A method for calibrating inner membrane heat transfer coefficient based on Wilson regression

Abstract

The present application belongs to the technical field of thermophysical parameter measurement and process control, and relates to a kind of inner membrane heat transfer coefficient calibration method based on Wilson regression, comprising the following steps: output composite broadband excitation signal and main clock trigger signal;Output the frame data set of the wall acceleration and the frame data set of micro heat flow fluctuation;Solve the output mechanical vibration transfer function;Get the thermodynamic thermal admittance function;Synthesize thermal coupling complex spectrum;The target inner membrane heat transfer coefficient is combined with the current working condition Reynolds number and packed to generate dynamic heat transfer calibration data model.The present application solves the problem that mechanical and thermal cross coupling interference is easy to introduce insufficiently quantized system deviation into measurement results.

Description

Technical Field

[0001] This invention belongs to the technical field of thermophysical parameter measurement and process control, and relates to a method for calibrating the inner membrane heat transfer coefficient based on Wilson regression. Background Technology

[0002] In fields such as chemical engineering, biopharmaceuticals, and food processing, characterizing the heat transfer performance of key equipment such as jacketed reactors is fundamental to achieving process optimization and scale-up. The inner membrane heat transfer coefficient, as a core parameter describing the heat exchange capacity between materials and the reactor's inner wall, is influenced by dynamic factors such as fluid physical properties, stirring intensity, and wall conditions, exhibiting nonlinear and time-varying characteristics, making its measurement quite challenging.

[0003] Chinese patent application number 201410629263.9 discloses an experimental apparatus for measuring the boiling heat transfer coefficient of refrigerant flow inside a pipe. Such prior art generally employs the Wilson graphical method based on steady-state experiments to determine the heat transfer coefficient. This method mainly involves conducting heat transfer experiments under multiple different and stable operating conditions to obtain a series of overall heat transfer coefficient values. Subsequently, regression or graphical extrapolation is performed using a thermodynamic theoretical model to separate the inner film heat transfer resistance and its corresponding heat transfer coefficient.

[0004] The aforementioned existing technologies exhibit certain limitations in practical applications. Current heat transfer coefficient calibration typically employs classic Wilson graphical methods, such as Wilson regression. This method requires the system to wait for a long time after changing the flow rate until multiple new thermodynamic steady states are reached before the inner membrane heat transfer coefficient can be extracted through linear fitting. This steady-state dependence results in an extremely long overall calibration cycle, making online dynamic monitoring impossible. Such calibration techniques usually require the system to reach thermodynamic steady state at each measurement point, leading to an excessively long overall calibration cycle and difficulty in reflecting the impact of dynamic process parameter changes on heat transfer performance in real time. Furthermore, these methods often treat the heat transfer process as a single thermodynamic problem, rarely considering the disturbance effects of the equipment system's own mechanical vibration on the boundary layer fluid for comprehensive analysis. This cross-coupling of mechanical and thermodynamic interference easily introduces unquantified systemic biases into the measurement results.

[0005] Therefore, the technical problem to be solved by the present invention is to provide a method for calibrating the inner membrane heat transfer coefficient that can overcome the interference of mechanical vibration cross-coupling and realize online dynamic measurement. Summary of the Invention

[0006] To address the aforementioned issues, this invention breaks away from the dependence of traditional Wilson regression on absolute steady state and proposes a novel multi-physics frequency domain decoupling method. By superimposing broadband micro-perturbations and utilizing digital phase-locked detection, a method for calibrating the inner membrane heat transfer coefficient based on Wilson regression is provided.

[0007] A method for calibrating the inner membrane heat transfer coefficient based on Wilson regression includes the following steps: S1. Obtain the basic control parameters of the target fluid reaction system and the target frequency band range to generate a low-frequency large-amplitude macroscopic frequency sweep sequence. Combine the preset phase coding feature matrix table to generate and superimpose a high-frequency detection sequence, and output a composite broadband excitation signal and a master clock trigger signal. S2. Feed the composite broadband excitation signal into the device motor to apply the actuation disturbance, and synchronously cut off the continuous data stream of the multi-physics field according to the master clock trigger signal, and output the set of device wall acceleration frame data and the set of micro heat flow fluctuation frame data. S3. Extract the high-frequency detection sequence as the mechanical domain reference local oscillator signal, perform phase-sensitive filtering on the set of acceleration frame data of the device wall, and calculate and output the mechanical vibration transfer function. S4. The high-frequency detection sequence is introduced into the thermodynamic domain and converted into a thermodynamic domain reference local oscillator signal. Parallel thermodynamic domain detection is applied to the microscopic heat flow fluctuation frame data set to obtain the thermodynamic thermal admittance function. S5. Extract the discrete envelope features of the mechanical vibration transfer function to generate a dynamic external resistance factor sequence. Divide the dynamic external resistance factor sequence with the thermodynamic thermal admittance function by performing decoupling and deconvolution numerical array to synthesize a thermo-coupled composite spectrum. S6. Locate the low-frequency approximation segment of the thermo-coupled composite spectrum to extract the initial steady-state value of heat transfer, calculate the area of ​​the continuous discrete data set in the characteristic working band interval and derive the target inner membrane heat transfer coefficient, and combine the target inner membrane heat transfer coefficient with the current operating condition Reynolds number to generate a dynamic heat transfer calibration data model.

[0008] A further aspect of the present invention involves generating and superimposing a high-frequency detection sequence based on a preset phase-encoded feature matrix table, comprising the following steps: Read the instantaneous frequency distribution and amplitude parameters of a low-frequency, large-amplitude macroscopic sweep frequency sequence; Using instantaneous frequency distribution and amplitude parameters as query indexes, access the preset phase coding feature matrix table and extract the matching symbol switching rate and modulation amplitude. A high-frequency detection sequence is generated based on the symbol switching rate and modulation amplitude. At the same time, the high-frequency detection sequence is superimposed and fused onto the low-frequency large-amplitude macroscopic sweep frequency sequence waveform to form a composite broadband excitation signal.

[0009] A further aspect of the present invention involves synchronously truncating the acquired multi-physics continuous data stream based on a master clock trigger signal, comprising the following steps: Activate the piezoelectric acceleration sensing element installed on the outer wall and the thin film heat flow sensing node laid on the inner side, and synchronously sample at a sampling frequency of not less than twice the maximum symbol switching rate of the high-frequency detection sequence to obtain the continuous wall vibration data stream and the continuous micro heat flow fluctuation data stream. The master clock trigger signal is used as the time gating cut reference to force synchronous truncation of the continuous wall vibration data stream and the continuous micro heat flow fluctuation data stream. The truncated stream data is integrated and output as a set of wall acceleration frame data and a set of microscopic heat flux fluctuation frame data with temporal dimension binding consistency.

[0010] A further aspect of this invention involves performing phase-sensitive filtering on the set of acceleration frame data of the vessel wall to calculate and output the mechanical vibration transfer function, including the following steps: Using the mechanical domain reference local oscillator signal and its orthogonal components, the device wall acceleration frame data set is subjected to multi-cycle coherent interactive orthogonal demodulation and rectangular time window sliding window integration to filter irregular environmental mechanical background noise and output a multidimensional complex vector sequence. Solve for the relative energy reduction ratio and physical time lag difference of the multidimensional complex vector sequence relative to the mechanical domain reference local oscillator signal; The relative energy reduction ratio and physical time lag difference of each frequency node are calculated and combined to draw discrete characteristic curves, and finally a mechanical vibration transfer function with a unique indicator of the metal excitation response state of the equipment is generated.

[0011] A further aspect of this invention involves applying parallel thermodynamic domain detection to a dataset of microscopic heat flow fluctuation frames to obtain the thermodynamic thermal admittance function, comprising the following steps: The independent digital phase-locked amplifier unit continuously applies the thermal domain reference local oscillator signal to the micro thermal flow fluctuation frame data set, reproduces and executes the orthogonal demodulation dot multiplication and cumulative mean integration operation mechanism of the equivalent mechanical demodulation processing rule; The redundant physical parameter components induced by the penetration of the medium layer are stripped off, and the remaining physical environment thermal interference is filtered out by cumulative mean integral calculation. The in-phase and orthogonal components after orthogonal demodulation are extracted as the thermal penetration pulsation index parameters. By integrating and mapping real heat permeation pulsation index parameters, we can present quantitative chart results that demonstrate the dynamic exchange of transient heat energy at the fluid medium interface, and generate a thermodynamic thermal admittance function free from perturbation-induced displacement contamination.

[0012] A further aspect of this invention involves extracting the discrete envelope features of the mechanical vibration transfer function to generate a dynamic external resistance factor sequence, comprising the following steps: The mechanical vibration transfer function amplitude-frequency characteristic scanning array is identified and extracted by applying an extreme value search and judgment algorithm. Capture the features of isolated mechanical resonance peak envelopes and valley convergence point clusters within the line segment of the amplitude-frequency characteristic scanning array results. The equivalent disturbance intensity parameter of the mechanical kinetic energy diffusion through the metal boundary layer to the external auxiliary cooling source is measured. Based on a specific associated physical model, numerical array analysis is performed to output a dynamic external resistance factor sequence that quantifies the unsteady thermal resistance fluctuation of the system.

[0013] A further aspect of this invention involves dividing the dynamic external resistance factor sequence by the thermodynamic thermal admittance function using a decoupling and deconvolution numerical array to synthesize a thermodynamically coupled composite spectrum, comprising the following steps: 5

[0014] The main core engine of the deployment device retrieves the comprehensive heat transfer admittance quantization set recorded in the frequency domain coordinate system as the basic divisor sequence term, which is the thermodynamic thermal admittance function recorded in the frequency domain coordinate system. The decoupled deconvolution numerical array is divided by the basic dividend sequence terms and the converted dynamic external resistance factor sequence, according to the requirement of iterative division operation with a predetermined bandwidth step; By using the frequency domain decoupling division algorithm, the false thermal resistance effect of the adhesion caused by the random impact feedback of the fluid on the outer wall of the system is filtered out, and the data array information that is not affected by the redundant response of the wrapping layer is purely retained and the thermo-mechanical coupling composite spectrum of the target system is inversely fitted.

[0015] A further aspect of the present invention involves locating the low-frequency approximation region of the thermo-coupled composite spectrum to extract the initial steady-state value of heat transfer, comprising the following steps: A continuous observational decay condition is constructed within the computer, where the dependent variable continuously approaches the zero baseline along the vertical axis of the computing engine. The continuous observation of the approaching attenuation condition is superimposed and substituted into the dynamic slope evaluation and judgment module of the thermo-coupled composite spectrum to retrieve and determine the low-frequency approximation section where the fluctuation amplitude of the output curve converges smoothly. The low-frequency approximation segment obtained by truncation is extracted and stripped of its inherent physical phase change limit, the basic limit heat transfer coefficient convergence steady-state initial value, and the heat transfer steady-state initial value specific to the steady-state conduction limit scenario is solidified and constructed.

[0016] A further aspect of the present invention involves calculating the area of ​​a continuous discrete data set within a characteristic operating band interval and deriving the target inner membrane heat transfer coefficient, comprising the following steps: Using the original location of the extracted heat transfer steady-state initial value as the geometric reference center point, a bidirectional extrapolation band fixed-frequency expansion is implemented to delineate the outer continuous discrete point cluster as the characteristic working band interval; In the generated characteristic working band interval, the two-dimensional area below the frequency distribution axis of the frequency band correlation data curve is integrated and taken as the area of ​​the continuous discrete data set representing the total true value of heat transfer. The area of ​​the continuous discrete data set mapped by the calculation results is sent to the internally configured smoothing normalization integral average transformation functional area module to perform coefficient back-reasoning correction operation, suppress and eliminate isolated and accidental abnormal test noise, and then derive the high-precision corrected target inner membrane heat transfer coefficient.

[0017] A further aspect of this invention involves merging the target inner membrane heat transfer coefficient with the current operating condition Reynolds number to generate a dynamic heat transfer calibration data model, including the following steps: The system's underlying average environmental rotation speed is recorded simultaneously during the synchronous operation phase when the target inner membrane heat transfer coefficient is generated, and the current operating condition Reynolds number associated with the instantaneous flow field characteristics is calculated. The solved target inner membrane heat transfer coefficient is combined with the current operating condition Reynolds number to establish a strong link dependency relationship. Combination encapsulation is then performed to establish the interlocked key-value pair data cluster required for a one-to-one linear query and retrieval mechanism. Multiple sets of interlocked key-value pairs data clusters, established iteratively at time points according to operating condition sequences, are written to the storage pool, and the storage units are refreshed according to structured management rules to generate a dynamic heat transfer calibration data model with open interfaces for external exposure.

[0018] In summary, the present invention has the following beneficial technical effects: 1. A composite broadband excitation signal was constructed by superimposing a high-frequency detection sequence onto a low-frequency macroscopic sweep sequence, and multi-physics data were acquired synchronously during system operation. In this technique, the low-frequency component drives the reaction system to smoothly traverse different operating condition ranges, while the high-frequency phase-encoded sequence serves as a traceable disturbance probe continuously injected into the system. This mechanism allows for calibration without interrupting production or waiting for the establishment of multiple thermodynamic steady-state points. Dynamic response data covering a wide range of operating conditions can be acquired in a single continuous operation, transforming offline, static heat transfer coefficient calibration into continuous, dynamic online monitoring, thus improving the ability to monitor dynamic changes in process parameters and the degradation of heat transfer performance.

[0019] 2. The frequency domain mathematically decouples the pure thermodynamic response from mechanical crosstalk, enabling transient extraction of the heat transfer coefficient without waiting for macroscopic steady-state conditions. A dual-channel digital lock-in amplification technique is employed, using a high-frequency probe sequence as the reference local oscillator signal. Orthogonal demodulation and sliding window integration are performed on the acquired frame data of vessel wall acceleration and microscopic heat flow fluctuations, respectively. By calculating the correlation between the mixed signal and a specific coded sequence, this feature effectively filters out background noise and other random interference unrelated to the excitation source. This method effectively separates the weak response signal induced by specific disturbances from noisy industrial environments, improving the signal-to-noise ratio of vessel wall vibration and interface heat flow fluctuation measurements, and providing a reliable data foundation for subsequent calculations of the transfer function and thermal admittance function.

[0020] 3. By parallel solving of the mechanical vibration transfer function and thermodynamic thermal admittance function, mechanical resonance features are extracted and converted into a dynamic external resistance factor sequence. Then, deconvolution and division operations are performed on the thermal admittance function and the dynamic external resistance factor sequence in the frequency domain. This technique quantifies the disturbance intensity of the external heat transfer boundary layer caused by the mechanical vibration of the vessel wall due to stirring, and removes the contribution of this mechanical interference from the measured total thermal admittance. This mechanism effectively eliminates the systematic errors caused by the mechanical-thermal cross-coupling effect often ignored in traditional measurements, purifying the thermodynamic coupling composite spectrum that only reflects the true convective heat transfer dynamics between the main fluid inside the vessel and the inner wall, thereby improving the objectivity and accuracy of the finally derived inner membrane heat transfer coefficient. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention.

[0022] Figure 1 A flowchart illustrating an embodiment of this application is disclosed.

[0023] Figure 2 The schematic diagram of the framework in the embodiments of this application is disclosed.

[0024] Figure 3 The frequency domain characteristics and area integral diagram of the mechanical vibration transfer function, thermodynamic thermal admittance function and the purely thermodynamic coupled composite spectrum in the embodiments of this application are disclosed.

[0025] Figure 4 A schematic diagram of the dynamic heat transfer calibration data model curve of the target inner membrane heat transfer coefficient as a function of Reynolds number in an embodiment of this application is disclosed. Detailed Implementation

[0026] The following is in conjunction with the appendix Figure 1 - Figure 4 A preferred description of the present invention is provided below.

[0027] See attached document Figure 1 This invention proposes a method for calibrating the inner membrane heat transfer coefficient based on Wilson regression, comprising the following steps: S1. Obtain the basic control parameters of the target fluid reaction system and the target frequency band range to generate a low-frequency large-amplitude macroscopic frequency sweep sequence. Combine the preset phase coding feature matrix table to generate and superimpose a high-frequency detection sequence, and output a composite broadband excitation signal and a master clock trigger signal. S2. Feed the composite broadband excitation signal into the device motor to apply the actuation disturbance, and synchronously cut off the continuous data stream of the multi-physics field according to the master clock trigger signal, and output the set of device wall acceleration frame data and the set of micro heat flow fluctuation frame data. S3. Extract the high-frequency detection sequence as the mechanical domain reference local oscillator signal, perform phase-sensitive filtering on the set of acceleration frame data of the device wall, and calculate and output the mechanical vibration transfer function. S4. The high-frequency detection sequence is introduced into the thermodynamic domain and converted into a thermodynamic domain reference local oscillator signal. Parallel thermodynamic domain detection is applied to the microscopic heat flow fluctuation frame data set to obtain the thermodynamic thermal admittance function. S5. Extract the discrete envelope features of the mechanical vibration transfer function to generate a dynamic external resistance factor sequence. Divide the dynamic external resistance factor sequence with the thermodynamic thermal admittance function by performing decoupling and deconvolution numerical array to synthesize a thermo-coupled composite spectrum. S6. Locate the low-frequency approximation segment of the thermo-coupled composite spectrum to extract the initial steady-state value of heat transfer, calculate the area of ​​the continuous discrete data set in the characteristic working band interval and derive the target inner membrane heat transfer coefficient, and combine the target inner membrane heat transfer coefficient with the current operating condition Reynolds number to generate a dynamic heat transfer calibration data model.

[0028] In one embodiment of the present invention, step S1 includes the following steps: Read the instantaneous frequency distribution and amplitude parameters of a low-frequency, large-amplitude macroscopic sweep frequency sequence; Using instantaneous frequency distribution and amplitude parameters as query indexes, access the preset phase coding feature matrix table and extract the matching symbol switching rate and modulation amplitude. A high-frequency detection sequence is generated based on the symbol switching rate and modulation amplitude. At the same time, the high-frequency detection sequence is superimposed and fused onto the low-frequency large-amplitude macroscopic sweep frequency sequence waveform to form a composite broadband excitation signal.

[0029] Specifically, in this embodiment of the invention, step S1 is executed by a built-in digital signal generator controller. The purpose of this step is to generate a broadband excitation signal with identifiable information encoding and to synchronously generate a hardware time reference for subsequent data acquisition.

[0030] The digital signal generator controller loads a pre-set fluid dynamics measurement database from local non-volatile memory. This database is a structured dataset that stores experimental data or theoretical calculation models, such as stirring speed and stirring power, required to achieve a specific Reynolds number in different fluids (e.g., water, ethanol) at different temperatures. The Reynolds number is set using a relational formula. Bridging with physical devices, where For fluid density, This refers to the stirring speed. The characteristic diameter of the impeller. This refers to the fluid dynamic viscosity. By accessing this fluid dynamics measurement database, the target Reynolds number can be accurately calculated into low-level servo control commands. This database stores the basic control parameters for specific target fluid reaction systems, such as jacketed stirred reactors, under different operating conditions.

[0031] The controller operates within the target frequency band range set according to the current calibration task, for example... to The target operating frequency band defines the frequency variation range of the low-frequency macroscopic sweep sequence, typically ranging from 0.01 Hz to 1 Hz. This frequency range is designed to ensure that the system has sufficient time to respond to macroscopic thermodynamic changes. Simultaneously, the controller incorporates preset measurement calibration intervals, such as the Reynolds number. to The corresponding stirring speed range, where the amplitude range covers the stirring intensity required for calibration, for example from the corresponding lowest Reynolds number. Smooth transition from 50 rpm to the corresponding maximum Reynolds number The controller calls the transfer function model or interpolation algorithm in the database to generate a low-frequency, large-amplitude macroscopic sweep frequency sequence with continuously varying time-domain waveform. This low-frequency, large-amplitude macroscopic sweep frequency sequence is typically designed as a linear frequency modulated (LFM) signal or a logarithmic frequency modulated (LFM) signal. As a LFM signal, its mathematical expression can be defined as:

[0032]

[0033] In the formula, For time variables, For its time The instantaneous rotational speed serves as the baseband carrier for the smooth transition of the macroscopic thermodynamic state of the driving system; The instantaneous frequency that varies with time is used for synchronous table lookup; and These are the start and end values ​​of the speed range, for example, the stirring speeds corresponding to the lowest and highest Reynolds numbers, respectively; The starting frequency of the target operating frequency band; is the frequency sweep rate, where The termination frequency of the target operating frequency band. The total sweep duration is expressed in seconds (s). Subsequently, the digital signal generator controller synchronously samples the instantaneous frequency at each time step of the real-time generation of the low-frequency, large-amplitude macroscopic sweep sequence. With instantaneous speed This refers to the instantaneous amplitude parameter. The controller uses this two-dimensional parameter pair as a query index to access a pre-defined phase-encoded feature matrix table that is internally stored.

[0034] The phase-encoded feature matrix table is established in advance through system simulation or experimental calibration. A numerical table of dimensions, in which, The number of discretized frequency intervals. N The number of discretized amplitude intervals is used to inject different coding features into high-frequency disturbances under different macroscopic conditions, thereby enhancing the robustness of subsequent signal recognition. This matrix is ​​a two-dimensional lookup table, where rows and columns correspond to the discretized frequency and amplitude intervals, respectively, and the storage units contain a pre-set symbol switching rate. With modulation amplitude Parameter pair.

[0035] In practical applications, the symbol switching rate of the high-frequency probe sequence Typically, the frequency is much higher than the sweep frequency base frequency, for example, set between 10 Hz and 200 Hz. The technical basis for setting this symbol switching rate is its lower limit. For example, 10 Hz needs to be at least one order of magnitude higher than the macroscopic sweep frequency base frequency to ensure strict scale separation during frequency domain demodulation and avoid aliasing of high and low frequency features. Its upper limit, such as 200 Hz, is limited by the dynamic response bandwidth of the variable frequency drive and the stirring motor, as well as the low-pass filter damping characteristics of the fluid. Too high a frequency will not be able to form an effective propagating mechanical wave in the fluid.

[0036] Its modulation amplitude This amplitude is much smaller than the baseband signal amplitude, typically ranging from 0.5% to 5% of the baseband amplitude. The basis for setting this parameter range is that if the amplitude percentage is less than 0.5%, the injected high-frequency energy will be submerged by the system's background mechanical noise, causing the heat flow sensor to fail to capture the response with an effective signal-to-noise ratio. If the amplitude percentage is greater than 5%, the excessive high-frequency disturbance energy will destroy the macroscopic quasi-steady-state boundary conditions created by the baseband signal, causing nonlinear cavitation or distortion in the local flow field, thereby causing the extracted linear thermal admittance parameter to fail. This ensures that the injected detection signal energy is sufficient to be captured by the sensor, while not having a dominant impact on the macroscopic flow field.

[0037] The controller, based on the query results and The value drives a pseudo-random sequence generator to generate a high-frequency probe sequence. This sequence is essentially a binary phase-shift keying signal, which serves as a dynamic, macroscopically related identifier superimposed on the system. Its generation process can be represented as follows:

[0038] In the formula, These are the instantaneous values ​​of the high-frequency detection sequence; and From the phase-coded feature matrix table The modulation amplitude and symbol switching rate are obtained by looking up the instantaneous characteristics in a table. It is based on a specific generator polynomial, such as that generated by a linear feedback shift register. A pseudo-random binary sequence with rate switching, the value of which is +1 or -1.

[0039] The signal synthesis unit inside the digital signal generator controller performs digital addition on the high-frequency detection sequence and the low-frequency large-amplitude macroscopic sweep frequency sequence to achieve layer-by-layer superposition and fusion. At this point, the composite broadband excitation signal... At the point of time The instantaneous values ​​are formed by linear superposition, and the synthesis formula is as follows:

[0040] The result of this addition operation is output as a digital signal stream, which is the composite broadband excitation signal used to drive the bottom stirring motor. Simultaneously, the digital signal generator controller performs level conversion and buffering on the internal symbol clock signal used to generate the high-frequency detection sequence, extracting and outputting its rising or falling edge to form a master clock trigger signal with guaranteed timing consistency. This master clock trigger signal is a standard transistor-to-transistor logic TTL level signal, with its pulse leading edge aligned with the start time of each symbol in the high-frequency detection sequence, achieving a timing accuracy at the microsecond level. This signal is provided as a standard digital logic signal to subsequent data acquisition units.

[0041] For example, suppose the current calibration task is a 5 L water-jacketed reactor, the target operating frequency range is set to 0.05 Hz to 0.5 Hz, the sweep time is 100 s, and the corresponding stirring speed range is 50 RPM to 500 RPM. At time s, the digital signal generator controller calculates the instantaneous rotational speed of the low-frequency, large-amplitude macroscopic sweep frequency sequence. RPM, instantaneous frequency The controller uses the instantaneous rotational speed of 140 RPM and the instantaneous frequency of 0.14 Hz as indexes to query the phase coding feature matrix table and retrieve the corresponding symbol switching rate. 50 Hz, modulation amplitude It is 5 RPM.

[0042] The controller generates a pseudo-random binary sequence that switches at a rate of 50 Hz and sets its amplitude to 5 RPM to obtain a high-frequency detection sequence. Therefore, in The instantaneous value of the composite broadband excitation signal output at time s is RPM ,in The value randomly jumps between +5 RPM and -5 RPM at a frequency of 50 Hz. Synchronously, the controller outputs a 50 Hz square wave signal as the master clock trigger signal, with each rising edge corresponding to... The update time of the data bits.

[0043] In one embodiment of the present invention, step S2 includes the following steps: Activate the piezoelectric acceleration sensing element installed on the outer wall and the thin film heat flow sensing node laid on the inner side, and synchronously sample at a sampling frequency of not less than twice the maximum symbol switching rate of the high-frequency detection sequence to obtain the continuous wall vibration data stream and the continuous micro heat flow fluctuation data stream. The master clock trigger signal is used as the time gating cut reference to force synchronous truncation of the continuous wall vibration data stream and the continuous micro heat flow fluctuation data stream. The truncated stream data is integrated and output as a set of wall acceleration frame data and a set of microscopic heat flux fluctuation frame data with temporal dimension binding consistency.

[0044] Specifically, following step S1, the method of this embodiment of the invention continues to execute step S2. This step is completed by the electromechanical control and data acquisition subsystem deployed at the reaction system site. Its core task is to apply the previously generated excitation signal to the physical system and simultaneously capture the system's multi-physics dynamic response. The electromechanical control and data acquisition subsystem receives a composite broadband excitation signal from a digital signal generator controller. This digital signal stream is converted into an analog voltage signal by a high-speed digital-to-analog converter (DAC) and fed into the electromechanical actuator of the target fluid reaction system.

[0045] The electromechanical actuator is an integrated physical system, including a variable frequency drive (VFD) or servo controller that receives control signals, a stirring drive motor that performs speed changes, a stirring shaft that transmits torque, and a stirring paddle that interacts with the fluid. The analog voltage signal serves as a speed command, which controls the actual speed of the stirring drive motor to reproduce the time-domain waveform of the composite broadband excitation signal in real time. This causes the stirrer to generate a composite motion mode in the operating medium.

[0046] Among them, the low-frequency, large-amplitude macroscopic sweep frequency sequence component in the signal causes the overall flow state of the operating medium, i.e. the macroscopic thermodynamic equilibrium state, to undergo a long-period nonlinear flow process from low Reynolds number to high Reynolds number; while the high-frequency detection sequence component superimposed on it induces high-frequency broadband physical disturbances at the contact interface between the fluid and the vessel wall. These disturbances specifically refer to the microscale eddies and pressure pulsations generated in the fluid boundary layer near the vessel wall caused by the high-frequency disturbance sequence.

[0047] Throughout the entire cycle of continuous change in motor torque based on the composite broadband excitation signal, the data acquisition unit synchronously activates pre-deployed sensing elements. Specifically, multiple piezoelectric accelerometers installed at specific locations on the outer wall of the stirred tank are activated. These sensing elements are industrial-grade piezoelectric ICP accelerometers with built-in integrated circuits. Their measurement frequency band needs to cover the frequency range of the high-frequency detection sequence, for example, from 1 Hz to 20 kHz. The lower limit of this measurement frequency band, 1 Hz, ensures backward compatibility with the slowest macroscopic operating condition scanning fundamental wave in this embodiment, while the upper limit, 20 kHz, is higher than the frequency of fluid boundary layer eddy current rupture at the microscale. This high-frequency margin ensures that the hardware system can completely capture all high-order transient physical pulsations caused by micro-disturbances, avoid spectral leakage caused by high-frequency signal aliasing, and capture the structural vibrations of the tank caused by internal fluid dynamics and mechanical stirring forces. At the same time, multiple thin-film heat flow sensing nodes integrated on the inner surface of the reactor jacket, i.e. the side in contact with the cooling or heating medium, are also activated. These nodes are fast-responding thermal sensors, such as Gordon meters or thin-film thermopile based on the Seebeck effect, with thermal response time constants on the order of milliseconds, which can capture rapid heat flow fluctuations caused by microscopic turbulence.

[0048] The analog output signals from these two sensing channels are fed into a multi-channel synchronous analog-to-digital converter (ADC). This converter, following the general principle in the art of satisfying the Nyquist theorem and preserving waveform information, samples the signals at the highest frequency threshold supported by the device. This highest frequency threshold sampling refers to the sampling frequency of the ADC. Its set value must be greater than the maximum symbol switching rate of the high-frequency detection sequence. Twice that, in practice it is usually set to This generates a continuous wall vibration data stream and a continuous microscopic heat flow fluctuation data stream, respectively.

[0049] The data acquisition unit references the master clock trigger signal synchronously generated in the preceding step S1 and configures it as the hardware-level time-gated cutting reference. When the effective level edge of the master clock trigger signal, such as the rising edge, arrives, the firmware or low-level driver of the data acquisition unit marks this moment as the start boundary of the data frame and marks the arrival time of the next effective edge as the end boundary of the frame.

[0050] In this way, the continuous vessel wall vibration data stream and the continuous microscopic heat flux fluctuation data stream are truncated into a series of data segments. These data segments are then organized into frame data structures, outputting a set of vessel wall acceleration frame data and a set of microscopic heat flux fluctuation frame data with phase dimension binding. Both sets are arrays of data structures, where each element is a numerical vector containing all reading samples acquired within one master clock trigger signal cycle. The indices of these two sets have a strict synchronous correspondence, i.e., the index value is... The acceleration frame and index value The heat flux fluctuation frames were acquired within the exact same time window for subsequent demodulation analysis.

[0051] For example, continuing the example of S1 above, the composite broadband excitation signal in The command at time s is a high-frequency disturbance with an amplitude of 5 RPM and a symbol switching rate of 50 Hz superimposed on a base rotation speed of 140 RPM. This signal is applied to the stirring motor via a digital-to-analog converter and a frequency converter driver. Simultaneously, the output signals from the accelerometer and heat flux sensor in the data acquisition subsystem are sampled at a frequency... The digitization was performed using a multi-channel ADC set to 10 kHz.

[0052] Since the frequency of the main clock trigger signal output by S1 is 50 Hz, its period is s stands for 20 ms. When the first rising edge of the master clock trigger signal arrives, the data acquisition system begins storing subsequent ADC sampling points into the buffer of the first frame. After 20 ms, the second rising edge arrives, the system completes the acquisition of the first frame and begins acquiring the second frame. Within this 20 ms time window, the system acquires a total of 200 sampling points were used. Therefore, a wall acceleration frame containing 200 acceleration readings and a microscopic heat flux fluctuation frame containing 200 heat flux readings were generated. These two frames were stored in the wall acceleration frame dataset and the microscopic heat flux fluctuation frame dataset, respectively, as elements with the same index number. This process will continue as the frequency sweep proceeds, generating a total of 200 frames over the entire 100-second calibration period. For such synchronized data frames.

[0053] In one embodiment of the present invention, step S3 includes the following steps: Using the mechanical domain reference local oscillator signal and its orthogonal components, the device wall acceleration frame data set is subjected to multi-cycle coherent interactive orthogonal demodulation and rectangular time window sliding window integration to filter irregular environmental mechanical background noise and output a multidimensional complex vector sequence. Solve for the relative energy reduction ratio and physical time lag difference of the multidimensional complex vector sequence relative to the mechanical domain reference local oscillator signal; The relative energy reduction ratio and physical time lag difference of each frequency node are calculated and combined to draw discrete characteristic curves, and finally a mechanical vibration transfer function with a unique indicator of the metal excitation response state of the equipment is generated.

[0054] Specifically, after step S2 is completed, the process proceeds to step S3, which is executed by the identification module in the central data processing unit. Its core function is to extract the system's mechanical response characteristics from a strong noise background using digital lock-in amplification (DLA). The identification module extracts a high-frequency detection sequence from memory or reconstructs it according to the generation rules of step S1, ensuring it is synchronized with the original excitation signal and aligned with its time zero point. This high-frequency detection sequence is locked as the mechanical domain reference local oscillator signal processed within the first-stage DLA unit.

[0055] The first-stage digital lock-in amplifier unit is an algorithm module, functionally equivalent to a hardware lock-in amplifier. It consists of three core parts: digital mixing, low-pass filtering, and amplitude-phase calculation. In this embodiment, the low-pass filtering is implemented using sliding window integration. The mechanical domain reference local oscillator signal is a clean digital copy. This operation ensures that the key used in the demodulation process is consistent with the probe signal injected into the physical system in frequency, phase, and coding structure.

[0056] The identification module combines the device wall acceleration frame dataset generated in step S2 into a signal to be processed, and inputs it frame by frame into the first-stage digital lock-in amplifier unit. For each data frame in the set, for example, the first... In a frame, this unit performs a complete cyclic quadrature demodulation operation, which utilizes the orthogonality principle of trigonometric functions, i.e., the integral result of sine waves of different frequencies within one cycle is zero. Specifically, the digital mixer inside this unit simultaneously multiplies the input acceleration data stream point-by-point with the mechanical domain reference local oscillator signal and its orthogonal components, such as the -90° phase shift component, to generate an in-phase component I data stream and an orthogonal component Q data stream.

[0057] A rectangular time-window sliding integrator integrates or averages the two data streams over the entire data frame time, thereby suppressing any signal components unrelated to the reference signal frequency. This includes background noise from the environment and mechanical background introduced by factors such as motor imbalance and environmental vibration, retaining only the response signal component that is frequency and phase-locked with the high-frequency detection sequence, and outputting an in-phase integral value. and an orthogonal integral value .

[0058] To further explain, for the first There are 1 data frames, and their time length is 1. Its corresponding center frequency is Let the acceleration signal within this frame be... ,in As mentioned earlier, the time variable is specifically the in-phase component obtained by demodulation, which is the actual heat permeation pulsation index parameter. and For the first The center frequency is The data frame, which serves as the integration variable, is used here to extract the equivalent acceleration amplitude. The normalized carrier of the subsequent mechanical domain reference local oscillator signal is Its orthogonal components are The generated in-phase and orthogonal integrals are then calculated as follows:

[0059]

[0060] The identification module generates a multidimensional complex vector sequence based on the above integral processing, which is a complex array, and its i-th... The elements are In the formula Let be the imaginary unit of the complex number, and calculate the system response at each frequency point. This is achieved by calculating the complex vector. The modulus is calculated and compared with the amplitude of the mechanical domain reference local oscillator signal to determine the relative energy reduction ratio characterizing the amplification or attenuation of the mechanical excitation by the device wall structure, i.e., the system gain. The argument of this complex vector is then calculated to determine the physical time lag difference characterizing the vibration delay of the device wall, i.e., the system phase shift. Specifically, the gain calculated in this step... and phase The calculation formula is as follows:

[0061]

[0062] In the formula, For system gain; For system phase; The value represents the normalized constant of the equivalent input acceleration amplitude corresponding to the generated reference local oscillator signal, in units of... ; and These are the in-phase and orthogonal integral values ​​obtained from the previous equation, respectively; It is a four-quadrant arctangent function. Due to the existence of low-frequency, large-amplitude macroscopic sweep frequency sequences, each data frame actually corresponds to a specific sweep frequency.

[0063] Therefore, by concatenating the discrete calculation results of all data frames and performing continuous frequency domain interpolation fitting, a curve can be plotted with frequency as the continuous abscissa and gain and phase shift as the ordinate. This curve is the unique indicator of the mechanical vibration transfer function that induced the micro-excitation characteristics of the equipment's metal cladding by high-frequency perturbation injection. This continuous mechanical vibration transfer function is a complex function. It fully describes the linear dynamic relationship between high-frequency disturbances of the stirrer and the vibration response of the vessel wall.

[0064] For example, continuing with the example of S2 above, select the corresponding The index number collected at time s is The data frame is processed. The center sweep frequency of this frame is... Hz, containing 200 acceleration sampling points, with a time length of The identification module generates a digital cosine wave with a frequency of 50 Hz and a digital sine wave as reference signals.

[0065] Assuming that the acceleration data of this frame acquired in step S2, after analysis, can be equivalent to the component at a frequency of 50 Hz as follows: m / s², the remainder being noise of other frequencies. Quadrature demodulation and integration are performed with the reference signal. Based on the lock-in amplification principle, noise is filtered out from the integration result, yielding... m / s², m / s².

[0066] Assuming the equivalent acceleration amplitude of the reference signal Normalized to 0.1 m / s². Based on this, the transfer function value at this frequency is calculated to have a relative energy reduction ratio of 0.1 m / s². The physical time lag difference is rad, which is -60°. This calculation result. This becomes a data point of the mechanical vibration transfer function at a frequency of 50 Hz. Repeating this process for all 5000 data frames yields the complete function curve.

[0067] In one embodiment of the present invention, step S4 includes the following steps: The independent digital phase-locked amplifier unit continuously applies the thermal domain reference local oscillator signal to the micro thermal flow fluctuation frame data set, reproduces and executes the orthogonal demodulation dot multiplication and cumulative mean integration operation mechanism of the equivalent mechanical demodulation processing rule; The redundant physical parameter components induced by the penetration of the medium layer are stripped off, and the remaining physical environment thermal interference is filtered out by cumulative mean integral calculation. The in-phase and orthogonal components after orthogonal demodulation are extracted as the thermal penetration pulsation index parameters. By integrating and mapping real heat permeation pulsation index parameters, we can present quantitative chart results that demonstrate the dynamic exchange of transient heat energy at the fluid medium interface, and generate a thermodynamic thermal admittance function free from perturbation-induced displacement contamination.

[0068] Specifically, step S4 is executed in parallel with or following step S3. This step is also performed by the identification module in the central data processing unit, but its processing target shifts to the thermodynamic domain. Through a technical path parallel to S3, the thermodynamic response function of the system is analyzed. To perform demodulation in the thermodynamic domain, the identification module retrieves a high-frequency detection sequence from memory that has no phase slip on the time axis and whose structural parameters are equivalent to the mechanical domain reference local oscillator signal.

[0069] The sequence is fed into the second-stage digital phase-locked amplifier (PLA) unit within the identification module and designated as the thermodynamic domain reference local oscillator signal for thermodynamic signal demodulation. The second-stage PLA unit is identical in algorithm structure and function to the first stage and is an independent computation channel; the thermodynamic domain reference signal, serving as a digital copy, ensures consistency of reference standards between the two demodulation channels. This redundant backup and independent import operation eliminates system errors introduced by inconsistencies in reference standards between the mechanical and thermodynamic domain demodulation processes.

[0070] The identification module mobilizes the computing resources of the second-stage digital lock-in amplifier unit to take the microscopic heat flow fluctuation frame dataset generated in step S2 as the input signal. For each heat flow data frame in the set, this unit completely reiterates and executes the entire computational process equivalent to the mechanical demodulation processing in the previous step S3, that is, iteratively applying orthogonal deconvolution dot product and cumulative mean integral.

[0071] This process multiplies the input microscopic heat flow frame data with the thermodynamic domain reference local oscillator signal and its orthogonal components, and then integrates the results to achieve phase-sensitive detection in the thermodynamic dimension. The effect is to remove redundant physical parameter components induced by factors such as macroscopic temperature changes, environmental thermal radiation fluctuations, and random collisions of fluid turbulence with the outer wall of the jacket. It retains the real heat permeation pulsation index parameters induced by microflow vortices touching the inner membrane boundary of the reaction system driven by high-frequency detection sequences from the complex mixed thermal signal.

[0072] Specifically, the actual heat permeation pulsation index parameter is the in-phase component obtained by demodulation. and orthogonal components For the first The center frequency is The data frame, assuming the microscopic heat flow signal is... Assuming that the normalized constant of the equivalent heat flux fluctuation amplitude is extracted. The normalized carrier of the subsequent thermal domain reference local oscillator signal is = The integral value of its thermal domain is calculated as follows:

[0073]

[0074] Finally, by integrating the demodulation results of all data frames—that is, the thermal response gain and phase at each frequency point—this module generates and outputs a thermodynamic thermal admittance function that can visualize the ability of the operating medium to absorb and even exchange transient thermal energy at the microscopic level. This thermodynamic thermal admittance function is also a complex function. It evaluates the dynamic performance of interfacial heat transfer from a thermodynamic perspective, and its gain and phase The calculation method is as follows:

[0075]

[0076] In the formula, For the gain of the thermodynamic domain system; Phase of the thermodynamic domain system; and These are the in-phase and orthogonal integral values ​​obtained from the previous equation, respectively; is the normalization constant for the amplitude of the equivalent heat flux fluctuation; atan2 is the arctangent function in the four quadrants.

[0077] For example, continue the aforementioned calculation process and process the same index number as in the example of step S3. exist The microscopic heat flux fluctuation frame data was synchronously acquired at time s. The center sweep frequency of this data frame is also... Hz. The identification module uses an identical 50 Hz digital reference signal.

[0078] Assuming the heat flux data collected by S2 in this frame can be analyzed to have an effective component at a frequency of 50 Hz that is equivalent to... W / m², which indicates that the amplitude and phase of the thermal response differ from those of the mechanical response. The second-stage digital lock-in amplifier unit performs quadrature demodulation to obtain the integral result. W / m², W / m².

[0079] Assuming the equivalent heat flux amplitude of the reference signal The system was calibrated and normalized to 20 W / m². Based on this, the thermal admittance function at this frequency was calculated, and the gain was... Phase is rad, which is -30°. This calculation result. This becomes a data point for the thermodynamic thermal admittance function at a frequency of 50 Hz. Repeating this parallel processing for all data frames yields the complete thermodynamic thermal admittance function curve, which can then be used for cross-domain decoupling calculations in subsequent steps.

[0080] In one embodiment of the present invention, step S5 includes the following steps: The main core engine of the deployment device retrieves the comprehensive heat transfer admittance quantization set recorded in the frequency domain coordinate system as the basic divisor sequence term, which is the thermodynamic thermal admittance function recorded in the frequency domain coordinate system. The decoupled deconvolution numerical array is divided by the basic dividend sequence terms and the converted dynamic external resistance factor sequence, according to the requirement of iterative division operation with a predetermined bandwidth step; By using the frequency domain decoupling algorithm, the additional heat transfer gain superimposed by the random impact feedback of the fluid on the outer wall of the system is filtered out, and the data array information that is not affected by the pseudo-enhanced heat transfer interference caused by the redundant mechanical vibration of the wrapping layer is purely retained. The thermo-mechanical coupling composite spectrum of the target system is then fitted in reverse.

[0081] Specifically, after the parallel computations of steps S3 and S4 are completed, step S5 is initiated. This step is executed by a spectral analysis and linkage solution module deployed within the central data processing unit. Its goal is to synthesize a spectrum that reflects only the pure inner membrane heat transfer characteristics by mathematically decoupling the acquired dual-domain dynamic function and eliminating physical cross-domain interference. This module imports the mechanical vibration transfer function generated in step S3 and performs feature identification on its spectral lines. By applying a peak detection algorithm, such as a discrimination method based on the sign change of the first derivative combined with amplitude threshold and frequency interval, multiple discrete mechanical resonance peak envelopes and trough convergence point clusters are identified and extracted from the amplitude-frequency characteristic curve of the mechanical vibration transfer function.

[0082] The mechanical resonance peak envelope refers to the local maxima region appearing on the amplitude-frequency curve, representing the sensitive response of the equipment structure to mechanical excitation at that frequency. The magnitude of these characteristics is related to the efficiency with which the mechanical kinetic energy generated by the stirrer crosses the metal boundary layer and diffuses to the external cooling medium at a specific frequency. Therefore, the module will use the amplitude sequence of the mechanical vibration transfer function. As the core input, the intensity of the dissipation process is calculated through a preset physical model, generating a frequency-dependent, dimensionless dynamic external resistance factor sequence.

[0083] This sequence essentially quantifies the equivalent convective heat transfer enhancement effect contributed by mechanical vibration to the overall heat transfer process. This dynamic external resistance factor sequence... The physical meaning of describes the change in equivalent thermal resistance caused by the enhanced external fluid disturbance due to the vibration of the vessel wall. Because mechanical vibration leads to a thinner boundary layer and intensified external fluid disturbance, an additional amplification of heat transfer capacity is generated during measurement. In this embodiment, it is simplified to a value related to the amplitude. A sequence of real numbers that are directly proportional, i.e.:

[0084] in, The thermo-mechanical coupling coefficient characterizes the mechanism of the conversion intensity of mechanical vibration into equivalent thermal resistance. This value can be determined through finite element simulation or experimental calibration, with a typical range of 0.05 to 0.3. The physical basis for this value lies in the fact that it depends on the acoustic impedance characteristics of the reactor's metal cladding and the Prandtl number of the external fluid. If... A value below 0.05 indicates that mechanical excitation has extremely difficulty penetrating the metal wall to produce a peeling effect on the outer boundary layer; if A value higher than 0.3 indicates strong fluid-structure interaction heat transfer. In this embodiment, the specific value of this coefficient is obtained by approximating it using the least squares method through offline heat transfer temperature difference experiments under no-load and full-load conditions in a standard fluid medium, such as pure water, thereby solidifying the parasitic thermal resistance transfer characteristics of the system.

[0085] The spectral analysis linkage solution module deploys its collaborative computing core, which divides the quantized set of comprehensive heat transfer admittance recorded by the thermodynamic thermal admittance function generated in step S4 with the dynamic external resistance factor sequence obtained in the previous step by a decoupled deconvolution numerical array calculated in steps with a predetermined bandwidth. In the frequency domain, this decoupled deconvolution operation is equivalent to performing a pointwise division operation on elements with the same index in two complex arrays.

[0086] Based on the decoupling and deconvolution numerical array division operation, this module mathematically eliminates the spurious thermal resistance effect of the outer wall boundary layer disturbance, caused by the mechanical process of random collisions between the fluid turbulence and the vessel wall outside the jacket, which is transmitted through the vessel wall vibration and mixed with the heat flow signal, from the total heat transfer admittance. The data array left after the operation is a data set that is no longer disturbed by the extralayer parasitic dynamic response. Then, by connecting the calculation results of all frequency points, a thermo-mechanically coupled composite spectrum that purely characterizes the internal convective exchange state of the stirred tank is formed. Specifically, this thermo-mechanically coupled composite spectrum... Obtained through the following frequency domain division operation:

[0087] In the formula, It is a continuous frequency variable; The thermo-coupled composite spectrum is to be determined; Let be the thermodynamic thermal admittance function, and its amplitude be . Phase is ; It is a dynamic external resistance factor sequence; This is the amplitude sequence of the mechanical vibration transfer function; The mechanism-thermal coupling coefficient; is the base of the natural logarithm; It is the imaginary unit of complex numbers.

[0088] For example, continuing the previous steps, we focus on the calculation at a frequency of 50 Hz. From the example in step S3, we know that the amplitude of the mechanical vibration transfer function at 50 Hz... As can be seen from the example in step S4, the thermodynamic thermal admittance function at 50 Hz is [value missing]. Its amplitude phase Assuming that the pre-defined mechanism-thermal coupling coefficient is achieved through prior calibration... Calculate the dynamic external resistance factor at 50 Hz. Since this factor is a real number in this model, its phase is 0°.

[0089] Subsequently, a decoupling and deconvolution numerical array division operation was performed to calculate the value of the thermo-coupled composite spectrum at 50 Hz. Its amplitude is Its phase is Therefore, after decoupling, the thermo-coupled composite spectrum characterizing the heat transfer properties of the pure inner membrane is at 50 Hz. This calculation process will be applied to all frequency points to generate a complete spectrum.

[0090] In one embodiment of the present invention, step S6 includes the following steps: The system's underlying average environmental rotation speed is recorded simultaneously during the synchronous operation phase when the target inner membrane heat transfer coefficient is generated, and the current operating condition Reynolds number associated with the instantaneous flow field characteristics is calculated. The solved target inner membrane heat transfer coefficient is combined with the current operating condition Reynolds number to establish a strong link dependency relationship. Combination encapsulation is then performed to establish the interlocked key-value pair data cluster required for a one-to-one linear query and retrieval mechanism. Multiple sets of interlocked key-value pairs data clusters, established iteratively at time points according to operating condition sequences, are written to the storage pool, and the storage units are refreshed according to structured management rules to generate a dynamic heat transfer calibration data model with open interfaces for external exposure.

[0091] Specifically, after generating the thermo-coupled composite spectrum, the method proceeds to step S6, which is executed by the system data analysis and model generation module. Its purpose is to extract the key static heat transfer parameters required for engineering applications from the dynamic spectrum purified in the previous steps and to complete the final model construction. This module sets the dependent variable analysis conditions in the frequency domain, that is, on the frequency axis of the thermo-coupled composite spectrum, it continuously approaches the position of zero frequency along the axis-centered coordinate system and retrieves the changing trend of the function value during this approach.

[0092] According to heat transfer theory, when the perturbation frequency approaches zero, the response is equivalent to a steady-state response. Therefore, this module analyzes the behavior of the thermo-coupled composite spectrum in the low-frequency range and automatically locates the low-frequency DC approximation region where the function value no longer fluctuates significantly with frequency. This low-frequency DC approximation region refers to a flat frequency band below a certain threshold, such as 0.1 Hz, set based on the fact that the thermal inertia time constant of industrial-grade stirred tanks is typically on the order of 10 s. When the physical perturbation period is greater than 10 s, i.e., the frequency is below 0.1 Hz, the fluid temperature gradient near the inner membrane of the reactor has sufficient time to follow the mechanical action to reach pseudo-thermal equilibrium. At this point, the calculated dynamic thermal admittance can be equivalently replaced by the traditional static inner membrane heat transfer coefficient.

[0093] Within this section, the module extracts a limiting value that conforms to the macroscopic steady-state heat transfer law, representing the steady-state heat transfer capacity; this is the initial value of the limiting heat transfer coefficient convergence in steady state. This represents the dimensionless heat transfer capacity under ideal steady-state conditions, which is obtained by interpolation or by averaging within the low-frequency DC approximation range. The calculation formula is:

[0094] in, This represents the initial value of the limiting heat transfer coefficient at steady-state convergence. For thermo-coupled composite spectra at frequency The amplitude at that point.

[0095] See Figure 3 As shown in the figure, this diagram visually illustrates the quantitative geometric relationship between multiphysics spectral decoupling and feature segment extraction in the embodiments of this application. The dashed line represents the mechanical vibration transfer function obtained from mechanical domain demodulation, which exhibits a significant mechanical resonance peak envelope near 40 Hz, reflecting the nonlinear amplification effect of mechanical disturbance energy dissipation to the boundary layer. The dotted line represents the thermodynamic thermal admittance function obtained from direct detection in the thermodynamic domain; affected by mechanical crosstalk, it shows spurious bulges in the mid-to-high frequency range. The solid line represents the purely thermodynamically coupled composite spectrum synthesized after performing deconvolution division on the basic dividend sequence and the dynamic external resistance factor sequence. .

[0096] It is evident that, after algorithmic stripping, the solid line effectively suppresses the spurious thermal resistance increment in the mechanical resonant frequency band, restoring the true heat transfer state. Further reading... Figure 3 The gray shaded area on the left represents the low-frequency DC approximation region automatically located by the system when the frequency approaches zero. This region is used to extract the pure initial steady-state value of the limiting heat transfer coefficient. The module uses this frequency point as a reference to define its outer boundary. Figure 3 The working band integration interval is marked by a dotted line in the middle. This module calculates the total geometric area of ​​the two-dimensional space enclosed by the solid line of the purely thermo-coupled composite spectrum and the frequency axis within this interval, which is the area segment below the solid line in the figure. This area physically represents the true value of the total dynamic energy exchange of the inner membrane heat transfer process within this characteristic frequency range after excluding external random impact interference.

[0097] The module performs a smoothed, normalized, integral-average transformation on the area value. This is understandable; it's a calculation method that reduces the randomness of single-point noise by integrating and averaging information over a frequency band. The result of this transformation is the target inner membrane heat transfer coefficient, which truly reflects the interfacial heat transfer capacity determined by the turbulent physical disturbances within the stirred tank under current macroscopic operating conditions. This target inner membrane heat transfer coefficient is a core output parameter. In the operating frequency band range The derived formula within is:

[0098] In the formula, and These represent the lower and upper frequency limits of the operating band, respectively. For thermo-coupled composite spectra at frequency The amplitude at that point; It is an absolute dimensionless conversion coefficient that includes the inherent geometric dimensions of the system, such as the characteristic length of the heat transfer area and the sensor hardware gain. Its actual value can be obtained through a benchmark experiment. That is, before the system is put into production, a standard medium with known properties, such as pure water, is injected. At a fixed Reynolds number, such as Re=30000, a benchmark value of the absolute overall heat transfer coefficient is measured using traditional steady-state convective heat transfer formulas, such as classical Newton's law of cooling or the Wilson regression method. Dividing this benchmark value by the dimensionless area characteristic value obtained by integration under the benchmark operating condition of this system yields the inverse solution for the solidification process. In subsequent industrial applications, this coefficient remains unchanged as a constant.

[0099] The data parsing and model generation module performs data packaging and recording operations, concatenating the currently calculated target inner membrane heat transfer coefficient with the actual operating conditions of the stirring generator at the time of coefficient generation, forming a data cluster of matching query interlocked key-value pairs with a one-to-one correspondence. The core operating parameter is the actual Reynolds number, a dimensionless number calculated from stirring speed, fluid density, and viscosity. Therefore, this data cluster can be represented as a structured data record, for example, {Reynolds_Number:35000, h_in:1250.5 W / (m²·K)}. This data cluster is then recorded and written to a solid-state data analysis platform, such as a database based on Structured Query Language (SQL) or NoSQL. This write operation completes the self-updating iteration of the heat transfer calibration model and outputs a dynamic heat transfer calibration data model that is publicly available and queried by the entire plant control decision-making body. This model, as an ever-expanding dataset or function model, fully records the relationship curve between the Reynolds number and the inner membrane heat transfer coefficient.

[0100] For example, the preceding steps have generated a complete thermo-coupled composite spectrum. The data analysis module first analyzed its amplitude in the low-frequency range of 0 Hz to 0.1 Hz, finding that it remained stable at... The module determines the initial, convergent steady-state value of the limiting heat transfer coefficient based on this. .

[0101] The module defines a working band centered on the spectral peak, for example... Hz to Hz. Calculations show that the area enclosed by the amplitude curve of the thermocoupled composite spectrum within this interval and the frequency axis is 6.98. The module then performs a smoothed normalized integral-average transform. Assume the system has preset dimension transformation and calibration coefficients. W / (m²·K). Therefore, the final calculated target inner membrane heat transfer coefficient is: W / (m²·K).

[0102] At this point, the system records the current time period, for example... s to Within s, the average rotational speed of the low-frequency, large-amplitude macroscopic sweep frequency sequence is 140 RPM, and the corresponding Reynolds number is calculated from the physical property database. Finally, a key-value pair data cluster {Reynolds_Number: 52000, h_in: 1256.4 W / (m²·K)} is generated and stored in the dynamic heat transfer calibration data model. As the frequency sweep process continues, this model will be continuously filled with new data clusters, forming a complete... Calibration curve as Reynolds number varies.

[0103] See appendix Figure 4 The figure presents a visualization of the final generated dynamic heat transfer calibration data model. The horizontal axis represents the Reynolds number, reflecting the flow field state, and the vertical axis represents the derived target inner membrane heat transfer coefficient. The discrete dots in the figure represent the massive clusters of measured key-value pairs calculated frame by frame by continuously capturing the responses of high-frequency probes during a single continuous large-amplitude frequency sweep operation.

[0104] As the macroscopic frequency sweep sequence smoothly transitions the equipment's operating conditions from low to high speed, these measured data points are densely clustered within the coordinate system. The black square nodes specifically highlighted in the figure correspond to the example operating point in the aforementioned textual deduction, such as... The solid line running through the scatter plot in the figure represents the dynamic heat transfer calibration data model curve generated by the data parsing and model generation module after performing smooth regression on these discrete data. This curve not only completely filters out discrete and free abnormal test noise, but also reveals the physical law of the nonlinear increase of the inner membrane heat transfer coefficient under dynamic process parameters with high continuity, realizing a technological leap from traditional multi-point steady-state offline calibration to single-time variable-condition continuous online mapping.

[0105] See appendix Figure 2 The present invention also proposes an inner membrane heat transfer coefficient calibration system based on Wilson regression, comprising the following modules: The composite excitation signal generation module acquires the basic control parameters of the target fluid reaction system and the target frequency band range to generate a low-frequency large-amplitude macroscopic frequency sweep sequence. It then combines the preset phase encoding feature matrix table to generate and superimpose a high-frequency detection sequence, and outputs a composite broadband excitation signal and a master clock trigger signal. The multi-physics data synchronous acquisition module feeds the composite broadband excitation signal into the device motor to apply actuation disturbance, and synchronously truncates the acquired multi-physics continuous data stream according to the master clock trigger signal, and outputs the set of device wall acceleration frame data and the set of micro heat flow fluctuation frame data. The mechanical transfer function demodulation module extracts the high-frequency detection sequence as the mechanical domain reference local oscillator signal, performs phase-sensitive filtering on the set of acceleration frame data of the device wall, and calculates and outputs the mechanical vibration transfer function. The thermodynamic admittance function demodulation module imports the high-frequency detection sequence into the thermodynamic domain and converts it into a thermodynamic domain reference local oscillator signal. It then applies parallel thermodynamic domain detection to the microscopic heat flow fluctuation frame data set to obtain the thermodynamic thermal admittance function. The spectral domain decoupling and purification module extracts the discrete envelope features of the mechanical vibration transfer function to generate a dynamic external resistance factor sequence. The dynamic external resistance factor sequence is then divided by the decoupling and deconvolution numerical array of the thermodynamic thermal admittance function to synthesize a thermo-coupled composite spectrum. The heat transfer coefficient analysis and modeling module locates the low-frequency approximation segment of the thermo-coupled composite spectrum to extract the initial steady-state value of heat transfer, calculates the area of ​​the continuous discrete data set in the characteristic working band interval and derives the target inner membrane heat transfer coefficient, and combines the target inner membrane heat transfer coefficient with the current operating condition Reynolds number to generate a dynamic heat transfer calibration data model.

[0106] Each of the modules can be implemented in whole or in part through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0107] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for calibrating the inner membrane heat transfer coefficient based on Wilson regression, characterized in that, Includes the following steps: S1. Obtain the basic control parameters of the target fluid reaction system and the target frequency band range to generate a low-frequency large-amplitude macroscopic frequency sweep sequence. Combine the preset phase coding feature matrix table to generate and superimpose a high-frequency detection sequence, and output a composite broadband excitation signal and a master clock trigger signal. S2. Feed the composite broadband excitation signal into the device motor to apply the actuation disturbance, and synchronously cut off the continuous data stream of the multi-physics field according to the master clock trigger signal, and output the set of device wall acceleration frame data and the set of micro heat flow fluctuation frame data. S3. Extract the high-frequency detection sequence as the mechanical domain reference local oscillator signal, perform phase-sensitive filtering on the set of acceleration frame data of the device wall, and calculate and output the mechanical vibration transfer function. S4. The high-frequency detection sequence is introduced into the thermodynamic domain and converted into a thermodynamic domain reference local oscillator signal. Parallel thermodynamic domain detection is applied to the microscopic heat flow fluctuation frame data set to obtain the thermodynamic thermal admittance function. S5. Extract the discrete envelope features of the mechanical vibration transfer function to generate a dynamic external resistance factor sequence. Divide the dynamic external resistance factor sequence with the thermodynamic thermal admittance function by performing decoupling and deconvolution numerical array to synthesize a thermo-coupled composite spectrum. S6. Locate the low-frequency approximation segment of the thermo-coupled composite spectrum to extract the initial steady-state value of heat transfer, calculate the area of ​​the continuous discrete data set in the characteristic working band interval and derive the target inner membrane heat transfer coefficient, and combine the target inner membrane heat transfer coefficient with the current operating condition Reynolds number to generate a dynamic heat transfer calibration data model.

2. The method for calibrating the inner membrane heat transfer coefficient based on Wilson regression according to claim 1, characterized in that, The process of generating and superimposing a high-frequency detection sequence by combining a preset phase-encoded feature matrix table includes the following steps: Read the instantaneous frequency distribution and amplitude parameters of a low-frequency, large-amplitude macroscopic sweep frequency sequence; Using instantaneous frequency distribution and amplitude parameters as query indexes, access the preset phase coding feature matrix table and extract the matching symbol switching rate and modulation amplitude; A high-frequency detection sequence is generated based on the symbol switching rate and modulation amplitude. At the same time, the high-frequency detection sequence is superimposed and fused onto the low-frequency large-amplitude macroscopic sweep frequency sequence waveform to form a composite broadband excitation signal.

3. The method for calibrating the inner membrane heat transfer coefficient based on Wilson regression according to claim 1, characterized in that, The continuous data stream of the acquired multiphysics field is synchronously truncated according to the master clock trigger signal, including the following steps: Activate the piezoelectric acceleration sensing element installed on the outer wall and the thin film heat flow sensing node laid on the inner side, and synchronously sample at a sampling frequency of not less than twice the maximum symbol switching rate of the high-frequency detection sequence to obtain the continuous wall vibration data stream and the continuous micro heat flow fluctuation data stream. The master clock trigger signal is used as the time gating cut reference to force synchronous truncation of the continuous wall vibration data stream and the continuous micro heat flow fluctuation data stream. The truncated stream data is integrated and output as a set of wall acceleration frame data and a set of microscopic heat flux fluctuation frame data with temporal dimension binding consistency.

4. The method for calibrating the inner membrane heat transfer coefficient based on Wilson regression according to claim 1, characterized in that, Perform phase-sensitive filtering on the frame data of the vessel wall acceleration, and calculate the output mechanical vibration transfer function, including the following steps: Using the mechanical domain reference local oscillator signal and its orthogonal components, the device wall acceleration frame data set is subjected to multi-cycle coherent interactive orthogonal demodulation and rectangular time window sliding window integration to filter irregular environmental mechanical background noise and output a multidimensional complex vector sequence. Solve for the relative energy reduction ratio and physical time lag difference of the multidimensional complex vector sequence relative to the mechanical domain reference local oscillator signal; The relative energy reduction ratio and physical time lag difference of each frequency node are calculated and combined to draw discrete characteristic curves, and finally a mechanical vibration transfer function with a unique indicator of the metal excitation response state of the equipment is generated.

5. The method for calibrating the inner membrane heat transfer coefficient based on Wilson regression according to claim 1, characterized in that, Applying parallel thermodynamic domain detection to the microscopic heat flow fluctuation frame data set to obtain the thermodynamic thermal admittance function includes the following steps: The independent digital phase-locked amplifier unit continuously applies the thermal domain reference local oscillator signal to the micro thermal flow fluctuation frame data set, reproduces and executes the orthogonal demodulation dot multiplication and cumulative mean integration operation mechanism of the equivalent mechanical demodulation processing rule; The redundant physical parameter components induced by the penetration of the medium layer are stripped off, and the remaining physical environment thermal interference is filtered out by cumulative mean integral calculation. The in-phase and orthogonal components after orthogonal demodulation are extracted as the thermal penetration pulsation index parameters. By integrating and mapping real heat permeation pulsation index parameters, a quantitative chart of the fluid medium interface's ability to dynamically exchange transient heat energy is generated, producing a thermodynamic thermal admittance function free from perturbation-induced displacement contamination.

6. The method for calibrating the inner membrane heat transfer coefficient based on Wilson regression according to claim 1, characterized in that, Extracting the discrete envelope features of the mechanical vibration transfer function to generate a dynamic external resistance factor sequence includes the following steps: The mechanical vibration transfer function amplitude-frequency characteristic scanning array is identified and extracted by applying an extreme value search and judgment algorithm. Capture the features of isolated mechanical resonance peak envelopes and valley convergence point clusters within the line segment of the amplitude-frequency characteristic scanning array results. The equivalent disturbance intensity parameter of the mechanical kinetic energy diffusion through the metal boundary layer to the external auxiliary cooling source is measured. Based on a specific associated physical model, numerical array analysis is performed to output a dynamic external resistance factor sequence that quantifies the unsteady thermal resistance fluctuation of the system.

7. The method for calibrating the inner membrane heat transfer coefficient based on Wilson regression according to claim 1, characterized in that, The dynamic external resistance factor sequence is divided by the thermodynamic thermal admittance function through a decoupling and deconvolution numerical array to synthesize a thermodynamically coupled composite spectrum, including the following steps: The main core engine of the deployment device retrieves the comprehensive heat transfer admittance quantization set recorded in the frequency domain coordinate system as the basic divisor sequence term, which is the thermodynamic thermal admittance function recorded in the frequency domain coordinate system. The decoupled deconvolution numerical array is divided by the basic dividend sequence terms and the converted dynamic external resistance factor sequence, according to the requirement of iterative division operation with a predetermined bandwidth step; By using the frequency domain decoupling division algorithm, the false thermal resistance effect of the adhesion caused by the random impact feedback of the fluid on the outer wall of the system is filtered out, and the data array information that is not affected by the redundant response of the wrapping layer is purely retained and the thermo-mechanical coupling composite spectrum of the target system is inversely fitted.

8. The method for calibrating the inner membrane heat transfer coefficient based on Wilson regression according to claim 1, characterized in that, Locating the low-frequency approximation region of the thermo-coupled composite spectrum to extract the initial steady-state values ​​of heat transfer includes the following steps: A continuous observational decay condition is constructed within the computer, where the dependent variable continuously approaches the zero baseline along the vertical axis of the computing engine. The continuous observation of the approaching attenuation condition is superimposed and substituted into the dynamic slope evaluation and judgment module of the thermo-coupled composite spectrum to retrieve and determine the low-frequency approximation section where the fluctuation amplitude of the output curve converges smoothly. The low-frequency approximation segment obtained by truncation is extracted and stripped of its inherent physical phase change limit, the basic limit heat transfer coefficient convergence steady-state initial value, and the heat transfer steady-state initial value specific to the steady-state conduction limit scenario is solidified and constructed.

9. The method for calibrating the inner membrane heat transfer coefficient based on Wilson regression according to claim 1, characterized in that, The calculation of the area of ​​the continuous discrete data set within the characteristic operating band interval and the derivation of the target inner membrane heat transfer coefficient include the following steps: Using the original location of the extracted heat transfer steady-state initial value as the geometric reference center point, a bidirectional extrapolation band fixed-frequency expansion is implemented to delineate the outer continuous discrete point cluster as the characteristic working band interval; In the generated characteristic working band interval, the two-dimensional area below the frequency distribution axis of the frequency band correlation data curve is integrated and taken as the area of ​​the continuous discrete data set representing the total true value of heat transfer. The area of ​​the continuous discrete data set mapped by the calculation results is sent to the internally configured smoothing normalization integral average transformation functional area module to perform coefficient back-reasoning correction operation, suppress and eliminate isolated and accidental abnormal test noise, and then derive the high-precision corrected target inner membrane heat transfer coefficient.

10. The method for calibrating the inner membrane heat transfer coefficient based on Wilson regression according to claim 1, characterized in that, The target inner membrane heat transfer coefficient and the current operating condition Reynolds number are combined and packaged to generate a dynamic heat transfer calibration data model, including the following steps: The system's underlying average environmental rotation speed is recorded simultaneously during the synchronous operation phase when the target inner membrane heat transfer coefficient is generated, and the current operating condition Reynolds number associated with the instantaneous flow field characteristics is calculated. The solved target inner membrane heat transfer coefficient is combined with the current operating condition Reynolds number to establish a strong link dependency relationship. Combination encapsulation is then performed to establish the interlocked key-value pair data cluster required for a one-to-one linear query and retrieval mechanism. Multiple sets of interlocked key-value pairs data clusters, established through iterative time point sequences of operating conditions, are written to the storage pool. The storage units are then refreshed according to structured management rules to generate a dynamic heat transfer calibration data model with open interfaces that can be exposed to the outside world.

Images