex phase field, f represents the wave number, f = 2π / λ, λ is the wavelength, e i* represents the complex exponential function (Euler's formula), sign represents the sign function, represents the dot product of the vortex field and the temperature gradient;

[0083] Based on the complex phase field and real-time energy consumption, the spatial modulation field with integrated energy consumption characteristics is obtained through element-wise multiplication and inverse Fourier transform reconstruction. Where F(*) represents the Fourier transform operator, F -1 represents the inverse Fourier transform operator, S1 represents the modulation field, and tanh represents the hyperbolic tangent function (nonlinear modulation);

[0084] The spatial modulation field of the fusion energy consumption characteristics is multiplied by the first-order spatial partial derivative and the second-order spatial partial derivative with the original vortex field modulus to obtain the third-order characteristic tensor of the curvature phase energy. Specifically,

[0085] Among them S ijk represents the third-order eigentensor of curvature phase energy, represents the real part of the modulation field S1, represents the imaginary part of the modulated field S1, Represents the second-order spatial partial derivative, extracting the phase gradient change, Represents the first-order spatial partial derivative, extracting the phase gradient change, represents the original vortex field modulus, and i, j, k represent the spatial coordinate index.

[0086] Specifically, the mapping module is analyzed as follows:

[0087] Performing high-order singular value decomposition on the third-order eigentensor of curvature phase energy to obtain five main eigencomponents and their corresponding spatial distribution vectors, and defining five-dimensional manifold space coordinates for the five main eigencomponents and their corresponding spatial distribution vectors;

[0088] Based on the coordinates of the five-dimensional manifold space, the rate of change in the physical space is calculated to obtain the manifold coordinate Jacobian matrix. The intrinsic geometric structure metric tensor of the manifold is obtained by summing the products of the third-order eigentensor of the curvature phase energy and the elements of the manifold coordinate Jacobian matrix, which is marked as the first fundamental form.

[0089] Calculate the gradient based on the coordinates of the five-dimensional manifold space, perform a generalized cross product on the gradient, and obtain the normal vector;

[0090] The dot product operation of the normal vector and the second-order derivative of the coordinate space of the five-dimensional manifold gives the second fundamental form that characterizes the extrinsic curvature characteristics.

[0091] Based on the fusion of the first basic form and the second basic form representing the external curvature characteristics, the Gaussian curvature and the mean curvature are calculated, and the product of the Gaussian curvature and the mean curvature is negatively exponentially processed to obtain the basic membership function;

[0092] The Gaussian curvature gradient is processed by the inverse of the modulus length to obtain the diffusion time scale. The reaction diffusion equation of curvature-driven diffusion term and nonlinear reaction term is established in the five-dimensional manifold space coordinate. The diffusion field after spatiotemporal smoothing is obtained by taking the basic membership function as the initial condition.

[0093] Based on the diffusion field after spatiotemporal smoothing, the difference between the differential form of the current membership and the differential form of the diffusion field after spatiotemporal smoothing is calculated by minimizing the independent variable operator, and the final geometrically invariant curvature membership is obtained by projection.

[0094] Specifically, the HOSVD high-order singular value decomposition is performed on the third-order eigentensor of the curvature phase energy. where σ n represents the nth singular value (feature weight), U n (i) represents the component of the nth left singular vector at position i, V n (j) represents the component of the nth right singular vector at position j, W n (k) represents the component of the nth right singular vector at position k, and n represents the feature index (range 1-5);

[0095] Specifically, high-order singular value decomposition (HOSVD), as the core method of tensor decomposition, expands the third-order eigentensor of the curvature phase energy into matrix modes along three spatial dimensions. Singular value decomposition is then applied to each mode matrix to extract its dominant eigenvectors and singular value spectra, thereby decomposing the original tensor into the product of a set of orthogonal eigenmatrices and a core tensor. This decomposition extracts three key physical information from the third-order eigentensor of the curvature phase energy: three orthogonal eigenmatrices characterize the dominant spatial vibration modes of the temperature field, airflow field, and energy consumption field, respectively, while the core tensor quantifies the strength of the nonlinear coupling between different modes. By truncating low singular value components, a compact representation of high-dimensional data is achieved, eliminating measurement noise interference. Simultaneously, the eigenspace modes of the heat-flow-energy coupling system are accurately extracted, providing coordinate basis vectors with clear physical meaning for the subsequent construction of a five-dimensional characteristic manifold, enabling the system to identify the spatial configuration of thermal convection vortices, energy transfer paths, and abnormal dissipation regions.

[0096] The five-dimensional manifold space coordinates ξ are defined for the five main characteristic components and their corresponding spatial distribution vectors n =[U n , V n , W n ] T ,ξ n Represents the nth coordinate of the 5-dimensional feature manifold, T represents the transpose operator, and the rate of change in the physical space is calculated based on the coordinates of the five-dimensional manifold space to obtain the manifold coordinate Jacobian matrix. in Represents the partial derivative of the manifold space coordinates with respect to the physical coordinates, which is a 3×3 matrix, the Jacobian matrix, with row components U n ,V n ,W n , column: physical coordinate x i , represents the physical space position vector, represents the displacement increment in the i direction, Represents the modulus of the displacement increment; Based on the sum of the elements of the manifold coordinate Jacobian matrix and the third-order characteristic tensor of the curvature phase energy, the intrinsic geometric structure metric tensor of the manifold is obtained. Where n and m represent the manifold coordinate index (values ​​range from 1 to 5), h nm Represents the metric tensor of the intrinsic geometric structure of the manifold, the first basic form, calculates the gradient based on the coordinates of the five-dimensional manifold space, performs a generalized cross product on the gradient, and obtains the normal vector, represents the normal vector of the 5-manifold, Δξ1, ..., Δξ5 represent the gradient of the manifold space coordinates, and × represents the high-dimensional cross product operator (wedge product). The dot product operation of the normal vector and the second-order derivative of the 5-manifold space coordinates gives the second fundamental form that represents the extrinsic curvature characteristics. Π nm represents the second fundamental form characterizing the extrinsic bending properties, represents the manifold space position vector [ξ1,ξ2,...,ξ5], represents the partial derivative with respect to the m-th manifold coordinate, and represents the dot product operator;

[0097] Calculate Gaussian curvature based on the fusion of the first fundamental form and the second fundamental form that characterizes the external curvature characteristics Where det represents the determinant operator and I represents the first fundamental form h nm , and the mean curvature Its tr represents the matrix trace operator, Ι -1 Represents the inverse matrix, multiplying the Gaussian curvature and the mean curvature by the negative exponential to obtain the basic membership function e represents a natural constant, |*| represents an absolute value operator;

[0098] The diffusion time scale is obtained by performing the inverse modulus processing on the Gaussian curvature gradient in Represents the gradient of Gaussian curvature, ||*|| represents the modulus of the gradient vector, and establishes the reaction diffusion equation of curvature-driven diffusion term and nonlinear reaction term in the five-dimensional manifold space coordinate. Taking the basic membership function as the initial condition, the diffusion field after spatiotemporal smoothing is obtained. Specifically:

[0099] Where μ represents the time-varying membership field, t represents the time variable, represents the Laplace-Beltrami operator on the manifold, κ represents the nonlinear coupling coefficient, u(·, 0) represents the diffusion result obtained by taking the basic membership function u0 as the initial condition: the diffusion field u after spatiotemporal smoothing diff =u(·,t end ), represents the spatial position placeholder. Based on the diffusion field after spatiotemporal smoothing, the difference between the current membership differential form and the diffusion field differential form after spatiotemporal smoothing is calculated by minimizing the independent variable operator, and the final geometrically invariant curvature membership is obtained by projection. where u k represents the final geometrically invariant curvature membership, argmin represents the solution to the minimization problem, ||*|| represents the norm of the function space, represents the gradient of the logarithm of membership, c represents the conformal constraint constant, and st represents the abbreviation of the constraint condition.

[0100] Specifically, the activation module is analyzed as follows:

[0101] The final geometrically invariant curvature membership features are subjected to tensor product operations with learnable parameters, and three sets of basic rule core tensors are obtained through singular value decomposition. The three sets of basic rule core tensors are paired and connected using the Kronecker delta function to constrain the index sum to zero, resulting in a high-dimensional ring-interconnected initial rule network.

[0102] The quantum state information density matrix is ​​obtained by multiplying the regular components in the high-dimensional ring-connected initial regular network with their conjugate transpose. The entanglement entropy is obtained by combining the matrix trace operation and the natural logarithm function on the quantum state information density matrix.

[0103] Compare and filter based on the preset entropy threshold, retain the rule components that are greater than the threshold, and obtain a simplified core rule network;

[0104] The activation function is used to compress the rule strength value range of the simplified core rule network, and the hyperbolic tangent function is used to calculate the interaction strength of the transposed product between rules. Then, the full permutation direction sign tensor is combined to introduce spatial sensitivity to obtain the activation tensor field of nonlinear interaction features.

[0105] For the activation tensor field of nonlinear interaction features, the rotation-invariant regular activation tensor is obtained by minimizing the projection transformation between the current tensor and the orthogonal shape while maintaining the orthogonality of the regular components.

[0106] Specifically, the final geometric invariant curvature membership feature and the learnable parameter tensor product operation are subjected to singular value decomposition to obtain three sets of basic rule core tensors: Among them, β, γ represents the tensor ring index, A β 、 C γ Represents the basic rule core tensor, W β 、 W γ represents the learnable parameters, represents the tensor product operator, SVD represents the singular value decomposition, and the three sets of basic rule core tensors are paired and connected using the Kronecker delta function to constrain the index sum to zero, thus obtaining a high-dimensional ring-interconnected initial rule network. Where R represents a high-dimensional ring-connected initial regular network, is the Kronecker delta function, when 1 when it is, otherwise 0;

[0107] The quantum state information density matrix is ​​obtained by multiplying each regular component in the high-dimensional ring-connected initial regular network with its conjugate transpose. R i represents the regular component of the high-dimensional ring-connected initial regular network, i represents the number of the regular component, R Θ i represents the conjugate transpose of the regular component, ρ i Represents the quantum state information density matrix, and performs matrix trace operation and natural logarithm function combination on the quantum state information density matrix to obtain the entanglement entropy S i =-tr(ρ i lnρ i ), where tr represents the matrix trace operation, ln represents the entropy calculation function, and the pre-set entropy threshold is compared and screened, and the rule components greater than the threshold are retained to obtain the simplified core rule network R pru ={R i |S i >Sth}, where Sth represents the entropy threshold. By borrowing the density matrix and von Neumann mathematical framework from quantum information theory, the local regular components of high-dimensional ring networks are converted into a quantifiable "structural complexity" indicator. High-entropy substructures are screened based on a preset entropy threshold, thereby eliminating redundant parts and retaining the core associated components of the network. Ultimately, a streamlined core regular network is output, providing an information-intensive underlying architecture for subsequent network reconstruction or optimization.

[0108] The activation function is used to compress the rule strength value range of the simplified core rule network, and the hyperbolic tangent function is used to calculate the interaction strength of the transposed product between rules. Then, the full permutation direction symbol tensor is combined to introduce spatial sensitivity to obtain the activation tensor field of nonlinear interaction features. Where i, j and k represent rule indices, component identifiers in the simplified core rule network, σ represents the Sigmoid function, tanh represents the hyperbolic tangent function, and ε ijk Represents the fully permuted direction symbol tensor, R i 、R j 、R k Represents the regular component of the simplified core regular network, and obtains the rotation-invariant regular activation tensor by minimizing the projection transformation of the current tensor and the orthogonal shape while maintaining the orthogonality of the regular component for the activation tensor field of the nonlinear interaction feature. Where G represents a mathematical optimization variable, representing any candidate solution in the orthogonal shape, used to search for the solution closest to R per The orthogonal tensor, G T G=I represents orthogonal constraint, I represents orthogonal constraint, ||*|| F represents the Frobenius norm, i, j and k represent the three-dimensional coordinates of the corresponding device, in the length, width and height directions. When the rule is in an even arrangement, the interaction is enhanced by +1, and when the rule is in an odd arrangement, the interaction is suppressed by -1.

[0109] Specifically, the control module is analyzed as follows:

[0110] A fractional-order history aggregation model is constructed based on the rotation-invariant rule activation tensor. The trace information of the rotation-invariant rule activation tensor is time-weighted accumulated through the Riemann-Liouville integral formula. The initial control signal of the long-term memory characteristic is obtained by using the fractional-order decay fusion of the gamma function history control.

[0111] Applying fractional time differential operation to the initial control signal of the long-term memory characteristic and multiplying it by the Laplace operator modulus length of the current temperature field to obtain the hysteresis effect correction value, and adding the hysteresis effect correction value to the initial control value to obtain the modified heat conduction compensation control;

[0112] Define the energy state potential energy function of the box-type temperature control system, solve for the three steady-state points of the system temperature control, inject the modified heat conduction compensation control as the system initial state, determine the position relationship with the three steady-state points of the system temperature control, and obtain the initial potential well area;

[0113] Based on the calculation of the steady-state point position and the initial potential well area, the dynamic noise strategy is obtained, and the dynamic equation of adaptive noise driving is established based on the dynamic noise intensity and the steady-state point position;

[0114] Based on the initial state injection and the starting potential well state of the system, a discrete iterative formula is constructed, and the time step is adaptively adjusted according to the distance between the current state and the steady-state point. The dynamic equation driven by adaptive noise is numerically solved to obtain the numerical sequence of the system state.

[0115] Based on the previous and next states in the numerical sequence of the system state, the transition direction is determined and the moment of crossing the critical point of phase transition in the steady-state point is calculated through dynamic threshold and linear interpolation to obtain the time point of resonance occurrence;

[0116] The transition starting point and end point are determined based on the three steady-state points of the system temperature control. The original state sequence is extracted based on the time point when the resonance occurs. The steady-state positions of the transition starting point and end point are used as the benchmark, and energy calibration is performed in combination with the potential energy difference. The resonance control signal is obtained by superimposing the target control requirements.

[0117] The final control signal is obtained by coherently optimizing the function based on the resonance control signal.

[0118] Specifically, a fractional-order history aggregation model is constructed based on the rotation-invariant rule activation tensor. The trace information of the rotation-invariant rule activation tensor is time-weighted accumulated through the Riemann-Liouville integral formula. The initial control signal of the long-term memory characteristic is obtained by using the fractional-order attenuation fusion of the gamma function history control. The specific fractional-order history aggregation model includes:

[0119] The above formula is called the Riemann-Liouville integral formula, where u0(t) represents the initial control signal of the long-term memory characteristic, Γ represents the gamma function, α represents the memory decay exponent, τ represents the integration time variable, tr(*) represents the regular tensor trace, and the main diagonal of the regular tensor is summed;

[0120] The fractional order time differential operation is applied to the initial control signal of the long-term memory characteristic and the product of the Laplace operator modulus length of the current temperature field to obtain the hysteresis effect correction amount, which is superimposed on the initial control amount to obtain the modified heat conduction compensation control. where u comp (t) represents the modified heat conduction compensation control, represents the fractional derivative, β represents the order, Represents the Laplace operator of the current temperature field and the second-order spatial derivative of the temperature field;

[0121] Define the energy state potential function of the box-type temperature control system Where u represents the system state variable, V(u) represents the energy state potential function of the box-type temperature control system, and the three steady-state points of the system temperature control are obtained by solving Where u1, u2, and u3 represent the low-temperature stable state (energy minimum point), the phase transition critical point (potential barrier vertex), and the high-temperature stable state (local energy minimum), respectively. The modified heat conduction compensation control is injected as the initial state of the system into u(0) = u comp , and the position relationship between the three steady-state points of the system temperature control is determined to obtain the initial potential well area

[0122] Where Region0 represents the initial potential well region;

[0123] Based on the calculation of the steady-state point position and the initial potential well area, the dynamic noise strategy is obtained Where σ(u) represents the dynamic noise strategy, K(Region0) represents the regional gain coefficient, σ0 represents the basic noise intensity. Based on the dynamic noise intensity and the steady-state point position, the adaptive noise-driven dynamic equation du = -V′(u)dt + σ(u)dWt is established, where du represents the differential of the control quantity, the instantaneous change, V′(u) represents the derivative of the potential energy function, and dWt represents the differential of the Wiener process, the Brownian motion increment.

[0124] Based on the system initial state injection and the starting potential well state, a discrete iterative formula is constructed Where k is the number of iterations, Δt is the time step, and ξ k Represents the random variable at the kth step, used to simulate the discretization of the Wiener process differential dWt, and adaptively adjust the time step according to the distance between the current state and the steady-state point Where |V′(u k )| is the absolute value of the potential energy derivative at step k, which reflects the “speed of deterministic change” of the system: the larger the absolute value, the more drastic the state change, and the step size needs to be shortened to ensure accuracy. min(|u k -u1|,|u k -u3|) is the minimum distance from the k-th step state to the two stable states u1 and u3. The closer the distance, the more stable the system is, and the step size can be increased to accelerate the calculation; the farther the distance is, the closer it is to the potential barrier u2, and the step size needs to be shortened to accurately capture the transition. The dynamic equation driven by adaptive noise is numerically solved to obtain the numerical sequence u of the system state k ;

[0125] The difference Δu between the previous and next states in the numerical sequence based on the system state k =|u k -u k-1 |, through the dynamic threshold if the state difference Δu k >Δu th Set threshold Δu th And the state symbol changes sign(u k )≠sign(u k-1), sign represents the sign function, which is used to judge the positive and negative signs of the state u, then it is determined that at time (t k-1 , t k ) a potential barrier is crossed, and the time of crossing the critical point of phase transition in the steady-state point is calculated by linear interpolation to obtain the time point of resonance occurrence.

[0126] The transition starting point and end point are determined based on the three steady-state points of the system temperature control. The original state sequence is extracted based on the time point of resonance. The steady-state positions of the transition starting point and end point are used as the benchmark, and energy calibration is performed in combination with the potential energy difference. The resonance control signal is obtained by superimposing the target control requirements. where u start and u end They represent the transition starting point and jump end point, respectively, where the transition starting point corresponds to the steady-state point u1 or u3, and the jump end point corresponds to the steady-state point u3 or u1, ΔE oot represents the potential energy difference, which refers to the absolute value of the potential energy difference between the starting point and the end point of the transition, u targ represents the target control quantity, u(t) represents the numerical sequence of the system state;

[0127] Based on the resonance control signal, the final control signal u is obtained through the coherent optimization function. fit (t)=arg min||δu(t)-u res (t)||stδ 2 u(r)=0, where δu(t) represents the upper edge operator, the boundary operator in algebraic topology, δ 2 u(t) represents the second-order upper edge, the harmonic form constraint, and st represents the constraint condition.

[0128] Specifically, the adjustment module is analyzed as follows:

[0129] Based on the dynamic heat-fluid coupling coefficient in the thermodynamic vortex field model, the nonlinear coupling coefficient in the reaction-diffusion equation of the curvature-driven diffusion term and the nonlinear reaction term, and the memory decay exponent in the fractional-order history aggregation model, the adjustment values ​​of the three parameters are defined, and the parameter update vector Δθ=[Δχ,Δκ,Δα] is obtained. T , where Δ represents the adjustment value. This vector takes these three parameters as elements and forms a unified parameter adjustment unit through structured arrangement, which is used for subsequent model control or optimization process;

[0130] Based on the temperature field of the temperature control device, the square of the cumulative temperature gradient is calculated (to quantify the temperature non-uniformity). Based on the final control signal, the maximum absolute value of the deviation between the final control signal and the standard control signal is calculated to obtain the control signal offset peak value. The number of resonance events per unit time of the resonance control signal is obtained. These three indicators are sequentially integrated to obtain the vector of system performance. where θ = [χ,κ,α] T , the current parameter vector, u opt Represents the standard control signal, ResonCount represents the number of resonance events per unit time obtained by the resonance control signal;

[0131] Set the target performance vector. The first two items are the cumulative temperature non-uniformity and the control signal offset peak, which are set to zero. The third item is to obtain the number of resonance events corresponding to the optimal resonance frequency determined by the heat capacity of the material in the database. Apply a small disturbance to each parameter update vector, calculate the difference between the performance vector before and after the disturbance and divide it by the disturbance amplitude, and obtain the derivative of the performance with respect to the parameter update vector. Use the unit matrix and the projection term constructed by the parameter update vector to multiply the derivative with the deviation of the target performance vector and the system performance vector, the projection term and the learning rate to obtain the parameter update vector, realize the adaptive adjustment of the parameters, and dynamically adjust the final control signal in real time. Among them, M targe represents the target performance vector, M targe =[0,0,f pot ] T , where f pot represents the number of resonance events corresponding to the optimal resonance frequency determined by the heat capacity of the material, and η represents the learning rate.

[0132] Specifically, multi-physical field data such as temperature field, energy consumption and air flow velocity are collected through thermocouple arrays, smart meters and hot wire anemometers to construct a thermodynamic vortex field model, extract the divergence and curl modulus components of the vortex field and combine them into a scalar phase field, and combine the temperature gradient direction correlation information to obtain a complex phase field. After Fourier transform and other processing, the curvature phase energy third-order characteristic tensor is generated; the tensor is then subjected to high-order singular value decomposition, the five-dimensional manifold space coordinates are defined, the Jacobian matrix, the first and second basic forms and other geometric quantities of the manifold are calculated, the basic membership function is constructed, and the diffusion field is obtained through the reaction-diffusion equation, and then the final geometrically invariant curvature membership is obtained by projection, which is then combined with the learnable parameters to construct a high-dimensional rule network. After screening and activation And after rotational invariance processing, the final control signal is obtained by combining the fractional-order history aggregation model, the resonant drive strategy and the adaptive parameter adjustment. The heat-flow coupling effect is accurately captured by the curl operator and the dynamic heat-flow coupling coefficient, and the thermal convection configuration and the abnormal dissipation area are identified by manifold geometry analysis, so that the fuzzy rule network can dynamically evolve according to the real-time state; the fractional-order control introduces long-term memory characteristics, the resonant drive strategy improves the thermal balance speed and suppresses temperature fluctuations, and the parameter adaptive adjustment is based on the performance vector optimization model parameters, and finally realizes the coordinated optimization of temperature distribution uniformity, fan speed matching and power allocation efficiency, reduces the thermal balance time, reduces energy consumption, reduces the temperature standard deviation to the set threshold, improves the temperature control accuracy and energy efficiency and enhances the robustness.

[0133] The above is an illustration of the present invention and should not be considered as limiting thereof. Although several exemplary embodiments of the present invention have been described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined by the claims. It should be understood that the above is an illustration of the present invention and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. A fuzzy logic adaptive adjustment system for box-type temperature control equipment, comprising a feature extraction module, a mapping module, an activation module, a control module and an adjustment module, characterized in that: The feature extraction module measures the spatial distribution, energy consumption, and flow field dynamics of the box-type temperature control equipment to obtain the temperature field, energy consumption, and airflow velocity. Based on the temperature field, energy consumption, and airflow velocity, a thermodynamic vortex field model is established and feature extraction analysis is performed to obtain the third-order feature tensor of curvature phase energy. The mapping module performs high-order singular value decomposition on the third-order eigentensor of curvature phase energy, constructs a five-dimensional manifold space, calculates geometric quantities, and obtains the final geometrically invariant curvature membership through reaction-diffusion equations and minimization projection; The activation module constructs a high-dimensional regular network based on the final geometrically invariant curvature membership and the learnable parameter tensor integral decomposition. After quantum entropy screening, nonlinear activation and orthogonal fractal projection, a rotationally invariant regular activation tensor is obtained. The control module constructs a fractional-order model based on the rotationally invariant rule activation tensor, superimposes heat conduction correction, and obtains the final control signal through potential well resonance and coherent optimization analysis; The regulation module constructs a parameter update vector based on the key coefficients in the feature extraction module, mapping module and control module. It calculates the gradient of the performance index and the target deviation, combines the learning rate to dynamically optimize the parameters, and adjusts the final control signal in real time.

2. The fuzzy logic adaptive adjustment system for box-type temperature control equipment according to claim 1 is characterized in that: The steps of analyzing the third-order characteristic tensor of curvature phase energy are as follows: Based on the dot product calculation of the vortex field vector and the temperature gradient, the directional correlation information of the temperature gradient is obtained. The scalar phase field of the vortex topology structure and the directional correlation information of the temperature gradient are combined and the characteristic wavelength parameter function is used to calculate the complex phase field. The complex phase field is converted to frequency space through Fourier transform, and the real-time energy consumption is nonlinearly transformed using the hyperbolic tangent function. The spatial modulation field with integrated energy consumption characteristics is obtained by element-by-element multiplication and inverse Fourier transform reconstruction. The real part of the spatial modulation field of the fusion energy consumption characteristics is used to capture the phase curvature characteristics, and the imaginary part is used to extract the phase gradient change, which is multiplied with the original vortex field modulus to obtain the third-order characteristic tensor of the curvature phase energy.

3. The fuzzy logic adaptive adjustment system for box-type temperature control equipment according to claim 2, characterized in that: The steps of scalar phase field analysis of the vortex topology are as follows: The temperature field of the box-type temperature control device is obtained by performing spatially distributed measurements on the box-type temperature control device through a thermocouple array; the energy consumption of the box-type temperature control device is obtained through a smart meter, and the flow field of the box-type temperature control device is dynamically monitored through a hot wire anemometer to obtain the airflow velocity; Based on the temperature field, energy consumption and air flow velocity of the temperature control equipment, a thermodynamic vortex field model is established by multi-physics field coupling and nonlinear energy consumption modulation to obtain the vortex field vector. Based on the vortex field vector, the divergence component and curl modulus component of the vortex field are calculated and combined into a complex form to extract the phase angle, and the scalar phase field of the vortex topological structure is obtained.

4. The fuzzy logic adaptive adjustment system for box-type temperature control equipment according to claim 1, characterized in that: The final geometrically invariant curvature membership analysis steps are as follows: Calculate the gradient based on the coordinates of the five-dimensional manifold space, perform a generalized cross product on the gradient, and obtain the normal vector; The dot product operation of the normal vector and the second-order derivative of the coordinate space of the five-dimensional manifold gives the second fundamental form that characterizes the extrinsic curvature characteristics. Based on the fusion of the first basic form and the second basic form representing the external curvature characteristics, the Gaussian curvature and the mean curvature are calculated, and the product of the Gaussian curvature and the mean curvature is negatively exponentially processed to obtain the basic membership function; The Gaussian curvature gradient is processed by the inverse of the modulus length to obtain the diffusion time scale. The reaction diffusion equation of curvature-driven diffusion term and nonlinear reaction term is established in the five-dimensional manifold space coordinate. The diffusion field after spatiotemporal smoothing is obtained by taking the basic membership function as the initial condition. Based on the diffusion field after spatiotemporal smoothing, the difference between the differential form of the current membership and the differential form of the diffusion field after spatiotemporal smoothing is calculated by minimizing the independent variable operator, and the final geometrically invariant curvature membership is obtained by projection.

5. The fuzzy logic adaptive adjustment system for box-type temperature control equipment according to claim 4, characterized in that: The first basic form analysis steps are as follows: Performing high-order singular value decomposition on the third-order eigentensor of curvature phase energy to obtain five main eigencomponents and their corresponding spatial distribution vectors, and defining five-dimensional manifold space coordinates for the five main eigencomponents and their corresponding spatial distribution vectors; Based on the five-dimensional manifold space coordinates, the rate of change in the physical space is calculated to obtain the manifold coordinate Jacobian matrix. The intrinsic geometric structure metric tensor of the manifold is obtained by summing the products of the third-order eigentensor of the curvature phase energy and the elements of the manifold coordinate Jacobian matrix, which is marked as the first fundamental form.

6. The fuzzy logic adaptive adjustment system for box-type temperature control equipment according to claim 1, characterized in that: The steps of analyzing the rotation invariance rule activation tensor are as follows: Based on the entanglement entropy, the preset entropy threshold is compared and screened, and the rule components greater than the threshold are retained to obtain a simplified core rule network; The activation function is used to compress the rule strength value range of the simplified core rule network, and the hyperbolic tangent function is used to calculate the interaction strength of the transposed product between rules. Then, the full permutation direction sign tensor is combined to introduce spatial sensitivity to obtain the activation tensor field of nonlinear interaction features. For the activation tensor field of nonlinear interaction features, the rotation-invariant regular activation tensor is obtained by minimizing the projection transformation between the current tensor and the orthogonal shape while maintaining the orthogonality of the regular components.

7. The fuzzy logic adaptive adjustment system for box-type temperature control equipment according to claim 6, characterized in that: The entanglement entropy analysis steps are as follows: The final geometrically invariant curvature membership features are multiplied by tensors with learnable parameters, and then subjected to singular value decomposition to obtain three sets of basic rule core tensors. The three sets of basic rule core tensors are paired and connected using the Kronecker function to constrain the sum of indices to zero, resulting in a high-dimensional ring-interconnected initial rule network. The quantum state information density matrix is ​​obtained by multiplying each regular component in the high-dimensional ring-interconnected initial regular network with its conjugate transpose. The entanglement entropy is obtained by combining the matrix trace operation and the natural logarithm function on the quantum state information density matrix.

8. The fuzzy logic adaptive adjustment system for box-type temperature control equipment according to claim 1, characterized in that: The final control signal analysis steps are as follows: Based on the calculation of the steady-state point position and the initial potential well area, the dynamic noise strategy is obtained, and the dynamic equation of adaptive noise driving is established based on the dynamic noise intensity and the steady-state point position; Based on the initial state injection and the starting potential well state of the system, a discrete iterative formula is constructed, and the time step is adaptively adjusted according to the distance between the current state and the steady-state point. The dynamic equation driven by adaptive noise is numerically solved to obtain the numerical sequence of the system state. Based on the previous and next states in the numerical sequence of the system state, the transition direction is determined and the moment of crossing the critical point of phase transition in the steady-state point is calculated through dynamic threshold and linear interpolation to obtain the time point of resonance occurrence; The transition starting point and end point are determined based on the three steady-state points of the system temperature control. The original state sequence is extracted based on the time point when the resonance occurs. The steady-state positions of the transition starting point and end point are used as the benchmark, and energy calibration is performed in combination with the potential energy difference. The resonance control signal is obtained by superimposing the target control requirements. The final control signal is obtained by coherently optimizing the function based on the resonance control signal.

9. The fuzzy logic adaptive adjustment system for box-type temperature control equipment according to claim 8, characterized in that: The initial potential well region analysis steps are as follows: A fractional-order history aggregation model is constructed based on the rotation-invariant rule activation tensor. The trace information of the rotation-invariant rule activation tensor is time-weighted accumulated through the Riemann-Liouville integral formula. The initial control signal of the long-term memory characteristic is obtained by using the fractional-order decay fusion of the gamma function history control. Applying fractional time differential operation to the initial control signal of the long-term memory characteristic and multiplying it by the Laplace operator modulus length of the current temperature field to obtain the hysteresis effect correction value, and adding the hysteresis effect correction value to the initial control value to obtain the modified heat conduction compensation control; The energy state potential energy function of the box-type temperature control system is defined, and the three steady-state points of the system temperature control are obtained by solving it. The modified heat conduction compensation control is injected as the initial state of the system, and the position relationship with the three steady-state points of the system temperature control is determined to obtain the initial potential well area.

10. The fuzzy logic adaptive adjustment system for box-type temperature control equipment according to claim 1, characterized in that: The final control signal analysis steps are as follows: Based on the dynamic heat-fluid coupling coefficient in the thermodynamic vortex field model, the nonlinear coupling coefficient in the reaction-diffusion equation of the curvature-driven diffusion term and the nonlinear reaction term, and the memory decay exponent in the fractional-order history aggregation model, the adjustment values ​​of the three parameters are defined and the parameter update vector is obtained. The square of the cumulative temperature gradient is calculated based on the temperature field of the temperature control device. Based on the final control signal, the maximum absolute value of the deviation between the final control signal and the standard control signal is calculated to obtain the control signal offset peak value. The number of resonance events per unit time of the resonance control signal is obtained and sequentially integrated to obtain the system performance vector. Set the target performance vector. The first two items are the cumulative temperature non-uniformity and the control signal offset peak, which are set to zero target values. The third item is to obtain the number of resonance events corresponding to the optimal resonance frequency determined by the heat capacity of the material in the database. Apply a small disturbance to each parameter update vector, calculate the difference between the performance vector before and after the disturbance and divide it by the disturbance amplitude, and obtain the derivative of the performance with respect to the parameter update vector. Use the unit matrix and the projection term constructed by the parameter update vector to multiply the derivative with the deviation of the target performance vector and the system performance vector, the projection term and the learning rate to obtain the parameter update vector, and dynamically adjust the final control signal in real time.

Images