Membrane-based microelectromechanical systems (mems) thermopile sensor self-calibration method and system based on digital twinning

Abstract

The embodiment of the application provides a MEMS thermoelectric sensor self-calibration method and system based on digital twinning, acquires a measured thermoelectric electromotive force time sequence signal and a measured environmental temperature time sequence signal of a packaging cavity in a current working period of a micro-electro-mechanical system thermoelectric sensor, and combines the signals into a sensor real-time working state data set; a digital twinning body instance isoclitic with the physical entity of the sensor is called, the data set is loaded into a thermoelectric coupling simulation model of the digital twinning body instance to perform simulation deduction, and a simulated thermoelectric electromotive force time sequence signal is generated; signal difference analysis is performed on the measured and simulated thermoelectric electromotive force time sequence signals, and a thermoelectric electromotive force drift characteristic curve reflecting the output deviation between the physical entity and the digital twinning body is obtained; the drift characteristic curve is used to perform parameter reverse correction on a current response characteristic function of the sensor, an updated response characteristic function is generated and replaces the original function, and self-calibration is completed. The application realizes continuous self-calibration of the sensor in a full life cycle under a non-interrupted working state.

Description

Technical Field

[0001] This application relates to the fields of intelligent sensor calibration and digital twin technology, specifically to a self-calibration method and system for MEMS thermopile sensors based on digital twins. Background Technology

[0002] Microelectromechanical systems (MEMS) thermopile sensors are widely used in non-contact temperature measurement. Their core principle is based on the Seebeck effect, converting temperature difference into thermoelectric potential output. However, during long-term operation, MEMS thermopile sensors are affected by a combination of factors, including thermal stress relaxation of the encapsulation material, slow changes in electrode contact resistance, and cyclic loading of ambient temperature and humidity. This causes irreversible drift in their thermoelectric conversion characteristics, leading to a gradual deterioration in measurement accuracy over time. Existing calibration techniques primarily rely on external standard blackbody radiation sources or high-precision reference thermometers for periodic offline calibration. This method not only requires interrupting the sensor's normal operation but also suffers from limitations in calibration cycles due to manual operation costs, making it impossible to achieve real-time tracking and continuous correction of sensor drift.

[0003] Some existing technologies attempt to estimate the impact of ambient temperature changes on the output and perform software compensation by integrating auxiliary temperature-sensitive elements within the sensor. However, these solutions can only eliminate the linear drift component caused by ambient temperature changes and cannot capture the nonlinear characteristic drift caused by inherent factors such as material aging and microstructure degradation in thermopile sensors. Furthermore, the auxiliary elements themselves also suffer from drift problems, resulting in limited compensation accuracy. In addition, existing technologies lack the means to systematically quantify and analyze the deviation between the actual and ideal outputs using the sensor's own physical model and then use this to drive the update of the characteristic function. This makes it difficult to effectively guarantee the measurement reliability of MEMS thermopile sensors throughout their entire lifecycle. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a self-calibration method and system for MEMS thermopile sensors based on digital twins.

[0005] In conjunction with the first aspect of this application, a self-calibration method for MEMS thermopile sensors based on digital twins is provided, applied to a self-calibration system for MEMS thermopile sensors based on digital twins, the method comprising: The measured thermoelectric potential timing signal of the microelectromechanical system thermopile sensor during the current working period is acquired, and the measured ambient temperature timing signal of the packaging cavity where the microelectromechanical system thermopile sensor is located is acquired simultaneously. The measured thermoelectric potential timing signal and the measured ambient temperature timing signal are combined into a sensor real-time working status data set. A digital twin instance is invoked that is isomorphic to the physical entity of the thermopile sensor of the microelectromechanical system. The real-time operating status data set of the sensor is loaded into the thermoelectric coupling simulation model of the digital twin instance. The thermoelectric coupling simulation model is used to simulate and deduce the thermoelectric conversion physical process of the thermopile sensor of the microelectromechanical system, and generate the simulated thermoelectric potential time-series signal output by the digital twin instance. The measured thermoelectric potential timing signal and the simulated thermoelectric potential timing signal are subjected to signal difference analysis processing to obtain the thermoelectric potential drift characteristic curve reflecting the output deviation between the physical entity of the thermopile sensor of the microelectromechanical system and the digital twin instance. The current response characteristic function of the microelectromechanical system thermopile sensor is corrected by using the thermoelectric potential drift characteristic curve to generate an updated response characteristic function of the microelectromechanical system thermopile sensor. The original response characteristic function of the microelectromechanical system thermopile sensor is then replaced with the updated response characteristic function to complete the self-calibration process.

[0006] In conjunction with the second aspect of this application, a self-calibration system for MEMS thermopile sensors based on digital twins is provided. The self-calibration system for MEMS thermopile sensors based on digital twins includes a machine-readable storage medium and a processor. The machine-readable storage medium stores machine-executable instructions. When the processor executes the machine-executable instructions, the self-calibration system for MEMS thermopile sensors based on digital twins implements the aforementioned self-calibration method for MEMS thermopile sensors based on digital twins.

[0007] In conjunction with a third aspect of this application, a computer-readable storage medium is provided, wherein computer-executable instructions are stored therein, and when the computer-executable instructions are executed, the aforementioned self-calibration method for MEMS thermopile sensors based on digital twins is implemented.

[0008] Combining any of the above aspects, firstly, by combining the measured thermoelectric potential (TEP) time-series signal of the microelectromechanical system (MEMS) thermopile sensor during the current operating period with the synchronously acquired measured ambient temperature time-series signal of the encapsulation cavity, a real-time operating state data set of the sensor is formed. Then, a digital twin instance isomorphic to the physical entity is invoked, and the sensor's real-time operating state data set is loaded into its thermoelectric coupling simulation model. This thermoelectric coupling simulation model is used to simulate and deduce the thermoelectric conversion physical process and generate a simulated TEP time-series signal. Since the digital twin instance is unaffected by physical factors such as material aging and microscopic degradation under ideal conditions, its output simulated TEP time-series signal can be considered as the ideal reference output of the sensor under the current operating conditions. Based on this, by analyzing the measured TEP time-series signal... The time-series signals of the sequence signal and the simulated thermoelectric potential are analyzed to obtain the thermoelectric potential drift characteristic curve, which describes the evolution of the output deviation of the physical entity due to inherent degradation factors over time or operating conditions. More importantly, the thermoelectric potential drift characteristic curve is used to perform parameter reverse correction processing on the current response characteristic function of the microelectromechanical system thermopile sensor, generating an updated response characteristic function and replacing the original function. This allows the mathematical model of the sensor to be adaptively updated according to the real deviation revealed by the digital twin, and calibration can be completed without relying on any external standard source. This realizes continuous self-calibration of the microelectromechanical system thermopile sensor throughout its entire life cycle without interrupting operation, significantly improving the measurement accuracy, stability and reliability of the sensor during long-term operation. Attached Figure Description

[0009] Figure 1 A schematic flowchart of the self-calibration method for MEMS thermopile sensors based on digital twins provided in this application embodiment. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0012] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0013] Figure 1 This document illustrates a flowchart of a self-calibration method for MEMS thermopile sensors based on digital twins, provided in an embodiment of this application. It should be understood that in other embodiments, the order of some steps in the self-calibration method for MEMS thermopile sensors based on digital twins may be shared based on actual needs, or some steps may be omitted or maintained. The detailed self-calibration method for MEMS thermopile sensors based on digital twins includes: This method is applied to industrial temperature monitoring scenarios deploying edge microcontrollers and cloud-based digital twin engines. The edge microcontroller is integrated onto the signal conditioning circuit board of a MEMS thermopile sensor, while the cloud-based digital twin engine is deployed in a server cluster of an industrial IoT platform. Data transmission between the edge microcontroller and the cloud-based digital twin engine occurs via an industrial wireless communication gateway. Upon sensor delivery, the factory calibration device writes an initial response characteristic function parameter table to the sensor's internal storage. This table includes initial values ​​for the Seebeck coefficient scaling factor and the cold junction temperature compensation bias, calibrated under multiple standard temperature difference conditions. Simultaneously, the cloud-based digital twin engine creates a digital twin instance uniquely corresponding to the physical entity of the sensor. The thermal resistance network topology parameters of the initial thermoelectric coupling simulation model of the digital twin instance are consistent with the physical geometric dimensions of the sensor at the time of delivery. The implementation process of this method is detailed below using the same industrial kiln wall temperature monitoring scenario. In this scenario, the MEMS thermopile sensor is attached to the outer wall of the kiln and receives infrared thermal radiation from the kiln wall through the window of the encapsulation cavity. The cold end of the sensor is in contact with the base of the encapsulation cavity, and the hot end of the sensor absorbs radiated heat through a thin-film thermocouple.

[0014] Step S110: Obtain the measured thermoelectric potential timing signal of the MEMS thermopile sensor during the current working period, and simultaneously obtain the measured ambient temperature timing signal of the package cavity where the MEMS thermopile sensor is located. Combine the measured thermoelectric potential timing signal and the measured ambient temperature timing signal into a set of real-time working status data of the sensor.

[0015] The edge microcontroller drives an analog-to-digital converter at a fixed sampling frequency to sample the voltage at the thermoelectric potential output of the MEMS thermopile sensor. The sampled value is converted into a measured thermoelectric potential value Vmeas(t) after gain calibration by an instrumentation amplifier, in millivolts. Simultaneously, it drives a digital temperature sensor integrated within the package cavity to read the internal air temperature Tenv(t), in degrees Celsius. The edge microcontroller assigns a Coordinated Universal Time (UTC) timestamp to each sampling time t, organizing Vmeas(t) and Tenv(t) into sampling point data tuples (Vmeas(t), Tenv(t)). All sampling point data tuples are arranged in ascending order of timestamps to form the sensor's real-time operating status data set.

[0016] Step S120: Call the digital twin instance that is isomorphic to the physical entity of the thermopile sensor of the microelectromechanical system, load the real-time working status data set of the sensor into the thermoelectric coupling simulation model of the digital twin instance, and simulate and deduce the thermoelectric conversion physical process of the thermopile sensor of the microelectromechanical system through the thermoelectric coupling simulation model to generate the simulated thermoelectric potential time-series signal output by the digital twin instance.

[0017] The edge microcontroller uploads the real-time operating status data set of the sensors to the cloud-based digital twin engine via a wireless communication gateway. The cloud-based digital twin engine searches for the corresponding digital twin instance in the digital twin instance registry based on the sensor's unique device identifier, uses the sensor's real-time operating status data set as boundary condition input, and drives the thermo-electric coupling simulation model of the digital twin instance to perform simulation deduction.

[0018] Step S121: Analyze the thermal resistance network topology between the cold and hot ends of the sensor in the thermoelectric coupling simulation model of the digital twin instance. The thermal resistance network topology is composed of multiple thermal resistance nodes and multiple heat conduction paths. Each thermal resistance node corresponds to the thermal resistance attribute of the thermocouple pair in the thermopile sensor of the microelectromechanical system at its spatial location.

[0019] The thermoelectric coupling simulation model of the digital twin instance maintains a thermal resistance network topology data structure internally. The thermal resistance network topology is stored as a node-edge graph data structure. The graph node set includes cold-end node Ncold, hot-end node Nhot, thermocouple pair node sequence {Nth(1), Nth(2), ..., Nth(K)}, and package node sequence {Npkg(1), Npkg(2), ..., Npkg(M)}, where K is the total number of thermocouple pairs in the sensor, and M is the total number of thermal resistance nodes in the package. The graph edge set stores the heat conduction paths between nodes. Each edge E(i,j) connects the i-th node and the j-th node. The weight attribute of the edge is the thermal conductivity value Gij, which is equal to 1 divided by the thermal resistance value Rij. The initial value of the thermal resistance value Rij is calculated from the thermal conductivity, geometric dimensions, and contact area of ​​the sensor material at the time of manufacture.

[0020] Step S122: Extract the cavity internal temperature value from the measured ambient temperature time-series signal at the current acquisition time from the sensor real-time working status data set, determine the cavity internal temperature value as the transient boundary temperature condition of the sensor cold end node in the thermal resistance network topology, and use it as the reference parameter for inverting and calculating the hot end temperature.

[0021] Extract the Tenv(t) value from the sampling point data tuple (Vmeas(t), Tenv(t)) at the current acquisition time t from the sensor's real-time operating status data set. Assign the temperature boundary condition Tcold(t) of the cold end node Ncold to Tenv(t). This temperature boundary condition serves as the temperature reference point for the entire thermal resistance network topology, and the temperature values ​​of all subsequent nodes are iteratively calculated with reference to this.

[0022] Step S123: Extract the potential difference value from the measured thermoelectric potential time-series signal at the current acquisition time from the sensor real-time working status data set. Combine the transient boundary temperature condition of the cold end node of the sensor and the preset Seebeck coefficient, calculate the equivalent temperature estimate of the hot end node of the sensor by inverting the potential difference value, and determine the equivalent temperature estimate as the transient boundary temperature condition of the hot end node of the sensor in the thermal resistance network topology.

[0023] The Vmeas(t) value is extracted from the real-time operating status data set of the sensor. Based on the fundamental Seebeck effect equation V = N × α × (Thot - Tcold), where N is the number of thermocouple pairs and α is the Seebeck coefficient scaling factor in millivolts per degree Celsius, the estimated equivalent temperature of the hot junction, Thot(t) = Tcold(t) + Vmeas(t) / (N × α), is calculated. The temperature boundary condition Thot(t) at the hot junction node Nhot is then assigned the value Thot(t).

[0024] Step S124: Based on the transient boundary temperature conditions of the cold end node and the hot end node of the sensor, drive the thermal resistance network topology to iteratively solve the heat flux density distribution along each heat conduction path to obtain the node temperature distribution state of each thermal resistance node in the thermal resistance network topology at the current acquisition time.

[0025] Using Tcold(t) and Thot(t) as the first type of boundary conditions, the steady-state thermal balance equations of the thermal resistance network topology are iteratively solved. For any node i in the thermal resistance network topology other than the cold and hot nodes, its steady-state thermal balance equation is ΣGij×(Ti-Tj)=0, where the summation range j is all adjacent nodes connected to node i by an edge. A sparse linear equation system is constructed by simultaneously solving the thermal balance equations of all internal nodes, and the over-relaxation iterative method is used for numerical solution. The convergence condition of the iteration is that the maximum absolute value of the difference between the temperature values ​​of all nodes in two consecutive iterations is less than a preset convergence threshold. After iterative convergence, the node temperature distribution state vector Tnodes(t) of all internal nodes at the current acquisition time t is obtained.

[0026] Step S125: Based on the node temperature distribution and the Seebeck coefficient distribution parameters of the thermocouple pair, perform Seebeck voltage accumulation calculation along the thermal resistance nodes covered by the thermocouple pair to generate the simulated thermoelectric potential value of the digital twin instance at the current acquisition time.

[0027] For each thermocouple pair k, it is connected between two nodes in the thermal resistance network topology, with temperatures Tnode_a(t) and Tnode_b(t) respectively. The local Seebeck voltage of the k-th thermocouple pair is Vth(k) = α(k) × (Tnode_a(t) - Tnode_b(t)), where α(k) is the local Seebeck coefficient of the k-th thermocouple pair. The local Seebeck voltages of all thermocouple pairs are cascaded and summed to simulate the thermoelectric potential Vsim(t) = ΣVth(k), with the summation range k from 1 to K.

[0028] Step S126: Arrange the simulated thermoelectric potential values ​​at each acquisition time in chronological order to form a simulated thermoelectric potential time sequence signal.

[0029] The simulated thermoelectric potential values ​​Vsim(t) of all acquisition times t1, t2, ..., tN within the current working period are organized into a one-dimensional sequence in ascending order of time, which is used as the simulated thermoelectric potential time series signal.

[0030] Step S130: Perform signal difference analysis on the measured thermoelectric potential timing signal and the simulated thermoelectric potential timing signal to obtain the thermoelectric potential drift characteristic curve that reflects the output deviation between the physical entity of the thermopile sensor of the microelectromechanical system and the digital twin instance.

[0031] The cloud-based digital twin engine compares the measured thermoelectric potential time-series signal and the simulated thermoelectric potential time-series signal point by point, and extracts the output drift characteristics of the sensor physical entity relative to its ideal digital twin.

[0032] Step S131: Perform point-by-point difference calculation between the potential difference values ​​at each acquisition time in the measured thermoelectric potential time series signal and the simulated thermoelectric potential values ​​at the same acquisition time in the simulated thermoelectric potential time series signal to generate a deviation signal sequence. Each deviation signal point in the deviation signal sequence corresponds to an acquisition time.

[0033] For each acquisition time t, the deviation value D(t) = Vmeas(t) - Vsim(t) is calculated. The deviation values ​​D(t) for all acquisition times are arranged in ascending order of time to form a one-dimensional deviation signal sequence. The deviation signal sequence reflects the absolute difference in magnitude between the actual output of the sensor and the simulated output of the digital twin.

[0034] Step S132: Perform time-division processing on the deviation signal sequence, aggregate adjacent deviation signal points with continuous deviation signal change trends into deviation signal segment units, and determine the trend inflection boundary points between adjacent deviation signal segment units.

[0035] The deviation signal sequence D(t) is subjected to first-order difference calculation, resulting in the first-order difference sequence ΔD(t) = D(t + Δt) - D(t), where Δt is the sampling time interval. The sign change points of ΔD(t) are detected, defined as the time points where the signs of ΔD(t) and ΔD(t + Δt) are opposite. These sign change points are the trend reversal boundary points. Deviation signal points between two adjacent trend reversal boundary points constitute a deviation signal segment unit.

[0036] Step S133: Perform feature extraction on each deviation signal segment unit, extracting the starting amplitude value, ending amplitude value, and time span of the acquisition time across the deviation signal segment unit.

[0037] For the m-th deviation signal segment unit, it contains all deviation signal points within the time interval [tstart(m), tend(m)]. The starting amplitude value is Dstart(m) = D(tstart(m)). The ending amplitude value is Dend(m) = D(tend(m)). The time span is Tspan(m) = tend(m) - tstart(m).

[0038] Step S134: Calculate the amplitude drift rate descriptor and amplitude drift direction descriptor of each deviation signal segment unit based on the starting amplitude value, ending amplitude value, and time span.

[0039] The amplitude drift rate descriptor is Rdrift(m) = (Dend(m) - Dstart(m)) / Tspan(m). The amplitude drift direction descriptor is Sdrift(m) = sign(Dend(m) - Dstart(m)), where sign is the sign function, with a value of +1 indicating positive drift, a value of -1 indicating negative drift, and a value of 0 indicating no drift.

[0040] Step S135: Combine the amplitude drift rate descriptors and amplitude drift direction descriptors of all deviation signal segment units in chronological order to generate a thermoelectric potential drift characteristic curve. The thermoelectric potential drift characteristic curve describes the deviation evolution trajectory of the output of the physical entity of the thermopile sensor of the microelectromechanical system relative to the output of the digital twin instance in a segmented manner.

[0041] Each deviation signal segment unit's Rdrift(m), Sdrift(m), and corresponding time interval [tstart(m), tend(m)] are encapsulated into a piecewise drift description record. All records are organized in chronological order into a thermoelectric potential drift characteristic curve data structure, which is a set of piecewise linear functions.

[0042] Step S140: Use the thermoelectric potential drift characteristic curve to perform parameter reverse correction processing on the current response characteristic function of the microelectromechanical system thermopile sensor, generate the updated response characteristic function of the microelectromechanical system thermopile sensor, and replace the original response characteristic function of the microelectromechanical system thermopile sensor with the updated response characteristic function to complete the self-calibration process.

[0043] The cloud-based digital twin engine reverse-corrects the sensor's response characteristic function parameters based on the sensor's aging trend reflected by the thermoelectric potential drift characteristic curve, and sends the corrected parameters to the edge microcontroller to complete sensor calibration.

[0044] Step S141: Analyze the function structure type and the set of parameters to be corrected of the current response characteristic function of the thermopile sensor of the microelectromechanical system. The set of parameters to be corrected includes the Seebeck coefficient scaling factor and the cold junction temperature compensation bias.

[0045] The response characteristic function of a MEMS thermopile sensor is Vout = αeff × (Thot - Tcold) + Voffset, where αeff is the Seebeck coefficient scaling factor and Voffset is the cold junction temperature compensation bias. The set of parameters to be corrected contains the current values ​​of both αeff and Voffset. The cloud-based digital twin engine reads the current αeff and Voffset values ​​from the parameter storage area of ​​the digital twin instance.

[0046] Step S142: Extract the amplitude drift rate descriptor and amplitude drift direction descriptor of each deviation signal segment unit from the thermoelectric potential drift characteristic curve, map the amplitude drift rate descriptor to the attenuation correction amount of the Seebeck coefficient scaling factor, and map the amplitude drift direction descriptor to the offset correction amount of the cold junction temperature compensation bias.

[0047] The attenuation correction for the Seebeck coefficient scaling factor is Δαeff(m) = -β × Rdrift(m) × Tspan(m) / (Thot_avg(m) - Tcold_avg(m)), where Thot_avg(m) and Tcold_avg(m) are the estimated hot-end temperature and the average cold-end temperature within the time interval corresponding to the m-th bias signal segment unit, respectively, and β is the attenuation mapping coefficient. The offset correction for the cold-end temperature compensation bias is ΔVoffset(m) = Rdrift(m) × Tspan(m) × Sdrift(m).

[0048] Step S143: Based on the attenuation correction and offset correction corresponding to each deviation signal segment unit, the Seebeck coefficient scaling factor and cold junction temperature compensation bias in the set of parameters to be corrected are corrected segment by segment to generate the Seebeck coefficient scaling factor sequence and cold junction temperature compensation bias sequence after segment correction.

[0049] The piecewise correction value of the Seebeck coefficient scaling factor is αeff(m) = αeff(m-1) + Δαeff(m), where αeff(0) is the αeff value of the current response characteristic function. The piecewise correction value of the cold junction temperature compensation offset is Voffset(m) = Voffset(m-1) + ΔVoffset(m), where Voffset(0) is the Voffset value of the current response characteristic function. After all deviation signal segment units are calculated, the last piecewise correction value is taken as the final correction value.

[0050] Step S144: Using the piecewise corrected Seebeck coefficient scaling factor sequence and cold junction temperature compensation bias sequence, perform piecewise parameter replacement on the current response characteristic function to generate an updated response characteristic function.

[0051] Replace αeff in the response characteristic function Vout=αeff×(Thot-Tcold)+Voffset with the final corrected value αeff, and replace Voffset with the final corrected value Voffset to obtain the updated response characteristic function.

[0052] Step S145: Erase the parameter table of the original response characteristic function stored in the internal storage unit of the MEMS thermopile sensor, and write the parameter table of the updated response characteristic function into the internal storage unit to complete the replacement operation of the original response characteristic function with the updated response characteristic function.

[0053] The cloud-based digital twin engine sends a self-calibration command to the edge microcontroller via a wireless communication gateway. This command includes updated Seebeck coefficient scaling factor values ​​and updated cold junction temperature compensation bias values. Upon receiving the self-calibration command, the edge microcontroller erases the original flash memory sector containing the response characteristic function parameter table in its internal storage unit and rewrites the updated parameter values ​​into the same sector, completing the self-calibration operation.

[0054] Step S210: Obtain the aging characteristic data of the cavity packaging material of the microelectromechanical system thermopile sensor. The cavity packaging material aging characteristic data includes the sequence of changes in the transmittance of the packaging material and the sequence of changes in the thermal conductivity of the packaging material.

[0055] The cloud-based digital twin engine retrieves accelerated aging test data from the packaging material aging database, which is the same type of packaging material as the MEMS thermopile sensor's packaging cavity. The packaging material transmittance change sequence τ(tacc) represents the infrared window transmittance values ​​measured at different accelerated aging times tacc, and the packaging material thermal conductivity change sequence λ(tacc) represents the thermal conductivity values ​​of the packaging shell measured at the corresponding aging times.

[0056] Step S211: Convert the sequence of changes in the transmittance of the encapsulation material into the flux attenuation factor sequence of the radiative heat flux input channel in the digital twin instance, and convert the sequence of changes in the thermal conductivity of the encapsulation material into the time-varying parameter sequence of the thermal resistance nodes of the encapsulation in the thermal resistance network topology of the thermoelectric coupling simulation model.

[0057] The accelerated aging time is mapped to the actual service equivalent time based on the sensor's service life and operating temperature profile. The flux attenuation factor ρ(t) = τ(t) / τ(0) of the radiative heat flux input channel, where τ(t) is the transmittance corresponding to the equivalent time t, and τ(0) is the initial transmittance. The time-varying thermal resistance parameter Rpkg(t) of the thermal resistance node of the package is Rpkg(t) = Lpkg / (λ(t) × Apkg), where Lpkg is the thermal conduction path length of the package, and Apkg is the effective thermal conductivity cross-sectional area.

[0058] Step S212: Update the heat transfer parameters of the corresponding path in the thermoelectric coupling simulation model using the flux decay factor sequence and the thermal resistance time-varying parameter sequence to generate an aging-sensing thermoelectric coupling simulation model.

[0059] The thermal resistance value of the edge associated with the encapsulation node Npkg in the thermal resistance network topology is updated to Rpkg(t), and the radiative heat flux input of the hot end node Nhot is multiplied by the flux attenuation factor ρ(t). The updated thermal resistance network topology is the aging-sensing thermoelectric coupling simulation model.

[0060] Step S213: Replace the previous thermoelectric coupling simulation model with an aging-sensing thermoelectric coupling simulation model, perform simulation deduction, and generate a simulated thermoelectric potential time series signal after aging compensation.

[0061] In subsequent self-calibration cycles, the aging-sensing thermoelectric coupling simulation model is used to execute the simulation deduction process of steps S124 to S126, generating a simulated thermoelectric potential timing signal after aging compensation of the packaging material.

[0062] Step S310: Arrange the multiple thermoelectric potential drift characteristic curves generated by the digital twin instance in a time series over multiple consecutive self-calibration cycles to form a sensor lifecycle drift characteristic sequence.

[0063] The cloud-based digital twin engine stores the thermoelectric potential drift characteristic curves generated during each self-calibration process in the full lifecycle drift database of the digital twin instance, ordered by the self-calibration cycle timestamps. All thermoelectric potential drift characteristic curves are read in chronological order and organized into a sensor full lifecycle drift characteristic sequence.

[0064] Step S311: Extract the amplitude drift rate descriptor and amplitude drift direction descriptor of each thermoelectric potential drift characteristic curve in the sensor's full life cycle drift characteristic sequence, and establish the change trajectory of the amplitude drift rate descriptor and the change trajectory of the amplitude drift direction descriptor over time.

[0065] For each thermoelectric potential drift characteristic curve, the average absolute value of the amplitude drift rate descriptor of all deviation signal segment units is calculated as the average amplitude drift rate for that self-calibration cycle. The number of deviation signal segment units with amplitude drift direction descriptors of +1 and -1 within that self-calibration cycle is counted, and the direction distribution ratio is calculated. The average amplitude drift rate and direction distribution ratio of each cycle are arranged in chronological order to form a change trajectory.

[0066] Step S312: Based on the trajectory of the amplitude drift rate descriptor over time, predict the drift acceleration trend of the thermopile sensor of the microelectromechanical system during future operating periods.

[0067] The trajectory of the average amplitude drift rate over time is fitted using a quadratic polynomial, with the fitting function being Ravg(t) = a × t² + b × t + c. The fitting coefficient 'a' represents the drift acceleration, and 'a > 0' indicates that the drift is accelerating and worsening.

[0068] Step S313: Based on the trajectory of the amplitude drift direction descriptor over time, predict the probability of the drift direction reversal of the thermopile sensor in the future operating period.

[0069] Calculate the first-order difference sequence of the directional distribution ratio over time. If the sequence shows a trend of gradually transitioning from a bias of +1 to a bias of -1 or vice versa, then the probability of drift direction reversal is a function of the rate of change of the directional ratio and the number of consecutive cycles in the same direction. The reversal probability Prev = 1 / (1 + exp(-k × (Δratio - ratio_th))), where Δratio is the change in the directional ratio between the last two cycles, and k and ratio_th are preset parameters.

[0070] Step S314: Based on the drift acceleration trend and the drift direction reversal probability, generate a pre-failure risk indication signal for the thermopile sensor of the microelectromechanical system.

[0071] If the drift acceleration exceeds a preset acceleration threshold or the drift direction reversal probability exceeds a preset reversal probability threshold, a pre-failure risk indication signal is generated. The pre-failure risk indication signal includes the sensor device identifier, the current drift acceleration value, the drift direction reversal probability value, and a recommended preventive replacement time window.

[0072] Step S410: When the MEMS thermopile sensor is not in operation, apply a standard temperature difference excitation signal of known amplitude to the MEMS thermopile sensor. The standard temperature difference excitation signal is generated by a calibration heat source placed on the hot end side of the MEMS thermopile sensor.

[0073] The cloud-based digital twin engine, based on the sensor's operational cycle plan, triggers the calibration heat source to power on via an edge microcontroller during the sensor's non-operational periods. The calibration heat source is a miniature resistance heating element integrated within the sensor's package cavity. When powered on, it generates a fixed thermal radiation power, establishing a known standard temperature difference excitation signal ΔTref between the sensor's hot and cold ends.

[0074] Step S411: Acquire the standard excitation thermoelectric potential signal of the microelectromechanical system thermopile sensor under the action of the standard temperature difference excitation signal.

[0075] After the standard temperature difference excitation signal reaches a steady state, the edge microcontroller acquires the thermoelectric potential output of the sensor at a fixed sampling frequency, and the sampled data is organized into a standard excitation thermoelectric potential signal Vref_meas(t).

[0076] Step S412: Obtain the simulated standard excitation thermoelectric potential signal of the digital twin instance under the same boundary conditions of the standard temperature difference excitation signal.

[0077] The cloud-based digital twin engine uses the standard temperature difference excitation signal ΔTref as the boundary condition of the thermoelectric coupling simulation model. It sets the cold end temperature of the sensor to the current measured ambient temperature and the hot end temperature of the sensor to the cold end temperature plus ΔTref. It executes the simulation derivation process from steps S124 to S126 to generate the simulation standard excitation thermoelectric potential signal Vref_sim(t).

[0078] Step S413: Analyze the signal difference between the standard excitation thermoelectric potential signal and the simulated standard excitation thermoelectric potential signal to obtain the standard excitation drift characteristic curve.

[0079] Perform signal difference analysis processing steps S131 to S135 on Vref_meas(t) and Vref_sim(t) to obtain the standard excitation drift characteristic curve.

[0080] Step S414: Use the standard excitation drift characteristic curve to correct the parameters of the current response characteristic function of the MEMS thermopile sensor, and write the corrected response characteristic function as the initial response characteristic function of the sensor into the storage unit of the MEMS thermopile sensor.

[0081] The parameter reverse correction process in steps S141 to S144 is performed using the standard excitation drift characteristic curve to generate the corrected response characteristic function parameters. This corrected response characteristic function is used as the initial response characteristic function of the sensor and is sent and written to the sensor's internal storage unit via a self-calibration command.

[0082] Step S510: Obtain a cluster of digital twin instances of multiple sensors from the same batch as the MEMS thermopile sensor in the digital twin instance. The multiple sensors share the same thermoelectric coupling simulation model structure with the MEMS thermopile sensor.

[0083] The cloud-based digital twin engine queries the digital twin instance registry for all sensor device identifiers from the same production batch as the current sensor, and retrieves the corresponding digital twin instances for each of these sensors. Because they belong to the same production batch, these sensor digital twin instances share a single topology template for a thermoelectric coupling simulation model.

[0084] Step S511: Extract the batch thermoelectric potential drift characteristic curve generated by each digital twin instance of a sensor during its respective working period from the cluster of digital twin instances of multiple sensors.

[0085] Traverse the full lifecycle drift database of each digital twin instance of the same batch of sensors, extract the thermoelectric potential drift characteristic curves generated during the same working period of the current sensor, and organize them into a set of batch thermoelectric potential drift characteristic curves.

[0086] Step S512: Compare the batch thermoelectric potential drift characteristic curves with the thermoelectric potential drift characteristic curves of the MEMS thermopile sensor to identify the degree of group drift deviation of the MEMS thermopile sensor relative to multiple sensors.

[0087] Calculate the mean μbatch and standard deviation σbatch of the amplitude drift rate descriptors for each deviation signal segment unit of the same batch of sensors within the same time period. The current group drift deviation degree of the sensor is Zscore = (Rcurr - μbatch) / σbatch, where Rcurr is the amplitude drift rate descriptor of the current sensor for the corresponding time period.

[0088] Step S513: Based on the degree of population drift deviation, correct the amplitude drift rate descriptor in the thermoelectric potential drift characteristic curve to generate the population-corrected thermoelectric potential drift characteristic curve.

[0089] If the absolute value of Zscore exceeds the preset deviation threshold, it indicates that the current sensor has an abnormal individual bias. The corrected amplitude drift rate descriptor Rcorr = Rcurr - γ × (Rcurr - μbatch), where γ is the population correction weight coefficient. Replace Rdrift(m) in the original thermoelectric potential drift characteristic curve with Rcorr to generate the population-corrected thermoelectric potential drift characteristic curve.

[0090] Step S514: Use the thermoelectric potential drift characteristic curve after group correction to perform parameter reverse correction processing on the current response characteristic function of the thermopile sensor of the microelectromechanical system.

[0091] The original thermoelectric potential drift characteristic curve is replaced with the thermoelectric potential drift characteristic curve after group correction, and the parameter reverse correction process in step S140 is executed.

[0092] Step S610: Perform frequency domain transformation on the thermoelectric potential drift characteristic curve to convert the thermoelectric potential drift characteristic curve from time domain representation to frequency domain representation, and obtain the drift frequency domain characteristic spectrum.

[0093] Perform a Fast Fourier Transform on the deviation signal sequence D(t) in the thermoelectric potential drift characteristic curve to obtain the spectral sequence F(f). The drift frequency domain characteristic spectrum is a mapping table of frequency f and corresponding spectral amplitude |F(f)|.

[0094] Step S611: Extract the dominant frequency component and the frequency amplitude of the corresponding dominant frequency component from the drift frequency domain feature spectrum. The dominant frequency component is the frequency point in the drift frequency domain feature spectrum whose frequency amplitude exceeds the amplitude threshold.

[0095] Set the amplitude threshold Fth = μamp + k × σamp, where μamp is the mean amplitude of all frequency points, σamp is the standard deviation, and k is the threshold factor. Select frequency points where |F(f)|>Fth as the dominant frequency components.

[0096] Step S612: Compare and match the dominant frequency component and the frequency amplitude of the corresponding dominant frequency component with the known periodic environmental disturbance frequency of the working environment of the MEMS thermopile sensor.

[0097] The system retrieves a list of known periodic environmental disturbance frequencies from a knowledge base of periodic disturbance frequencies in the sensor installation environment. This list includes the daytime fluctuation frequency of ambient temperature and the periodic switching frequency of the heating equipment. Each dominant frequency component is matched against a known periodic environmental disturbance frequency. If the frequency difference is less than a preset matching tolerance, the match is considered successful.

[0098] Step S613: Based on the matching results, separate the periodic drift component caused by periodic environmental disturbance and the monotonic drift component caused by sensor aging from the thermoelectric potential drift characteristic curve.

[0099] The dominant frequency component that successfully matches the known periodic environmental disturbance frequency and its corresponding spectral amplitude are restored to the time-domain periodic drift component Dper(t) using inverse fast Fourier transform. Subtracting Dper(t) point-by-point from the original bias signal sequence D(t) yields the monotonic drift component Dmono(t).

[0100] Step S614: Perform parameter inverse correction processing on the current response characteristic function of the MEMS thermopile sensor using only the monotonic drift component.

[0101] Using Dmono(t) as the new deviation signal sequence, the processing flow from step S131 to step S140 is executed, and the parameters are corrected only based on the monotonic drift component caused by the aging of the sensor itself.

[0102] For example, step S710: acquire multiple measured thermoelectric potential timing signals of multiple similar sensors of the microelectromechanical system thermopile sensor during the same working period, and the multiple similar sensors are deployed in the same temperature measurement environment as the microelectromechanical system thermopile sensor.

[0103] The cloud-based digital twin engine queries the measured thermoelectric potential time-series signal records of other similar MEMS thermopile sensors in the same temperature measurement environment during the same working period through the industrial wireless communication gateway, and organizes them into a set of multiple measured thermoelectric potential time-series signals.

[0104] Step S711: Perform cross-correlation analysis on the multi-channel measured thermoelectric potential time series signals to establish the cross-correlation matrix between the multi-channel measured thermoelectric potential time series signals.

[0105] For the measured thermoelectric potential time-series signals of the i-th and j-th channels, calculate the Pearson cross-correlation coefficient Corr(i,j) = Cov(Vi,Vj) / (σi×σj), where Cov is the covariance and σ is the standard deviation. The cross-correlation coefficients of all sensor pairs are organized into a cross-correlation coefficient matrix.

[0106] Step S712: Determine the degree of deviation in sensor output consistency between the thermopile sensor of the microelectromechanical system and multiple similar sensors based on the cross-correlation coefficient matrix.

[0107] Calculate the mean cross-correlation coefficient (Corr_mean) between the current sensor and all similar sensors. The deviation in output consistency between sensors, Dcons = 1 - Corr_mean.

[0108] Step S713: Map the degree of output consistency deviation between sensors to the confidence weight of the thermoelectric potential drift characteristic curve, and use the confidence weight to adjust the amplitude drift rate descriptor of the thermoelectric potential drift characteristic curve to generate a consistency weighted drift characteristic curve.

[0109] The confidence weight is Wcons = exp(-η × Dcons), where η is the attenuation coefficient. The consistency-weighted amplitude drift rate descriptor is Rw(m) = Wcons × Rdrift(m). Replacing the amplitude drift rate descriptor in the original thermoelectric potential drift characteristic curve with Rw(m) generates the consistency-weighted drift characteristic curve.

[0110] Step S714: Use the uniform weighted drift characteristic curve to perform parameter reverse correction processing on the current response characteristic function of the thermopile sensor of the microelectromechanical system.

[0111] The original thermoelectric potential drift characteristic curve is replaced by a uniform weighted drift characteristic curve, and the parameter reverse correction process in step S140 is executed.

[0112] Step S810: Perform sensitivity analysis on the thermal resistance network topology between the cold end and hot end of the sensor in the thermoelectric coupling simulation model, and extract the sensitivity coefficient of the thermal resistance value of each thermal resistance node in the thermal resistance network topology to the output heat flux density distribution of the thermal resistance network topology.

[0113] For each thermal resistance node in the thermal resistance network topology, its thermal resistance value Rnode is increased by a preset small percentage perturbation δR. After the perturbation, the heat flux density distribution Qpert is recalculated, and the heat flux density distribution without perturbation is Qbase. The sensitivity coefficient Sens of this node is Sens = |Qpert - Qbase| / δR, where |·| represents the L2 norm.

[0114] Step S811: Arrange the sensitivity coefficients in descending order and select the set of sensitive thermal resistance nodes with the highest sensitivity coefficients. The contribution of each sensitive thermal resistance node in the set of sensitive thermal resistance nodes to the output heat flux density distribution exceeds that of the non-sensitive thermal resistance nodes.

[0115] Sort all thermal resistance nodes in descending order of their sensitivity coefficients. Select the nodes whose sum of sensitivity coefficients exceeds a preset percentage of the total sensitivity coefficients to form a set of sensitive thermal resistance nodes.

[0116] Step S812: In the digital twin instance, inject a small perturbation into the thermal resistance value of each sensitive thermal resistance node in the set of sensitive thermal resistance nodes to generate multiple sets of thermal resistance parameter perturbation combinations. Each set of thermal resistance parameter perturbation combinations corresponds to a set of perturbation offsets of the thermal resistance value of each sensitive thermal resistance node in the set of sensitive thermal resistance nodes.

[0117] The Latin hypercube sampling method is used to generate multiple sets of perturbation offsets within a preset range of thermal resistance values ​​of each node in the sensitive thermal resistance node set. Each set of perturbation offsets is denoted as {ΔR(1), ΔR(2), ..., ΔR(S)}, where S is the number of sensitive thermal resistance nodes.

[0118] Step S813: For each combination of thermal resistance parameter disturbances, drive the thermoelectric coupling simulation model to re-perform the simulation and generate the time series signal of the disturbance simulation thermoelectric potential corresponding to that combination of thermal resistance parameter disturbances.

[0119] Update the thermal resistance value of the s-th node in the set of sensitive thermal resistance nodes to R(s) + ΔR(s), while keeping the thermal resistance values ​​of other nodes in the thermal resistance network topology unchanged. Re-execute the simulation derivation in the manner of step S120 to obtain the time-series signal of the disturbance simulation thermoelectric potential corresponding to this set of disturbances.

[0120] Step S814: Compare the time series signals of the simulated thermoelectric potential corresponding to each group of thermal resistance parameter perturbation combinations with the time series signals of the simulated thermoelectric potential without perturbation, and generate the simulation output offset corresponding to each thermal resistance parameter perturbation combination.

[0121] For the p-th perturbation combination, calculate the sum of squares and the mean of the differences between the simulated thermoelectric potential time series signal under perturbation and the simulated thermoelectric potential time series signal without perturbation at each time point, and use it as the simulation output offset Dev(p).

[0122] Step S815: Select the thermal resistance parameter perturbation combination with the highest degree of matching between the deviation amplitude of the simulation output offset and the drift characteristic curve of the thermoelectric potential from multiple thermal resistance parameter perturbation combinations, and use it as the optimal thermal resistance parameter perturbation combination.

[0123] Calculate the Euclidean distance between the simulated output offset Dev(p) of each perturbation combination and the mean value of the amplitude drift rate descriptor of all deviation signal segment units in the thermoelectric potential drift characteristic curve. Select the perturbation combination with the smallest Euclidean distance as the optimal thermal resistance parameter perturbation combination.

[0124] Step S816: Determine the perturbation offset of the thermal resistance value of each sensitive thermal resistance node in the optimal thermal resistance parameter perturbation combination as the thermal resistance aging offset of the corresponding sensitive thermal resistance node in the thermal resistance network topology.

[0125] The perturbation offset ΔRopt(s) in the optimal thermal resistance parameter perturbation combination is the thermal resistance aging offset of the s-th sensitive thermal resistance node. This offset is then written into the thermal resistance aging record table of the digital twin instance.

[0126] Step S817: Based on the thermal resistance aging offset of each sensitive thermal resistance node, permanently update the thermal resistance parameters of the corresponding sensitive thermal resistance nodes in the thermal resistance network topology to generate the thermal resistance network topology after physical aging synchronization.

[0127] In the thermal resistance network topology, the thermal resistance values ​​of each node in the set of sensitive thermal resistance nodes are permanently replaced with the original thermal resistance value plus the corresponding thermal resistance aging offset. The updated thermal resistance network topology is the thermal resistance network topology after physical aging synchronization.

[0128] Step S818: Replace the previous thermoelectric coupling simulation model with a thermoelectric coupling simulation model that includes the thermal resistance network topology after physical aging synchronization, so that the thermal resistance parameter state of the digital twin instance is synchronized with the thermal resistance aging state of the physical entity of the MEMS thermopile sensor.

[0129] The reference pointer of the thermal resistance network topology in the thermoelectric coupling simulation model of the digital twin instance is set to the thermal resistance network topology after physical aging synchronization. Subsequent simulations are all performed based on this synchronized model.

[0130] Step S819: Based on the synchronized digital twin instance, perform simulation deduction in subsequent self-calibration cycles and generate a simulated thermoelectric potential timing signal that is synchronized with the thermal resistance aging state of the sensor physical entity.

[0131] In subsequent self-calibration cycles, the simulation of step S120 is performed using a synchronized digital twin instance. The generated simulated thermoelectric potential timing signal has taken into account the influence of the aging of the sensor's physical thermal resistance on the thermal field distribution.

[0132] Step S910: Perform multi-scale decomposition processing on the thermoelectric potential drift characteristic curve, decomposing the thermoelectric potential drift characteristic curve into a superposition of slow drift trend component, fast drift fluctuation component and residual drift noise component.

[0133] An integrated empirical mode decomposition (EMD) algorithm is applied to the deviation signal sequence D(t) in the thermoelectric potential drift characteristic curve. The EMD algorithm performs empirical mode decomposition by repeatedly superimposing white noise of different amplitudes onto the original signal, and takes the average of the multiple decomposition results as the final decomposition result. The decomposition yields multiple intrinsic mode function components, arranged from low to high frequency. The low-frequency components correspond to the slow drift trend component Dslow(t), the mid-frequency component combination corresponds to the fast drift fluctuation component Dfast(t), and the high-frequency components correspond to the residual drift noise component Dnoise(t).

[0134] Step S911: Extract the trend morphology features of the slow drift trend component. The trend morphology features include the monotonicity direction indicator and curvature change pattern of the slow drift trend component.

[0135] First-order difference sign statistics are performed on Dslow(t), counting the number of sampling points Npos with positive signs and Nneg with negative signs. The monotonicity direction is identified by the direction corresponding to the larger of Npos and Nneg. Second-order difference calculation is performed on Dslow(t), and the intervals where the second-order difference is continuously positive and continuously negative are extracted as curvature variation patterns.

[0136] Step S912: Extract the wave morphology features of the rapidly changing drift wave component. The wave morphology features include the wave amplitude envelope and wave period distribution density of the rapidly changing drift wave component.

[0137] Perform a Hilbert transform on Dfast(t) to calculate the instantaneous amplitude; the smoothed envelope of the instantaneous amplitude is the fluctuation amplitude envelope. Perform statistical analysis on the zero-crossing interval sequence of Dfast(t) to obtain the fluctuation period distribution density function.

[0138] Step S913: Analyze the first correlation mapping relationship between the trend morphology characteristics of the slowly varying drift trend component and the degree of oxidation on the hot end surface of the sensor of the microelectromechanical system thermopile sensor physical entity.

[0139] In the sensor digital twin example, a physical correlation model is established between the oxidation degree Ox of the hot end surface and the thermal resistance of the hot end node: Rhot_ox = Rhot × (1 + kox × Ox), where kox is the oxidation thermal resistance growth coefficient. Through simulation, the slow-varying drift trend component characteristics corresponding to different oxidation degrees are obtained, and a reverse mapping function from the trend morphology characteristics to the oxidation degree is fitted.

[0140] Step S914: Analyze the second correlation mapping relationship between the wave morphology characteristics of the rapidly changing drift wave component and the intensity of airflow disturbance inside the encapsulation cavity of the thermopile sensor physical entity of the microelectromechanical system.

[0141] In the sensor digital twin example, a physical correlation model is established between the airflow disturbance intensity Gust inside the encapsulation cavity and the temperature fluctuation variance σcold² at the cold end node. Simulations are used to obtain the characteristics of the rapidly changing drift fluctuation components under different disturbance intensities, and a reverse mapping function from the fluctuation morphology characteristics to the airflow disturbance intensity is obtained through fitting.

[0142] Step S915: Based on the first correlation mapping relationship, convert the trend morphology characteristics of the slow drift trend component into an estimated value of the oxidation state of the sensor hot end surface.

[0143] The monotonicity direction identifier and curvature change pattern extracted in step S911 are input into the inverse mapping function of the first correlation mapping relationship, and the estimated value of the oxidation state of the sensor hot end surface, Ox_est, is output.

[0144] Step S916: Based on the second correlation mapping relationship, convert the wave morphology characteristics of the fast-changing drift wave component into an estimated value of the airflow disturbance intensity inside the encapsulation cavity.

[0145] Input the fluctuation amplitude envelope and fluctuation period distribution density extracted in step S912 into the inverse mapping function of the second correlation mapping relationship, and output the estimated value of the airflow disturbance intensity inside the encapsulation cavity, Gust_est.

[0146] Step S917: Simultaneously use the estimated value of the oxidation state of the sensor hot end surface and the estimated value of the airflow disturbance intensity inside the packaging cavity as auxiliary correction factors to perform multi-dimensional parameter linkage correction on the current response characteristic function of the MEMS thermopile sensor, and generate a response characteristic function jointly compensated for oxidation and airflow disturbance.

[0147] The Seebeck coefficient scaling factor oxidation correction Δαox = -kalpha × Ox_est, where kalpha is the oxidation attenuation coefficient. The cold end temperature compensation bias perturbation correction ΔVgust = kgust × Gust_est, where kgust is the perturbation coupling coefficient. These two corrections are superimposed on the response characteristic function parameters corrected in step S140 to generate a response characteristic function jointly compensated for oxidation and airflow perturbation.

[0148] Step S918: Replace the original response characteristic function of the MEMS thermopile sensor with the response characteristic function of the combined compensation of oxidation and airflow disturbance to complete the self-calibration operation of the combined compensation of multidimensional degradation factors.

[0149] The parameters of the response characteristic function, which is jointly compensated for oxidation and airflow disturbance, are sent to the edge microcontroller via a self-calibration command and written into the sensor's internal storage unit.

[0150] Step S1010: Construct a multi-physics coupling model of the thermopile sensor of the microelectromechanical system in the digital twin instance. The multi-physics coupling model couples the stress field distribution of the encapsulation cavity and the stress sensitivity of the thermocouple to the thermal resistance of the contact interface based on the thermoelectric coupling simulation model.

[0151] The cloud-based digital twin engine adds an interface thermal resistance node to the contact interface of each thermocouple pair in the thermal resistance network topology. The relationship between the thermal resistance value of the interface thermal resistance node and the local contact stress is Rcontact = R0 × (1 + c × σstress)^n, where R0 is the initial contact thermal resistance at zero stress, σstress is the interface normal stress, and c and n are material stress-sensitive parameters.

[0152] Step S1011: Obtain the stress sensing signal inside the encapsulation cavity of the thermopile sensor of the microelectromechanical system, and convert the stress sensing signal inside the cavity into the stress field boundary load condition in the multiphysics coupling model.

[0153] Stress sensing signals from various measuring points inside the cavity are read from a piezoresistive stress sensor integrated within the cavity. A stress interpolation algorithm is used to map the stress values ​​at discrete measuring points to local stress values ​​at the thermocouple contact interface, which serve as the boundary load conditions for the stress field.

[0154] Step S1012: In the multiphysics coupling model, calculate the change in interfacial contact thermal resistance of the thermocouple pair under stress based on the stress field boundary load conditions.

[0155] Substitute the local stress value σstress at the thermocouple contact interface into the interface thermal resistance stress sensitivity model Rcontact=R0×(1+c×σstress)^n to calculate the change in interface contact thermal resistance ΔRstress=Rcontact-R0.

[0156] Step S1013: Feed back the change in interface contact thermal resistance to the thermal resistance node corresponding to the thermocouple pair contact interface in the thermal resistance network topology of the multiphysics coupling model, and update the thermal resistance value of the thermal resistance node.

[0157] Update the thermal resistance value of the interface thermal resistance node corresponding to the thermocouple pair contact interface in the thermal resistance network topology to the original thermal resistance value plus ΔRstress.

[0158] Step S1014: Drive the multi-physics coupling model after updating the thermal resistance value to perform multi-physics coupling simulation and deduction of the thermoelectric conversion physical process of the thermopile sensor of the microelectromechanical system, and generate a stress-sensing simulated thermoelectric potential time-series signal.

[0159] Using the updated interface thermal resistance node thermal resistance value and the current measured ambient temperature and measured thermoelectric potential as boundary conditions, the simulation derivation process of steps S124 to S126 is executed to generate a stress-sensing simulated thermoelectric potential time series signal.

[0160] Step S1015: Replace the simulated thermoelectric potential time series signal with the stress-sensing simulated thermoelectric potential time series signal, and perform signal difference analysis processing with the measured thermoelectric potential time series signal to obtain the stress-compensated thermoelectric potential drift characteristic curve.

[0161] The original simulated thermoelectric potential time series signal is replaced with the simulated thermoelectric potential time series signal with stress-sensing simulation. The signal difference analysis processing in step S130 is performed to obtain the stress-compensated thermoelectric potential drift characteristic curve.

[0162] Step S1016: Use the stress-compensated thermoelectric potential drift characteristic curve to perform parameter reverse correction processing on the current response characteristic function of the microelectromechanical system thermopile sensor, and generate a stress-decoupled response characteristic function.

[0163] Using the stress-compensated thermoelectric potential drift characteristic curve, the parameter reverse correction process in step S140 is executed to generate a stress-decoupled response characteristic function. The parameter correction process for this response characteristic function has eliminated the influence of packaging stress variations on the thermoelectric potential drift.

[0164] Step S1017: Replace the original response characteristic function of the MEMS thermopile sensor with the stress-decoupled response characteristic function to complete the self-calibration operation to eliminate packaging stress interference.

[0165] The parameters of the stress-decoupled response characteristic function are sent to the edge microcontroller via a self-calibration command, written into the sensor's internal storage unit, and replace the original response characteristic function parameter table.

[0166] In some embodiments, the digital twin-based MEMS thermopile sensor self-calibration system for performing the above method can be any electronic device with data computing, processing, and storage functions. This digital twin-based MEMS thermopile sensor self-calibration system can be used to implement the digital twin-based MEMS thermopile sensor self-calibration method provided in the above embodiments.

[0167] Typically, a self-calibration system for MEMS thermopile sensors based on digital twins includes a processor and a memory. The processor may include one or more processing cores, such as a 4-core processor or an 8-core processor. The processor can be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field Programmable Gate Array), or PLA (Programmable Logic Array). The processor may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor may also include an AI (Artificial Intelligence) processor, which handles computational operations related to machine learning.

[0168] The memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory is used to store a computer program configured to be executed by one or more processors to implement the above-described self-calibration method for MEMS thermopile sensors based on digital twins.

[0169] In an illustrative embodiment, a computer-readable storage medium is also provided, wherein a computer program is stored in the storage medium, and the computer program, when executed by a processor of a computer device, implements the aforementioned self-calibration method for MEMS thermopile sensors based on digital twins. Optionally, the aforementioned computer-readable storage medium may be ROM (Read-Only Memory), RAM (Random Access Memory), CD-ROM (CompactDisc Read-Only Memory), magnetic tape, floppy disk, and optical data storage device, etc.

[0170] This application provides a computer program product, which includes computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the processor will execute the self-calibration method for MEMS thermopile sensors based on digital twins provided in this application.

[0171] This application provides a computer-readable storage medium storing computer-executable instructions or computer programs. When the computer-executable instructions or computer programs are executed by a processor, the processor will execute the self-calibration method for MEMS thermopile sensors based on digital twins provided in this application.

[0172] In some embodiments, the computer-readable storage medium may be a read-only memory (ROM), random access memory (RAM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic surface memory, optical disk, or CD-ROM, etc.; or it may be a device that includes one or any combination of the above-mentioned memories.

[0173] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0174] As an example, computer-executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0175] As an example, computer-executable instructions can be deployed to execute on a single electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0176] Finally, it should be noted that the above-disclosed embodiments are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A self-calibration method for MEMS thermopile sensors based on digital twins, characterized in that, The method includes: The measured thermoelectric potential timing signal of the microelectromechanical system thermopile sensor during the current working period is acquired, and the measured ambient temperature timing signal of the packaging cavity where the microelectromechanical system thermopile sensor is located is acquired simultaneously. The measured thermoelectric potential timing signal and the measured ambient temperature timing signal are combined into a sensor real-time working status data set. A digital twin instance is invoked that is isomorphic to the physical entity of the thermopile sensor of the microelectromechanical system. The real-time operating status data set of the sensor is loaded into the thermoelectric coupling simulation model of the digital twin instance. The thermoelectric coupling simulation model is used to simulate and deduce the thermoelectric conversion physical process of the thermopile sensor of the microelectromechanical system, and generate the simulated thermoelectric potential time-series signal output by the digital twin instance. The measured thermoelectric potential timing signal and the simulated thermoelectric potential timing signal are subjected to signal difference analysis processing to obtain the thermoelectric potential drift characteristic curve reflecting the output deviation between the physical entity of the thermopile sensor of the microelectromechanical system and the digital twin instance. The current response characteristic function of the microelectromechanical system thermopile sensor is corrected by using the thermoelectric potential drift characteristic curve to generate an updated response characteristic function of the microelectromechanical system thermopile sensor. The original response characteristic function of the microelectromechanical system thermopile sensor is then replaced with the updated response characteristic function to complete the self-calibration process.

2. The self-calibration method for MEMS thermopile sensors based on digital twins according to claim 1, characterized in that, The process involves calling a digital twin instance that is isomorphic to the physical entity of the microelectromechanical system (MEMS) thermopile sensor, loading the real-time operating status data set of the sensor into the thermoelectric coupling simulation model of the digital twin instance, simulating and deducing the thermoelectric conversion physical process of the MEMS thermopile sensor through the thermoelectric coupling simulation model, and generating a simulated thermoelectric potential time-series signal output by the digital twin instance, including: The thermal resistance network topology between the cold and hot ends of the sensor in the thermoelectric coupling simulation model of the digital twin instance is analyzed. The thermal resistance network topology is composed of multiple thermal resistance nodes and multiple heat conduction paths. Each thermal resistance node corresponds to the thermal resistance attribute of the thermocouple pair in the space of the thermopile sensor of the microelectromechanical system. Extract the cavity internal temperature value from the measured ambient temperature time-series signal at the current acquisition time from the sensor real-time operating status data set, determine the cavity internal temperature value as the transient boundary temperature condition of the sensor cold end node in the thermal resistance network topology, and use it as the reference parameter for inverting and calculating the hot end temperature. The potential difference value is extracted from the measured thermoelectric potential time-series signal at the current acquisition time from the real-time working status data set of the sensor. Combined with the transient boundary temperature condition of the cold end node of the sensor and the preset Seebeck coefficient, the potential difference value is used to calculate the equivalent temperature estimate of the hot end node of the sensor. The equivalent temperature estimate is then determined as the transient boundary temperature condition of the hot end node of the sensor in the thermal resistance network topology. Based on the transient boundary temperature conditions of the cold-end node and the hot-end node of the sensor, the thermal resistance network topology is driven to iteratively solve the heat flux density distribution along each heat conduction path to obtain the node temperature distribution state of each thermal resistance node in the thermal resistance network topology at the current acquisition time. Based on the node temperature distribution and the Seebeck coefficient distribution parameters of the thermocouple pair, Seebeck voltage accumulation calculation is performed along the thermal resistance nodes covered by the thermocouple pair to generate the simulated thermoelectric potential value of the digital twin instance at the current acquisition time. The simulated thermoelectric potential values ​​at each acquisition time are arranged in chronological order of acquisition time to form a simulated thermoelectric potential time sequence signal.

3. The self-calibration method for MEMS thermopile sensors based on digital twins according to claim 1, characterized in that, The step of performing signal difference analysis on the measured thermoelectric potential time-series signal and the simulated thermoelectric potential time-series signal to obtain a thermoelectric potential drift characteristic curve reflecting the output deviation between the physical entity of the microelectromechanical system thermopile sensor and the digital twin instance includes: The potential difference values ​​at each acquisition time in the measured thermoelectric potential time series signal are compared with the simulated thermoelectric potential values ​​at the same acquisition time in the simulated thermoelectric potential time series signal to generate a deviation signal sequence. Each deviation signal point in the deviation signal sequence corresponds to an acquisition time. The deviation signal sequence is segmented in the time dimension, and adjacent deviation signal points with continuous deviation signal change trends are aggregated into deviation signal segment units, and the trend inflection boundary points between adjacent deviation signal segment units are determined. For each deviation signal segment unit, feature extraction is performed to extract the starting amplitude value, ending amplitude value, and time span of the acquisition time span of the deviation signal within the deviation signal segment unit. Based on the initial amplitude value, the final amplitude value, and the time span of each deviation signal segment unit, calculate the amplitude drift rate descriptor and the amplitude drift direction descriptor of that deviation signal segment unit; The amplitude drift rate descriptors and amplitude drift direction descriptors of all deviation signal segment units are combined in chronological order to generate the thermoelectric potential drift characteristic curve. The thermoelectric potential drift characteristic curve describes the deviation evolution trajectory of the output of the microelectromechanical system thermopile sensor physical entity relative to the output of the digital twin instance in a segmented manner.

4. The self-calibration method for MEMS thermopile sensors based on digital twins according to claim 1, characterized in that, The process of using the thermoelectric potential drift characteristic curve to perform parameter reverse correction on the current response characteristic function of the microelectromechanical system thermopile sensor, generating an updated response characteristic function of the microelectromechanical system thermopile sensor, and replacing the original response characteristic function of the microelectromechanical system thermopile sensor with the updated response characteristic function to complete the self-calibration process includes: The function structure type and the set of parameters to be corrected of the current response characteristic function of the thermopile sensor of the microelectromechanical system are analyzed. The set of parameters to be corrected includes Seebeck coefficient scaling factor and cold junction temperature compensation bias. The amplitude drift rate descriptor and amplitude drift direction descriptor of each deviation signal segment unit are extracted from the thermoelectric potential drift characteristic curve. The amplitude drift rate descriptor is mapped to the attenuation correction amount of the Seebeck coefficient scaling factor, and the amplitude drift direction descriptor is mapped to the offset correction amount of the cold junction temperature compensation bias. Based on the attenuation correction amount and the offset correction amount corresponding to each deviation signal segment unit, the Seebeck coefficient scaling factor and cold junction temperature compensation bias amount in the set of parameters to be corrected are corrected segment by segment to generate the Seebeck coefficient scaling factor sequence and cold junction temperature compensation bias amount sequence after segment correction. Using the piecewise corrected Seebeck coefficient scaling factor sequence and the cold junction temperature compensation bias sequence, the current response characteristic function is piecewise parameter replaced to generate an updated response characteristic function; The parameter table of the original response characteristic function stored in the internal storage unit of the microelectromechanical system thermopile sensor is erased, and the parameter table of the updated response characteristic function is written into the internal storage unit, thus completing the replacement operation of the original response characteristic function with the updated response characteristic function.

5. The self-calibration method for MEMS thermopile sensors based on digital twins according to claim 1, characterized in that, The method further includes: Acquire aging characteristic data of the cavity packaging material of the packaging cavity where the microelectromechanical system thermopile sensor is located. The aging characteristic data of the cavity packaging material includes a sequence of changes in the transmittance of the packaging material and a sequence of changes in the thermal conductivity of the packaging material. The sequence of changes in the transmittance of the encapsulation material is converted into the flux attenuation factor sequence of the radiative heat flux input channel in the digital twin instance, and the sequence of changes in the thermal conductivity of the encapsulation material is converted into the time-varying parameter sequence of the thermal resistance of the thermal resistance node in the thermal resistance network topology of the thermoelectric coupling simulation model. The heat transfer parameters of the corresponding path in the thermoelectric coupling simulation model are updated using the flux attenuation factor sequence and the thermal resistance time-varying parameter sequence to generate an aging-sensing thermoelectric coupling simulation model. The aging-sensing thermoelectric coupling simulation model is used to replace the previous thermoelectric coupling simulation model, and the simulation deduction is performed to generate a simulated thermoelectric potential time series signal after aging compensation.

6. The self-calibration method for MEMS thermopile sensors based on digital twins according to claim 1, characterized in that, The method further includes: The multiple thermoelectric potential drift characteristic curves generated by the digital twin instance in multiple consecutive self-calibration cycles are arranged in time series to form the sensor's full life cycle drift characteristic sequence. The amplitude drift rate descriptor and amplitude drift direction descriptor of each thermoelectric potential drift characteristic curve in the full life cycle drift characteristic sequence of the sensor are extracted, and the change trajectory of amplitude drift rate descriptor and amplitude drift direction descriptor over time are established. Based on the time-varying trajectory of the amplitude drift rate descriptor, predict the drift acceleration trend of the microelectromechanical system thermopile sensor during future operating periods; Based on the time-varying trajectory of the amplitude drift direction descriptor, the probability of drift direction reversal of the microelectromechanical system thermopile sensor during future operating periods is predicted. Based on the drift acceleration trend and the drift direction reversal probability, a pre-failure risk indication signal is generated for the microelectromechanical system thermopile sensor.

7. The self-calibration method for MEMS thermopile sensors based on digital twins according to claim 1, characterized in that, The method further includes: When the microelectromechanical system thermopile sensor is not in operation, a standard temperature difference excitation signal with a known amplitude is applied to the microelectromechanical system thermopile sensor. The standard temperature difference excitation signal is generated by a calibration heat source placed on the hot end side of the microelectromechanical system thermopile sensor. The standard excitation thermoelectric potential signal of the microelectromechanical system thermopile sensor under the action of the standard temperature difference excitation signal is acquired. Obtain the simulated standard excitation thermoelectric potential signal of the digital twin instance under the same boundary conditions as the standard temperature difference excitation signal; The standard excitation thermoelectric potential signal and the simulated standard excitation thermoelectric potential signal are analyzed to obtain the standard excitation drift characteristic curve. The current response characteristic function of the microelectromechanical system thermopile sensor is corrected using the standard excitation drift characteristic curve, and the corrected response characteristic function is written into the internal storage unit of the microelectromechanical system thermopile sensor as the initial response characteristic function of the sensor.

8. The self-calibration method for MEMS thermopile sensors based on digital twins according to claim 1, characterized in that, The method further includes: Obtain a cluster of digital twin instances of multiple sensors from the same batch as the microelectromechanical system thermopile sensor in the digital twin instance, wherein the multiple sensors share the same thermoelectric coupling simulation model structure with the microelectromechanical system thermopile sensor; From the cluster of digital twin instances of the multiple sensors, extract the batch thermoelectric potential drift characteristic curves generated by each digital twin instance of the sensors during its respective working period; The group deviation of the batch thermoelectric potential drift characteristic curves is compared with that of the thermoelectric potential drift characteristic curves of the microelectromechanical system thermopile sensor to identify the degree of group drift deviation of the microelectromechanical system thermopile sensor relative to the multiple sensors. Based on the degree of population drift deviation, the amplitude drift rate descriptor in the thermoelectric potential drift characteristic curve is corrected to generate a population-corrected thermoelectric potential drift characteristic curve. The current response characteristic function of the thermoelectric pile sensor of the microelectromechanical system is reverse-parameterized using the thermoelectric potential drift characteristic curve after the group correction.

9. The self-calibration method for MEMS thermopile sensors based on digital twins according to claim 1, characterized in that, The method further includes: The thermoelectric potential drift characteristic curve is transformed by frequency domain transformation to convert the time domain representation of the thermoelectric potential drift characteristic curve into the frequency domain representation, thereby obtaining the drift frequency domain characteristic spectrum; Extract the dominant frequency component and the corresponding frequency amplitude of the drift frequency domain feature spectrum, wherein the dominant frequency component is the frequency point in the drift frequency domain feature spectrum whose frequency amplitude exceeds the amplitude threshold; The dominant frequency component and the frequency amplitude of the corresponding dominant frequency component are compared and matched with the known periodic environmental disturbance frequency of the working environment of the microelectromechanical system thermopile sensor. Based on the matching results, the periodic drift component caused by the periodic environmental disturbance and the monotonic drift component caused by the aging of the sensor itself are separated from the thermoelectric potential drift characteristic curve. The current response characteristic function of the microelectromechanical system thermopile sensor is corrected by using only the monotonic drift component.

10. A self-calibration system for MEMS thermopile sensors based on digital twins, characterized in that, The device includes a processor and a computer-readable storage medium storing machine-executable instructions, which, when executed by a computer, implement the self-calibration method for MEMS thermopile sensors based on digital twins as described in any one of claims 1-9.

Images