Cooling and heating integrated heating table control system and control method

Abstract

The invention discloses a cooling and heating integrated heating table control system and method and a multi-mode coupling field actuator. The multi-mode coupling field actuator comprises a heating unit, a cooling unit, an adsorption layer and a thermal control layer. The heating unit comprises a near-infrared laser and a scanning galvanometer and irradiates a high-absorptivity photo-thermal coating arranged on the back face of the substrate, and rapid and local heating is achieved. A microporous electrostatic adsorption film with high insulativity and a microporous structure is arranged on the outer surface of the base plate, and the micro thermoelectric refrigerating unit is integrated below the base plate to jointly construct a solid execution surface capable of realizing zoned temperature control; the space-time field information fusion sensing module integrates a distributed fiber grating temperature sensor and a thermal imaging sensor, and realizes synchronous, global and undisturbed sensing of physical temperature and optical effect; and the adaptive prediction controller establishes a real-time digital twinborn model of a controlled object, maps set temperature distribution into actuator driving parameters through a field domain inverse problem solving algorithm, and introduces a closed-loop multi-mode coupling field actuator for online correction.

Description

Technical Field

[0001] This invention belongs to the field of experimental instrument technology, specifically, it relates to a control system and control method for an integrated heating and cooling platform. Background Technology

[0002] Microscope heating stages with integrated heating and cooling are crucial for in-situ microscopic observation of samples under varying temperatures. However, current mainstream technologies suffer from fundamental limitations in system architecture, sensing capabilities, and control logic, making it difficult to meet the stringent requirements of cutting-edge scientific research for high spatial resolution, high temperature uniformity, and long-term dynamic observation. These shortcomings are specifically manifested in the following three aspects: Existing technologies mostly use integral, contact-type thermoelectric coolers (Peltier plates) or resistance wires as the core actuators. This single structure of heating / cooling leads to two inherent drawbacks: first, it has high thermal inertia, resulting in slow heating and cooling rates and making it difficult to achieve rapid and precise temperature switching; second, it has extremely low spatial resolution (usually on the order of centimeters), making it impossible to perform fine-grained temperature control of the sample stage, let alone generate or cancel complex local temperature gradients, which severely limits its application in fields such as micro-area observation and biomimetic temperature field simulation. The current temperature sensing system suffers from insufficient dimensionality, resulting in blind spots and a disconnect between observation and feedback: Existing systems heavily rely on a few embedded contact temperature sensors (such as thermocouples or platinum resistance thermometers). This sparse, single-point or few-point temperature measurement method provides insufficient feedback information, making it impossible to reconstruct or assess the overall temperature distribution on the sample stage. It is also powerless to address spatial inhomogeneities caused by heat sinks, differences in sample heat capacity, or environmental disturbances. More importantly, the existing sensing system is completely independent of the core microscopic optical observation path. There is no informational link between the temperature control loop and the optical observation results, causing the control system to be unaware of the actual impact of its temperature control actions on the final observation target (such as the fluorescence intensity of cells or the Raman peak position of materials), resulting in a severe control-observation disconnect. The control logic is simple and lagging, unable to achieve global coordination and adaptive optimization: Based on the aforementioned simple actuators and sensors, existing control logic generally adopts the classic setpoint-error-PID local feedback control paradigm. This paradigm can only drive the actuators, striving to make the temperature at a limited number of sensor measurement points approach the set value. It cannot sense or control the temperature conditions in the sensor blind zone, and completely lacks the ability to model and respond to thermal coupling between actuators, thermal disturbances of the sample itself, and environmental interference. This kind of control method, which only sees the trees and not the forest, inevitably leads to poor temperature uniformity of the platform and slow dynamic response.

[0003] In summary, existing technologies are limited by the traditional architecture of integrated actuators, sparse point sensing, and local feedback control, and have reached bottlenecks in terms of spatial resolution, temperature uniformity, dynamic response, and observation correlation. A systematic innovation from system architecture to control logic is urgently needed. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a control system and control method for an integrated heating and cooling platform.

[0005] To achieve the aforementioned objectives, the technical solution adopted by this invention includes: a heating and cooling integrated heating stage control system, comprising a heating and cooling stage and a substrate for heating / cooling, and further comprising: The multimodal coupled field actuator includes a heating unit, a cooling unit, an adsorption layer, and a thermal control layer. The heating unit includes a near-infrared laser and a scanning galvanometer, which irradiates a high-absorptivity photothermal coating disposed on the back of the substrate (the high-absorptivity photothermal coating is such as a titanium nitride nanofilm prepared by atomic layer deposition, which has an absorptivity >95% and a thermal conductivity >20 ppm near a wavelength of 1550 nm). The heating unit (W / m·K) enables rapid, localized, and precise heating. The cooling unit includes multiple miniature thermoelectric coolers (TEC cooling arrays) and / or fans integrated within the heating stage to cool the substrate. The adsorption layer includes a microporous electrostatic adsorption film with high insulation and microporous structure on the outer surface of the substrate (the microporous electrostatic adsorption film is such as transparent indium tin oxide conductive glass with a silicon dioxide insulating layer on its surface, formed into an array of mutually insulated electrodes through photolithography, capable of independently applying 0-1000V DC voltage). The thermal control layer includes miniature thermoelectric coolers integrated under the substrate, together constructing a solid-state, zoned temperature-controlled execution surface. It also includes a fixing unit, located on the microporous electrostatic adsorption film on the outer surface of the substrate, used to fix the sample and suppress thermal drift. The heating and cooling units are physically decoupled, together forming a zoned temperature-controlled execution surface.

[0006] The all-time and all-space field information fusion sensing module integrates a distributed fiber optic temperature sensor and a thermal imaging sensor to achieve synchronous, full-domain, and non-disruptive sensing of physical temperature and optical effects. The all-time, all-space field information fusion sensing module specifically includes: a. Constructing a spatiotemporal hypergraph model: Each sensor data point (fiber optic grating point, thermal imager pixel) is modeled as a hypergraph node. The connections between nodes include not only spatial proximity but also physical associations (such as multiple sensor nodes under the same TEC control unit being strongly connected). This hypergraph is dynamically updated, and its edge weights are related to the real-time confidence of the nodes.

[0007] b. Confidence assessment and weighted fusion: A confidence score C_i(t) is calculated in real time for each data source.

[0008] For the fiber Bragg grating sensor C_fbg: evaluation is based on its historical data stability (reciprocal of variance) and consistency with readings from adjacent sensors.

[0009] For thermal imager pixel C_ir: the image sharpness (gradient), signal-to-noise ratio, and correlation with physical sensor readings over time are evaluated based on the region where the pixel is located.

[0010] c. Confidence-based diffusion completion algorithm: The final fused temperature field T_fused is obtained by solving the following optimization problem: argmin{ Σ_i C_i * || T_fused(s_i) - T_i || 2 + λ * Σ_{(j,k)∈E} w_jk * || T_fused(s_j) - T_fused(s_k) || 2} The first term forces the fused field to approximate the high-confidence original measurement value T_i; the second term is a smoothing term based on the hypergraph edge weights w_jk, allowing values ​​in low-confidence regions (such as areas where the thermal imager is obstructed) to "diffuse" from the high-confidence neighborhood; λ is the smoothing coefficient; s_i is the spatial coordinate of the i-th sensor; T_i is the actual measured original temperature value at the i-th sensor position s_i; C_i is the real-time confidence weight of the i-th sensor measurement value T_i; and E is the set of all edges in the hypergraph. Each edge connects two spatial positions (nodes); (j,k) is an edge belonging to the edge set E, connecting positions s_j and s_k.

[0011] The adaptive predictive controller driven by digital twins establishes a real-time digital twin model of the controlled object, maps the set temperature distribution to actuator driving parameters through a field inverse problem solving algorithm, and introduces closed-loop multimodal coupled field actuator feedback for online correction.

[0012] A method for controlling an integrated heating and cooling stage, employing a method similar to that of an integrated heating and cooling stage control system, includes: A1. The characteristics of the heating unit and the microfluidic electrostatic adsorption composite substrate are calibrated by a multimodal coupled field actuator. Temperature field information is obtained by a thermal imaging sensor. The spatiotemporal evolution of the full field temperature under arbitrary control input is predicted by a high-fidelity real-time digital twin model. Specifically, step A1 includes: A1.1. The heating and cooling units are calibrated using a low-power scanning method. The transient temperature field distribution T_physical(x, y, t) measured by a distributed fiber Bragg grating temperature sensor network (spatial resolution up to 0.5 mm) under different driving parameters is recorded. A1.2 Simultaneously start the microscope to image the standard grid sample placed on the stage. Based on the thermal imaging sensor, the temperature field information T_optical(x, y, t) is obtained by analyzing the astigmatism and sub-pixel displacement of feature points caused by the thermal refractive index change. A1.3. By fusing T_physical and T_optical data and combining the three-dimensional thermal parameters (thermal conductivity, heat capacity, and density) of the substrate and sample, a high-fidelity real-time digital twin model for this specific experimental configuration is constructed using the finite volume method. This model can predict the spatiotemporal evolution of the full-field temperature under arbitrary control inputs at millisecond speeds.

[0013] A2. Set up the inverse problem of temperature field solution and initial control. Set the difference between the target temperature and the current state of the digital twin model as input, run the field inverse problem solution algorithm, and efficiently solve the optimization problem through the adjoint variable method to obtain the optimal control vector. Specifically, step A2 includes: A2.1. Set the target temperature distribution T_target(x, y) (which can be a constant value or a specific gradient mode), and the temperature value T_target that is expected to be reached and maintained at the coordinate point (x, y). A2.2 The adaptive predictive controller does not directly drive the actuator. Instead, it takes the difference between T_target and the current state of the digital twin model as input and runs an algorithm to solve the inverse field problem. The core of this algorithm is to minimize the following loss function L: L = ∫∫[α(T_model(u) - T_target) 2 + β(▽ 2 T_model(u)) 2 + γ||u|| 2 ] dx dy Where T_model(u) is the predicted temperature field of the digital twin model under the control vector u (including heating unit and wavelength, flow velocity and fluid temperature at each node of the microfluidic channel, etc.), and α, β, and γ are weighting coefficients that respectively weigh tracking accuracy, temperature field smoothness (uniformity), and control energy consumption. ▽ 2It is a Laplace operator used to penalize drastic temperature changes and ensure uniformity. The control vector u is obtained by discretizing the spatial domain (x, y) into a grid, and then discretizing the control vector u into the driving value vector U of each execution unit. The digital twin model is expressed as a linear time-varying or linearizable state-space equation: dT / dt = A(U)T + B(U), or its steady-state approximation T = G * U (G is the transfer matrix, an N × N matrix where N is the number of control units (laser / TEC units). Matrix element G_{ij} represents the steady-state temperature change (°C / W or °C / A) caused by a unit driving force applied by the j-th control unit at the i-th temperature sensor location. This matrix is ​​obtained through experimental calibration: during system initialization, a series of known low-power excitations are applied to each control unit sequentially. After the system stabilizes, the temperature response distribution across the entire platform measured by the distributed fiber optic grating sensor network is recorded. Each column of matrix G can be fitted using linear regression. At this point, the loss function L is discretized as L = (GU - T_target). T Wα(GU - T_target) + U T WγU + (DGU) T Wβ(DGU), where W is the weight diagonal matrix and D is the discrete Laplacian operator matrix. Ultimately, the optimal solution U* can be obtained by solving the linear equation system (GTWαG +G). T D T WβDG + Wγ) U = G T Wα T_target is obtained.

[0014] A2.3. The optimal control vector u* is obtained by efficiently solving the optimization problem using the adjoint variable method. This step directly resolves the macroscopic temperature target into a set of coordinated working instructions for the underlying actuators.

[0015] A3. Adaptive multimodal coupling field precision execution: Laser of a specific wavelength (e.g., 1550nm infrared light targeting the absorption characteristics of the sample) and power is precisely projected onto a high-absorptivity photothermal coating on the back of the substrate through a scanning galvanometer, efficiently converting light energy into heat energy. Then, a micro thermoelectric cooler is used to achieve independent cooling of each micro-region by changing the direction and magnitude of the current, thus achieving localized cooling. Specifically, step A3 includes: A3.1. According to the instructions in u*, a laser of a specific wavelength (e.g., 1550nm infrared light targeting the sample's absorption characteristics) and power is precisely projected onto a high-absorptivity photothermal coating on the back of the substrate using a scanning galvanometer. This coating efficiently converts light energy into heat energy, achieving non-contact, high spatial resolution (approximately 10μm) targeted heating. This coating efficiently converts light energy into heat energy, enabling rapid, targeted heating of a specified area on the sample stage; then, a micro-thermoelectric cooler is used to achieve independent cooling of each micro-region by changing the direction and magnitude of the current, thus achieving localized cooling. A3.2 The micro thermoelectric cooler and the microporous electrostatic adsorption film adjust the temperature of the circulating fluid according to the overall temperature setting in U*; A4. Closed-loop multimodal coupled field actuator feedback and online dynamic correction: The adaptive predictive controller compares the optical signal f(T_model) predicted based on the physical sensor with the actual observed optical signal S_measuredv in real time. Using the corrected model, it generates a correction amount for a model parameter (such as local thermal conductivity) online, returns to the second step, and performs a fast inverse problem solution again (the computational cost is very small at this time because the initial solution u is very close to the optimum), and updates the control vector u_new. Specifically, step A4 includes: A4.1 During the temperature control process, the thermal imaging sensor continuously monitors the characteristic optical signals S (e.g., fluorescence intensity ratio, Raman specific peak position, interference fringe phase) of specific regions of the sample (such as organelles, material grain boundaries). A4.2 Establish an empirical mapping function f between the optical signal S and the local real temperature T_local (which can be obtained through prior calibration): S = f(T_local); A4.3 Adaptive predictive controllers utilize the inverse mapping function f -1 (Calibration required) Convert the actually observed optical signal S_measured into an equivalent observation temperature T_optical_inferred = f -1 (S_measured); A4.4 Directly calculate the temperature residual ΔT = T_optical_inferred - T_model (where T_model is the predicted temperature of the spatial point corresponding to T_optical_inferred by the digital twin). Feed this ΔT into a model parameter updater (e.g., based on Kalman filtering) to correct the thermal parameters (e.g., thermal conductivity k or heat capacity C) of local regions in the digital twin model online, obtaining δk or δC. The feedback is clearly positioned as a correction for the difference between the model-predicted temperature and the independent temperature measurement results inverted from the optical signal; A4.5. Using the corrected model, return to step two and perform another round of fast inverse problem solving (the computational cost is very small at this point because the initial solution u is very close to the optimum), and update the control vector u_new. This forms a closed loop with the goal of stabilizing the final optical observation effect, fundamentally eliminating the systematic error between pure physical temperature control and the actual optical response of the sample.

[0016] A computer storage medium storing a computer program, characterized in that: when the program is executed by a processor, it performs the steps of a method for controlling an integrated hot and cold heating stage.

[0017] Compared with the prior art, the advantages of the present invention include: (1) The present invention provides a heating stage control system and control method that integrates heating and cooling, which fundamentally solves the problem of thermodynamic drift: electrostatic adsorption force counteracts thermal stress, and the sample does not drift. It achieves ultra-high spatial resolution and fast response: the laser heating point can reach 10μm, the TEC array is independently controlled, complex temperature patterns can be realized, and the heating and cooling rates are much higher than those of the overall temperature control. (2) The integrated heating stage control system and method provided by this invention achieves, for the first time, synchronous full-domain closed-loop control of the temperature field and the optical effect field. It not only controls the surface temperature but also directly uses the stability of the final observed target as the control objective, resulting in a qualitative leap in accuracy and reliability. (3) The present invention provides a control system and control method for an integrated heating and cooling stage, which upgrades temperature control from passive adjustment to active design. By solving the inverse problem, the control command that simultaneously satisfies the optimal accuracy, uniformity and energy efficiency of the temperature across the entire field is calculated at one time, achieving precision from the open loop, and using photon feedback to correct model errors in real time. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an overall schematic diagram of a cold and hot integrated heating table control system and control method according to the present invention; Figure 2 This is a schematic diagram of a heating stage for a heating stage that integrates heating and cooling systems and control methods, as described in this invention.

[0020] Figure label: 11. Heating unit; 12. Cooling unit; 13. Adsorption layer; 14. Thermal control layer; 15. Hot and cold stage; 16. Substrate. Detailed Implementation

[0021] In view of the shortcomings of the prior art, the inventors of this invention, through long-term research and extensive practice, have proposed the technical solution of this invention. The technical solution, its implementation process, and principles will be further explained below with reference to the accompanying drawings and specific implementation examples in the embodiments of this application.

[0022] It should be noted that the embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, the present invention covers any substitutions, modifications, equivalent methods and solutions made on the spirit, principles and scope of the present invention as defined by the claims. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] In the description of this application, the terms "first," "second," "third," and similar terms do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms "one" or "a" do not indicate a quantity limitation, but rather indicate the presence of at least one. Terms such as "include" or "contain" mean that the elements or objects preceding "include" cover the elements or objects listed following "include" or their equivalents, and do not exclude other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.

[0024] In the description of this application, the terms center, up, down, front, back, left, right, vertical, horizontal, top, bottom, inside, outside, etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. In addition, when using positional terms such as sides, outer side, top and bottom, etc., it should be understood that they are used only for the convenience of understanding and description, taking into account that the structure may be oriented to other positions.

[0025] In the description of this application, unless otherwise expressly specified and limited, the technical or scientific terms used shall have the ordinary meaning understood by a person with ordinary skills in the art to which this application pertains. Terms such as installation, connection, and linking shall be interpreted broadly, for example, they may refer to fixed connection, detachable connection, mating connection, or integral connection. For a person skilled in the art, the specific meaning of the above terms in this application may be understood according to the specific circumstances.

[0026] The present invention aims to introduce and explain the structural composition of a cold and hot integrated heating stage control system and control method, as well as the cooperation relationship between the various components. Unless otherwise specified, the dimensions, materials, and manufacturing processes of the various components in the cold and hot integrated heating stage control system and control method of the present invention can be selected according to specific circumstances, and no special limitations or explanations are made here.

[0027] Furthermore, to provide the public with a better understanding of the present invention, certain specific details are described in detail in the following description of the invention. However, those skilled in the art will fully understand the invention even without these detailed descriptions.

[0028] Example 1 Please see Figure 1 and Figure 2 A heating and cooling integrated stage control system includes a heating and cooling stage 15 and a base plate 16 for heating / cooling, and further includes: The multimodal coupled field actuator includes a heating unit 11, a cooling unit 12, an adsorption layer 13, and a thermal control layer 14. The heating unit 11 includes a near-infrared laser and a scanning galvanometer, which irradiates a high-absorptivity photothermal coating disposed on the back of the substrate 16 (the high-absorptivity photothermal coating is such as a titanium nitride nanofilm prepared by atomic layer deposition, which has an absorptivity >95% and a thermal conductivity >20 ppm near a wavelength of 1550 nm). W / m·K) to achieve rapid, localized, and precise heating; the cooling unit 12 includes multiple micro thermoelectric coolers and / or fans integrated within the substrate to dissipate heat from the substrate 16, and vents are provided on the outer surface of the heating and cooling stage for air intake and exhaust; the adsorption layer 13 includes a microporous electrostatic adsorption film with high insulation and microporous structure on the outer surface of the substrate 16 (the microporous electrostatic adsorption film is such as transparent indium tin oxide conductive glass with a silicon dioxide insulating layer on its surface, which is formed into an array of mutually insulated electrodes through photolithography and can be independently applied with a DC voltage of 0-1000V); the thermal control layer 14 includes micro thermoelectric coolers integrated inside or below the substrate, together constructing a solid-state, zone-controlled temperature execution surface; The all-time and all-space field information fusion sensing module integrates a distributed fiber optic temperature sensor and a thermal imaging sensor to achieve synchronous, full-domain, and non-disruptive sensing of physical temperature and optical effects. The all-time, all-space field information fusion sensing module specifically includes: a. Constructing a spatiotemporal hypergraph model: Each sensor data point (fiber optic grating point, thermal imager pixel) is modeled as a hypergraph node. The connections between nodes include not only spatial proximity but also physical associations (such as multiple sensor nodes under the same TEC control unit being strongly connected). This hypergraph is dynamically updated, and its edge weights are related to the real-time confidence of the nodes.

[0029] b. Confidence assessment and weighted fusion: A confidence score C_i(t) is calculated in real time for each data source.

[0030] For the fiber Bragg grating sensor C_fbg: evaluation is based on its historical data stability (reciprocal of variance) and consistency with readings from adjacent sensors.

[0031] For thermal imager pixel C_ir: the image sharpness (gradient), signal-to-noise ratio, and correlation with physical sensor readings over time are evaluated based on the region where the pixel is located.

[0032] c. Confidence-based diffusion completion algorithm: The final fused temperature field T_fused is obtained by solving the following optimization problem: argmin{ Σ_i C_i * || T_fused(s_i) - T_i || 2 + λ * Σ_{(j,k)∈E} w_jk * || T_fused(s_j) - T_fused(s_k) || 2} The first term forces the fused field to approximate the high-confidence original measurement value T_i; the second term is a smoothing term based on the hypergraph edge weights w_jk, allowing values ​​in low-confidence regions (such as areas where the thermal imager is obstructed) to "diffuse" from the high-confidence neighborhood; λ is the smoothing coefficient; s_i is the spatial coordinate of the i-th sensor; T_i is the actual measured original temperature value at the i-th sensor position s_i; C_i is the real-time confidence weight of the i-th sensor measurement value T_i; and E is the set of all edges in the hypergraph. Each edge connects two spatial positions (nodes); (j,k) is an edge belonging to the edge set E, connecting positions s_j and s_k.

[0033] This algorithm no longer treats sensors as equal individuals, but dynamically assesses their reliability and uses physical correlation as a guide to perform intelligent information completion and denoising, thereby generating a spatiotemporally continuous, high-confidence, and higher-resolution "super-resolution" fused temperature field than any single sensor, providing unprecedented high-quality input for digital twin models.

[0034] The adaptive predictive controller driven by digital twins establishes a real-time digital twin model of the controlled object, maps the set temperature distribution to actuator driving parameters through a field inverse problem solving algorithm, and introduces closed-loop multimodal coupled field actuator feedback for online correction.

[0035] The specific components of a digital twin-driven adaptive predictive controller include: Constructing a differentiable shadow model: e. In the initialization phase, a massive amount of {U, T} data pairs are generated using a high-fidelity model, and a lightweight deep neural network is trained as its shadow. This network takes the control vector U as input and directly predicts the overall temperature T_pred = NN(U; θ). Crucially, this network architecture is designed to be fully differentiable, and its output gradient with respect to the input U... It can perform calculations efficiently and analytically.

[0036] f. Transform the inverse problem into a differentiable optimization: T_model in the loss function L(U) is provided by the shadow model NN(U; θ). Since the model is differentiable, solving for min L(U) can be done using efficient gradient-based algorithms (such as the quasi-Newton method), directly solving U* for hundreds of control units in parallel on the GPU, which is several orders of magnitude faster than traditional methods.

[0037] g. Online Adaptation Based on Meta-Learning: When the closed-loop photon feedback calculates the model parameter correction δp (such as changes in local thermal conductivity), instead of modifying the complex high-fidelity model, it updates the parameters θ of the shadow model. The system pre-trains a meta-learner that learns the mapping from "physical parameter perturbation δp" to "neural network parameter update Δθ". When δp is identified, the meta-learner instantly provides Δθ, causing the behavior of the shadow model NN(U; θ+Δθ) to immediately match the new physical conditions.

[0038] This scheme creatively distinguishes between simulation models (high-fidelity, for understanding) and control models (lightweight, differentiable, for real-time decision-making). Through differentiable programming and meta-learning, it achieves: 1) millisecond-level optimal control solutions; and 2) second-level online model adaptation. This enables the system to truly possess the intelligent capability of "learning by doing" during a single experiment.

[0039] A method for controlling an integrated heating and cooling stage, employing a method similar to that of an integrated heating and cooling stage control system, includes: A1. The characteristics of the heating unit and the microfluidic electrostatic adsorption composite substrate are calibrated by a multimodal coupled field actuator. Temperature field information is obtained by a thermal imaging sensor. The spatiotemporal evolution of the full field temperature under arbitrary control input is predicted by a high-fidelity real-time digital twin model. Specifically, step A1 includes: A1.1. The heating and cooling units are calibrated using a low-power scanning method. The transient temperature field distribution T_physical(x, y, t) measured by a distributed fiber Bragg grating temperature sensor network (spatial resolution up to 0.5 mm) under different driving parameters is recorded. A1.2 Simultaneously start the microscope to image the standard grid sample placed on the stage. Based on the thermal imaging sensor, the temperature field information T_optical(x, y, t) is obtained by analyzing the astigmatism and sub-pixel displacement of feature points caused by the thermal refractive index change. A1.3. By fusing T_physical and T_optical data and combining the three-dimensional thermal parameters (thermal conductivity, heat capacity, and density) of the substrate and sample, a high-fidelity real-time digital twin model for this specific experimental configuration is constructed using the finite volume method. This model can predict the spatiotemporal evolution of the full-field temperature under arbitrary control inputs at millisecond speeds.

[0040] A2. Set up the inverse problem of temperature field solution and initial control. Set the difference between the target temperature and the current state of the digital twin model as input, run the field inverse problem solution algorithm, and efficiently solve the optimization problem through the adjoint variable method to obtain the optimal control vector. Specifically, step A2 includes: A2.1. Set the target temperature distribution T_target(x, y) (which can be a constant value or a specific gradient mode), and the temperature value T_target that is expected to be reached and maintained at the coordinate point (x, y). A2.2 The adaptive predictive controller does not directly drive the actuator. Instead, it takes the difference between T_target and the current state of the digital twin model as input and runs an algorithm to solve the inverse field problem. The core of this algorithm is to minimize the following loss function L: L = ∫∫[α(T_model(u) - T_target) 2 + β(▽ 2 T_model(u)) 2 + γ||u|| 2 ] dx dy Where T_model(u) is the predicted temperature field of the digital twin model under the control vector u (including heating unit and wavelength, flow velocity and fluid temperature at each node of the microfluidic channel, etc.), and α, β, and γ are weighting coefficients that respectively weigh tracking accuracy, temperature field smoothness (uniformity), and control energy consumption. ▽ 2It is a Laplace operator used to penalize drastic temperature changes and ensure uniformity. The control vector u is obtained by discretizing the spatial domain (x, y) into a grid and then discretizing the control vector u into the drive value vector U of each execution unit. The digital twin model is expressed as a linear time-varying or linearizable state-space equation: dT / dt = A(U)T + B(U), or its steady-state approximation T = G * U (G is the transfer matrix, an N × N matrix where N is the number of control units (laser / TEC units)). In addition to steady-state response, the model can predict transient behavior through the state-space equation or transfer function, which is the basis for achieving fast, over-regulation-free temperature control. Matrix element G_{ij} represents the steady-state temperature change (°C / W or °C / A) caused by applying a unit drive at the i-th temperature sensor location when the j-th control unit applies a unit drive. This matrix is ​​obtained through experimental calibration: during system initialization, a series of known low-power excitations are applied to each control unit sequentially. After the system stabilizes, the temperature response distribution across the entire platform measured by the distributed fiber optic grating sensor network is recorded. The matrix G can be fitted using linear regression. For each column), the loss function L is discretized as L = (GU - T_target). T Wα(GU - T_target) + U T WγU + (DGU) T Wβ(DGU), where W is the weight diagonal matrix and D is the discrete Laplacian operator matrix. Ultimately, the optimal solution U* can be obtained by solving the linear equation system (G... T WαG + G T D T WβDG + Wγ) U = G T Wα T_target is obtained.

[0041] A2.3. The optimal control vector u* is obtained by efficiently solving the optimization problem using the adjoint variable method. This step directly resolves the macroscopic temperature target into a set of coordinated working instructions for the underlying actuators.

[0042] A3. The adaptive multimodal coupling field is precisely executed. A laser of a specific wavelength (e.g., 1550nm infrared light for the absorption characteristics of the sample) and power is precisely projected onto a high-absorptivity photothermal coating on the back of the substrate through a scanning galvanometer to efficiently convert light energy into heat energy. By switching to another wavelength of laser and adjusting the projection mode, the radiation cooling effect of the coating can be excited to achieve local cooling. Specifically, step A3 includes: A3.1. According to the instructions in u*, a laser of a specific wavelength (e.g., 1550nm infrared light specific to the sample's absorption characteristics) and power is precisely projected onto a high-absorptivity photothermal coating on the back of the substrate using a scanning galvanometer. This coating efficiently converts light energy into heat energy, achieving contactless, high spatial resolution (approximately 10μm) point heating. Alternatively, a micro-thermoelectric cooler can be used to achieve independent cooling of each micro-region by changing the direction and magnitude of the current, thus achieving localized cooling. A3.2 The micro thermoelectric cooler and microporous electrostatic adsorption film regulate the temperature of the circulating fluid according to the overall temperature setting in U*. In the control vector U and matrix G, one or more control dimensions need to be allocated to this overall temperature control block. Its temperature setpoint is also obtained by solving the inverse problem and works in conjunction with the zone temperature control. The voltage control of the electrostatic adsorption film should be an independent, parallel closed loop. It automatically generates the adsorption pattern based on the sample image, does not participate in solving the inverse problem of the temperature field, and only receives the adsorption enable command. A4. Closed-loop multimodal coupled field actuator feedback and online dynamic correction: The adaptive predictive controller compares the optical signal f(T_model) predicted based on the physical sensor with the actual observed optical signal S_measuredv in real time. Using the corrected model, it generates a correction amount for a model parameter (such as local thermal conductivity) online, returns to the second step, and performs a fast inverse problem solution again (the computational cost is very small at this time because the initial solution u is very close to the optimum), and updates the control vector u_new. Specifically, step A4 includes: A4.1 During the temperature control process, the thermal imaging sensor continuously monitors the characteristic optical signals S (e.g., fluorescence intensity ratio, Raman specific peak position, interference fringe phase) of specific regions of the sample (such as organelles, material grain boundaries). A4.2 Establish an empirical mapping function f between the optical signal S and the local real temperature T_local (which can be obtained through prior calibration): S = f(T_local); A4.3 Adaptive predictive controllers utilize the inverse mapping function f -1 (Calibration required) Convert the actually observed optical signal S_measured into an equivalent observation temperature T_optical_inferred = f -1 (S_measured); A4.4 Directly calculate the temperature residual ΔT = T_optical_inferred - T_model (where T_model is the predicted temperature of the spatial point corresponding to T_optical_inferred by the digital twin). Feed this ΔT into a model parameter updater (e.g., based on Kalman filtering) to correct the thermal parameters (e.g., thermal conductivity k or heat capacity C) of local regions in the digital twin model online, obtaining δk or δC. The feedback is clearly positioned as a correction for the difference between the model-predicted temperature and the independent temperature measurement results inverted from the optical signal; A4.5. Using the corrected model, return to step two and perform another round of fast inverse problem solving (the computational cost is very small at this point because the initial solution u is very close to the optimum), and update the control vector u_new. This forms a closed loop with the goal of stabilizing the final optical observation effect, fundamentally eliminating the systematic error between pure physical temperature control and the actual optical response of the sample.

[0043] This system adopts a collaborative architecture of FPGA (real-time control layer) + GPU / CPU (high-performance computing layer). Real-time control layer (FPGA): Input interfaces: The wavelength data stream of the fiber Bragg grating demodulator can be read directly via a high-speed SPI or LVDS interface; the raw images of the thermal imaging sensor can be captured directly via a Camera Link or CoaXPress interface.

[0044] Core task: Perform millisecond-level preprocessing on the raw data (such as converting wavelength offsets into temperature values ​​T_physical, executing image algorithms to obtain T_optical or optical signals S), and perform timestamp alignment.

[0045] Output interface: Upload the preprocessed fused temperature field data T_fused to the computing layer via Gigabit Ethernet or PCIe; Receive the control vector U* issued by the computing layer and solve it into specific galvanometer position commands, laser power modulation signals, and TEC drive currents, and send them to each driver card via analog output or digital bus.

[0046] High-performance computing layer (GPU / CPU workstations): Core task: Run the digital twin model and inverse problem solving algorithm. Receive T_fused from the FPGA and update the model state; when a new control quantity needs to be calculated, solve the equation H * U* = b, obtain U*, and then send it to the FPGA.

[0047] Synchronization mechanism: PTP (Precision Clock Protocol) is used to synchronize the clocks of the FPGA, workstation, and all sensors. The FPGA triggers data acquisition, processing, and transmission at a fixed period (e.g., 1ms) to ensure a strictly defined control cycle.

[0048] Working principle of the invention: Setup and Modeling: The process begins with the user setting the macroscopic target temperature field T_target (e.g., isothermal or gradient). The high-fidelity digital twin model built during system initialization forms the computational basis for all subsequent predictions and optimizations.

[0049] Prediction and Solving: The inverse problem solver is the intelligent brain of the system. Instead of performing traditional PID calculations, it takes the difference between T_target and the model-predicted T_model as input and directly calculates the optimal control instruction set U* that enables the overall temperature to reach the desired state by solving an optimization problem that minimizes the loss function. This achieves a paradigm shift from passive correction to proactive design.

[0050] Execution and observation: The optimal command U* is sent to the multimodal actuator (laser heating array, TEC cooling array, electrostatic adsorption, etc.), the physical world generates the actual temperature field T_real, and the microscope observes it simultaneously.

[0051] Sensing and Fusion: The global fusion sensing network simultaneously acquires two key data sources: physical temperature T_physical (measured by a distributed fiber Bragg grating sensor) and sample optical signal S_optical (obtained by a thermal imaging sensor or spectral detection).

[0052] Feedback and Correction: A unique optical feedback corrector inverts the optical signal S_optical into independent temperature information and compares it with the model's predicted value to generate a residual ΔT. ΔT is used to update the parameters of the digital twin model (such as local thermal conductivity) online, making the model increasingly approximate the real system. The updated model is immediately used for the next round of control calculations, forming an enhanced closed loop of perception-learning-control.

[0053] In this way, the global inverse problem solving algorithm (β(▽) in the loss function) is used. 2 T) 2 The device, along with a high-resolution actuator, can actively generate and maintain a highly uniform or arbitrarily preset temperature field. Electrostatic adsorption physically locks the sample in place, ensuring its position remains unchanged regardless of thermal expansion and contraction of the stage.

[0054] Laser heating exhibits virtually no thermal inertia, with TEC arrays responding in milliseconds. Digital twin models can accurately predict dynamic processes, and inverse problem solving can generate optimal dynamic control trajectories, enabling rapid, overshoot-free temperature changes.

[0055] The innovative closed-loop photon feedback directly uses the microscope's observations as the feedback input to the control system. The ultimate goal of system control is no longer for the stage temperature to equal the set value, but for the observed sample optical signal to equal the expected value. It automatically compensates for all disturbances such as uncertainties in the sample's own thermal properties, differences in thermal contact, and model errors, ensuring the direct reliability and repeatability of scientific data. This is a function that no existing temperature control equipment can achieve.

[0056] Based on digital twins and online parameter calibration, the system can adapt to different samples, slides, and liquid environments. Users only need to care about the desired temperature environment for the sample, without having to worry about how to adjust complex hardware parameters. This greatly reduces the operational threshold and failure rate of precision temperature-controlled experiments, allowing researchers to focus their energy on the scientific questions themselves.

[0057] A computer storage medium storing a computer program, characterized in that: when the program is executed by a processor, it performs the steps of a method for controlling an integrated hot and cold heating stage.

[0058] It should be understood that the above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. It should not be considered that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, several simple deductions or substitutions can be made without departing from the concept of the present invention. All equivalent changes or modifications made in accordance with the spirit and essence of the present invention should be covered within the protection scope of the present invention.

Claims

1. A control system for an integrated heating and cooling stage, comprising a heating / cooling stage and a substrate for heating / cooling, characterized in that: Also includes: A multimodal coupled field actuator includes a heating unit, a cooling unit, an adsorption layer, and a thermal control layer. The heating unit includes a near-infrared laser and a scanning galvanometer, which irradiate a high-absorptivity photothermal coating disposed on the back of a substrate to achieve rapid, localized heating. The cooling unit includes multiple miniature thermoelectric coolers integrated within a heating stage to cool the substrate. The adsorption layer includes a microporous electrostatic adsorption film with high insulation and a microporous structure on the outer surface of the substrate. The thermal control layer includes miniature thermoelectric coolers integrated under the substrate, together forming a solid-state, zone-controlled temperature-controlled actuation surface. The all-time and all-space field information fusion sensing module integrates a distributed fiber optic temperature sensor and a thermal imaging sensor to achieve synchronous, full-domain, and non-disruptive sensing of physical temperature and optical effects. An adaptive predictive controller establishes a real-time digital twin model of the controlled object, maps the set temperature distribution to actuator driving parameters through a field inverse problem solving algorithm, and introduces a closed-loop multimodal coupled field actuator for online correction.

2. The integrated heating and cooling platform control system according to claim 1, characterized in that: The all-time, all-space field information fusion sensing module specifically includes: a. Construct a spatiotemporal hypergraph model: model each sensor data point as a hypergraph node; b. Dynamic confidence assessment and weighted fusion: A confidence score C_i(t) is calculated in real time for each data source; For the fiber Bragg grating sensor C_fbg: the evaluation is based on its historical data stability and consistency with readings from adjacent sensors; For thermal imager pixel C_ir: the image sharpness, signal-to-noise ratio, and correlation with physical sensor readings over time are evaluated based on the region where the pixel is located; c. Confidence-based diffusion completion algorithm: The final fused temperature field T_fused is obtained by solving the following optimization problem: argmin{ Σ_i C_i * || T_fused(s_i) - T_i || 2 + λ * Σ_{(j,k)∈E} w_jk * || T_fused(s_j) - T_fused(s_k) || 2 } The first term is the forced fusion field approximation of the high-confidence original measurement value T_i; the second term is the smoothing term based on the hypergraph edge weights w_jk; λ is the smoothing coefficient; s_i is the spatial coordinate of the i-th sensor; T_i is the actual measured original temperature value at the i-th sensor position s_i; C_i is the real-time confidence weight of the i-th sensor measurement value T_i; E is the set of all edges in the hypergraph; each edge connects two spatial positions; (j,k) is an edge belonging to the edge set E, connecting positions s_j and s_k.

3. A control method for an integrated heating and cooling platform, employing the control method of the control system described in any one of claims 1-2, characterized in that: Includes the following steps: A1. The characteristics are calibrated by a multimodal coupled field actuator, the temperature field information is obtained by a thermal imaging sensor, and the spatiotemporal evolution of the full field temperature under arbitrary control input is predicted by a real-time digital twin model. A2. Solving the inverse problem of the set temperature field and initial control: Set the difference between the target temperature and the current state of the digital twin model as input, run the inverse problem solving algorithm of the field, and efficiently solve the optimization problem through the adjoint variable method to obtain the optimal control vector; A3. Adaptive multimodal coupling field precision execution: Lasers of specific wavelengths and powers are precisely projected onto a high-absorptivity photothermal coating on the back of a substrate through a scanning galvanometer to efficiently convert light energy into heat energy. Then, a micro thermoelectric cooler is used to achieve independent cooling of each micro-region by changing the direction and magnitude of the current, thus achieving localized cooling. A4. Closed-loop multimodal coupled field actuator feedback and online dynamic correction: The adaptive predictive controller compares the predicted optical signal with the actual observed optical signal in real time, uses the corrected model to generate a correction amount for the model parameters online, returns to step A2, and performs another round of fast inverse problem solving to update the control vector.

4. The method for controlling a combined heating and cooling platform according to claim 3, characterized in that: Step A1 includes: A1.

1. The heating unit and cooling unit are calibrated by low-power scanning, and the transient temperature field distribution T_physical(x, y, t) measured by the distributed fiber optic temperature sensor is recorded under different driving parameters. A1.2 Simultaneously start the microscope to image the standard grid sample placed on the stage. Through the thermal imaging sensor, the temperature is inverted by analyzing the astigmatism and sub-pixel displacement of feature points caused by the thermally induced refractive index change to obtain another dimension of temperature field information T_optical(x, y, t). A1.

3. The T_physical and T_optical data are fused, and the three-dimensional thermal parameters of the substrate and sample are combined to construct a high-fidelity real-time digital twin model for this specific experimental configuration using the finite volume method.

5. The method for controlling a combined heating and cooling platform according to claim 3, characterized in that: Step A2 includes: A2.

1. Define a target temperature distribution T_target(x, y), which is the temperature value T_target that is expected to be reached and maintained at the coordinate point (x, y). A2.2 The adaptive predictive controller takes the difference between the temperature value T_target and the current state of the digital twin model as input and runs an algorithm to solve the inverse field problem in order to minimize the following loss function L: L = ∫∫[α(T_model(u) - T_target) 2 + β(▽ 2 T_model(u)) 2 + γ||u|| 2 ] dx dy Where T_model(u) is the predicted temperature field of the digital twin model under the control vector u, and α, β, γ are weighting coefficients that weigh tracking accuracy, temperature field smoothness and control energy consumption, respectively. ▽ 2 The Laplace operator is used to penalize drastic temperature changes. A2.3 Solve the optimization problem using the adjoint variable method to obtain the optimal control vector u*.

6. The method for controlling a combined heating and cooling platform according to claim 5, characterized in that: The control vector u is obtained through the following method: The spatial domain (x, y) is discretized into a grid, and the control vector u is discretized into the driving value vector U of each execution unit; The digital twin model is expressed as a linear time-varying or linearizable state-space equation: dT / dt = A(U)T + B(U), or its steady-state approximation T = G * U, where G is the transfer matrix. In this case, the loss function L is discretized as L = (GU - T_target). T Wα(GU - T_target) + U T WγU + (DGU) T Wβ(DGU), where W is the weight diagonal matrix and D is the discrete Laplacian operator matrix. Ultimately, the optimal solution U* can be obtained by solving the linear equation system (G... T WαG + G T D T WβDG + Wγ) U = G T Wα T_target is obtained.

7. The method for controlling a combined heating and cooling platform according to claim 3, characterized in that: Step A3 includes: A3.

1. According to the instructions in u*, a laser of a specific wavelength and power is precisely projected onto a high-absorptivity photothermal coating on the back of the substrate via a scanning galvanometer. This coating efficiently converts light energy into heat energy, achieving rapid, targeted heating of a specified area on the sample stage. Then, a miniature thermoelectric cooler is used to achieve independent cooling of each micro-region by changing the direction and magnitude of the current, thus achieving localized cooling. A3.2 The micro thermoelectric cooler and the microporous electrostatic adsorption film adjust the temperature of the circulating fluid according to the overall temperature setting in U*.

8. The method for controlling a combined heating and cooling platform according to claim 3, characterized in that: Step A4 includes: A4.1 During the temperature control process, the thermal imaging sensor continuously monitors the characteristic optical signal S of a specific area of ​​the sample; A4.2 Establish an empirical mapping function f between the optical signal S and the local real temperature T_local: S = f(T_local); A4.3 Adaptive predictive controllers utilize the inverse mapping function f -1 The observed optical signal S_measured is converted into an equivalent observation temperature T_optical_inferred = f -1 (S_measured); A4.4 Directly calculate the temperature residual ΔT = T_optical_inferred - T_model, and feed this ΔT into a model parameter updater to correct the thermal parameters of local areas in the digital twin model online, to obtain δk or δC; The feedback is clearly positioned as a correction for the difference between the model-predicted temperature and the independent temperature measurement results inverted from the optical signal; A4.5 Using the corrected model, return to step A2 and perform another round of fast inverse problem solving to update the control vector u_new.

9. A computer storage medium storing a computer program, characterized in that: When the program is executed by the processor, it performs the steps of the method as described in any one of claims 2-7.

Images