Remote intelligent diagnosis and dynamic calibration system and method for pressure gauge

Abstract

The invention relates to the technical field of dynamic calibration and multi-parameter cooperative compensation, in particular to a remote intelligent diagnosis and dynamic calibration system and method for a pressure gauge, and aims to solve the problem that a traditional pressure gauge adopts a fixed calibration mode, and the calibration precision is high. The problem that the measurement accuracy is insufficient due to the fact that the viscosity change of a measured medium, the pressure fluctuation frequency and an aging accumulation error caused by accumulated use of instruments cannot be dynamically compensated is solved, real-time viscosity data are collected through a viscosity sensor, and pressure fluctuation frequency data are extracted through a pressure sensor; the counter records the starting times and the operation duration of the instrument and integrates the starting times and the operation duration into accumulated use frequency data; and the dynamic calibration coefficient correction model unit receives the three types of data, performs normalization processing, inputs the data into a pre-trained machine learning regression model, analyzes the nonlinear influence of viscosity on fluid resistance, the coupling effect of pressure fluctuation frequency and an aging accumulative error, performs fusion to generate a dynamic calibration coefficient, realizes dynamic compensation, and improves the measurement accuracy of a pressure gauge.

Description

Technical Field

[0001] This invention relates to the field of dynamic calibration and multi-parameter collaborative compensation technology, and more specifically, to a remote intelligent diagnostic and dynamic calibration system and method for pressure gauges. Background Technology

[0002] Dynamic calibration and multi-parameter collaborative compensation is an important technology, specifically applied to the precise measurement and remote calibration of pressure gauges in industrial scenarios. Its core principle is to improve pressure measurement accuracy by adapting to multiple dynamic factors, thus meeting the core requirements of industrial production for the accuracy and stability of pressure gauge measurements. In industrial measurement environments, the viscosity of the measured medium affects fluid resistance, the frequency of pressure fluctuations interferes with the stability of sensor response, and the cumulative frequency of instrument use leads to component aging and accumulated errors. These factors cause the pressure gauge's measurement error to dynamically change with operating conditions and usage status. Because traditional pressure gauges use fixed calibration standards, they cannot adjust calibration parameters in real time according to these dynamic factors, resulting in deviations between measured values ​​and actual pressure. This affects the accuracy of pressure monitoring in industrial production and makes it difficult to meet the needs of high-precision production control. To solve this problem, we provide a remote intelligent diagnostic and dynamic calibration system and method for pressure gauges. Summary of the Invention

[0003] The purpose of this invention is to provide a remote intelligent diagnostic and dynamic calibration system and method for pressure gauges to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, a remote intelligent diagnostic and dynamic calibration system for pressure gauges is provided, comprising: The viscosity sensor unit measures the viscosity of the medium being measured in real time and outputs real-time viscosity data through signal conversion; The pressure sensor unit monitors the pressure signal and extracts and outputs pressure fluctuation frequency data from the pressure signal through the built-in frequency analysis module; The counter unit records the cumulative usage frequency of the instrument, including the number of starts and the running time, and outputs the cumulative usage frequency data; The dynamic calibration coefficient correction model unit calculates dynamic calibration coefficients based on real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data, and applies these dynamic calibration coefficients to correct the pressure measurement value. This unit includes an embedded processor and a pre-stored dynamic calibration model. The dynamic calibration model is trained using historical experimental data. The embedded processor takes real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data as input to the dynamic calibration model. The dynamic calibration model is a regression algorithm based on machine learning. By analyzing the impact of viscosity changes on fluid resistance, the coupling effect of pressure fluctuation frequency on sensor response stability, and the cumulative error caused by cumulative usage frequency to instrument aging, it internally performs multi-parameter weight adjustment calculations and outputs dynamic calibration coefficients. These coefficients are then multiplied by the pressure fluctuation frequency to generate the calibrated pressure value. This process dynamically compensates for the real-time viscosity of the measured medium, the pressure fluctuation frequency, and the cumulative usage frequency of the instrument, thus achieving calibration.

[0005] The second objective of this invention is to provide a method for implementing a remote intelligent diagnostic and dynamic calibration system for a pressure gauge, comprising the following steps: S1. The vibrating probe contacts the medium being tested and generates a mechanical vibration signal based on the change in viscous resistance. This signal is then converted into an electrical signal by a piezoelectric converter, and further processed digitally by an analog-to-digital converter to filter out environmental noise and output real-time viscosity data. S2. The pressure signal is collected by the strain gauge and amplified by the amplifier. The spectrum of the amplified signal is analyzed by the Fast Fourier Transform algorithm to identify the dominant fluctuation frequency component. After eliminating high-frequency noise by the filter, the pressure fluctuation frequency data is output. S3. The instrument running time is accumulated using a hardware timer, and the number of startups is recorded by an event counter. The two are integrated into a cumulative usage frequency data in the form of a comprehensive numerical index, stored in non-volatile memory, and output through a digital interface. S4. The embedded processor normalizes the input real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data, and inputs them into the pre-trained dynamic calibration model. S5. The dynamic calibration model analyzes the nonlinear effect of viscosity on fluid resistance, the coupling effect of pressure fluctuation frequency on response delay, and the cumulative error of component aging due to usage frequency through cross-correlation algorithms. S6. Based on the coefficient ratio obtained from training, the three factors are fused to generate dynamic calibration coefficients in floating-point format. The dynamic calibration coefficients are multiplied by the pressure fluctuation frequency data to generate a temporary calibration value. Finally, the calibrated pressure value is added to the original pressure measurement value to perform dynamic compensation.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes a viscosity sensor unit to accurately acquire real-time viscosity data of the measured medium, a pressure sensor unit to extract pressure fluctuation frequency and filter out noise, and a counter unit to integrate instrument start-up counts and runtime to generate comprehensive usage frequency data. This comprehensively covers the core dynamic factors affecting measurement accuracy. The dynamic calibration coefficient correction model unit, based on a pre-trained machine learning regression model, analyzes the nonlinear influence of viscosity on fluid resistance, the coupling effect of pressure fluctuation frequency, and the cumulative error due to instrument aging. It quantifies the multi-factor coupling distortion through a cross-correlation algorithm and fuses the results according to the training coefficient ratio to generate dynamic calibration coefficients. These coefficients are then multiplied by the pressure fluctuation frequency and superimposed with the original measured value to achieve dynamic compensation. This completely solves the problem that traditional fixed calibration cannot adapt to multiple dynamic factors, enabling real-time intelligent calibration of pressure measurements. It improves measurement accuracy and stability under different operating conditions, meets the high-precision pressure monitoring needs of industrial production, and provides reliable data support for remote diagnosis and precise control. Attached Figure Description

[0007] Figure 1 This is an overall block diagram of the present invention; Figure 2 This is the overall flowchart of the present invention.

[0008] The meanings of the labels in the diagram are as follows: 1. Viscosity sensor unit; 2. Pressure sensor unit; 3. Counter unit; 4. Dynamic calibration coefficient correction model unit. Detailed Implementation

[0009] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0010] This invention provides a remote intelligent diagnostic and dynamic calibration system for pressure gauges. Please refer to [link / reference]. Figure 1 As shown, it includes: Viscosity sensor unit 1 measures the viscosity of the medium being measured in real time and outputs real-time viscosity data through signal conversion; Pressure sensor unit 2 monitors the pressure signal and extracts and outputs pressure fluctuation frequency data from the pressure signal through the built-in frequency analysis module. Counter unit 3 records the cumulative usage frequency of the instrument, including the number of starts and the running time, and outputs the cumulative usage frequency data; The dynamic calibration coefficient correction model unit 4 calculates dynamic calibration coefficients based on real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data, and applies these dynamic calibration coefficients to correct the pressure measurement value. The dynamic calibration coefficient correction model unit 4 includes an embedded processor and a pre-stored dynamic calibration model. The dynamic calibration model is trained using historical experimental data. The embedded processor takes real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data as input to the dynamic calibration model. The dynamic calibration model is a regression algorithm based on machine learning. By analyzing the influence of viscosity changes on fluid resistance, the coupling effect of pressure fluctuation frequency on sensor response stability, and the cumulative error of instrument aging due to cumulative usage frequency, it internally performs multi-parameter weight adjustment calculations and outputs dynamic calibration coefficients. These coefficients are then multiplied by the pressure fluctuation frequency to generate the calibrated pressure value. This dynamically compensates for the real-time viscosity of the measured medium, the pressure fluctuation frequency, and the cumulative usage frequency of the instrument, thus achieving calibration.

[0011] The viscosity sensor unit 1 uses a vibrating probe to contact the medium being measured. The vibrating probe generates a mechanical vibration signal based on the change in the viscosity resistance of the medium. The mechanical vibration signal is converted into an electrical signal by a built-in piezoelectric converter, and the electrical signal is digitized by an analog-to-digital converter to remove environmental noise and output as continuous real-time viscosity data, which is then transmitted to the dynamic calibration coefficient correction model unit 4.

[0012] The pressure signal of pressure sensor unit 2 is acquired by strain gauge and amplified by amplifier to obtain the amplified pressure signal. The frequency analysis module uses fast Fourier transform algorithm to analyze the amplified pressure signal, identify the dominant fluctuation frequency component in the spectrum of the amplified pressure signal, and extract the dominant fluctuation frequency component as pressure fluctuation frequency data. The pressure fluctuation frequency data is output in numerical form. At the same time, the frequency analysis module eliminates the influence of high frequency noise through built-in filter.

[0013] Counter unit 3 records the instrument's runtime and number of starts using a hardware timer and an event counter, respectively. The runtime starts accumulating from each start, and the number of starts increments with each activation. The accumulated usage frequency data is output through a digital interface. The accumulated usage frequency data integrates the number of starts and runtime into a comprehensive numerical index, which is directly associated with pressure fluctuation frequency data and real-time viscosity data, reflecting the instrument's aging status.

[0014] The dynamic calibration model uses historical experimental datasets during the training phase. These datasets contain real pressure values ​​and error data collected under different temperature, viscosity, pressure fluctuation frequency, and instrument usage frequency conditions. The training process employs a machine learning regression algorithm to iteratively optimize the parameter weights of the dynamic calibration model, enabling the model to output error correction values. After training, the dynamic calibration model is solidified into embedded code and stored in the memory of the dynamic calibration coefficient correction model unit 4.

[0015] The embedded processor receives real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data via a data bus. Before input, the data is normalized to unify the dimensions. The data is then input into the dynamic calibration model. After receiving the input, the dynamic calibration model performs forward propagation calculations and uses parameter weights for data processing to ensure real-time fusion and synchronization of the input data.

[0016] In its internal processing, the dynamic calibration model uses the nonlinear effect of real-time viscosity data on fluid resistance as the first weighting factor, the coupling effect of pressure fluctuation frequency data on sensor response delay as the second weighting factor, and the cumulative error of component aging as the cumulative usage frequency data as the third weighting factor. The first, second, and third weighting factors are used to calculate the degree of mutual influence through the cross-correlation algorithm built into the dynamic calibration model to quantify the multi-factor coupling distortion.

[0017] The first weighting factor, the second weighting factor, and the third weighting factor are fused according to the coefficient ratio obtained during training to generate a dynamic calibration coefficient. The dynamic calibration coefficient represents the error correction value in the current state. When output, it is temporarily stored as a floating-point number in the register of the embedded processor for subsequent calibration operations.

[0018] The embedded processor acquires the dynamic calibration coefficient and reads the pressure fluctuation frequency data. It performs a multiplication operation to multiply the dynamic calibration coefficient by the pressure fluctuation frequency to generate a temporary calibration value. Then, it adds the temporary calibration value to the original pressure measurement value and outputs the calibrated pressure value.

[0019] Further explanation is needed regarding the viscosity sensor unit 1, which uses a vibrating probe to directly contact the measured medium. As the core sensing component that directly interacts with the medium, the vibrating probe incorporates a miniature piezoelectric ceramic excitation source, capable of generating a fixed-frequency mechanical vibration under the action of the drive circuit. The measured medium, i.e., the fluid medium monitored by the pressure gauge, exhibits viscous resistance that changes with its viscosity. The specific process of generating the mechanical vibration signal is as follows: The vibrating probe is fixedly installed in a detection cavity connected to the measured medium. The vibrating end of the probe is completely immersed in the medium. When the piezoelectric ceramic excitation source drives the probe to generate basic vibration, the viscous resistance of the medium will hinder the probe vibration. The higher the viscosity of the medium, the greater the viscous resistance, and the corresponding decrease in the vibration amplitude and slight shift in the vibration frequency of the probe. The lower the viscosity of the medium, the smaller the viscous resistance, the greater the vibration amplitude and the closer the frequency of the probe is to the basic value. This change in vibration parameters caused by the change in viscous resistance forms a mechanical vibration signal reflecting the viscosity of the medium. The dynamic change of the vibration signal is fixedly correlated with the viscosity of the medium. The mechanical vibration signal is converted into an electrical signal by a built-in piezoelectric converter. The piezoelectric converter is an energy conversion component based on the piezoelectric effect (some crystal materials generate charges under mechanical stress, and mechanical energy is converted into electrical energy). It is rigidly connected to the vibrating end of the vibrating probe and can capture the mechanical vibration of the probe in real time. The specific conversion process is as follows: When the probe generates mechanical vibrations containing viscosity information, the resulting mechanical stress is transmitted to the piezoelectric crystal inside the piezoelectric transducer, causing a slight deformation of the crystal and generating a weak charge at its ends corresponding to the vibration parameters. The transducer's built-in charge amplification circuit converts this weak charge into a measurable voltage signal. The amplitude of this voltage signal is proportional to the probe's vibration amplitude, and its frequency matches the probe's vibration frequency. This converts the mechanical vibration signal into an easily processed electrical signal, with the conversion response time controlled within 1 millisecond to ensure real-time signal transmission. Subsequently, the electrical signal is digitized by an analog-to-digital converter (ADC). The ADC is the core component that converts continuously changing analog electrical signals (voltage signals) into discrete digital signals. Here, a 16-bit high-precision ADC is selected, with a sampling frequency set to 100kHz to ensure complete retention of the viscosity information in the electrical signal. The specific digitization process is as follows: An analog-to-digital converter periodically samples the continuous voltage signal output by the piezoelectric converter at a fixed frequency of 100kHz, recording an instantaneous voltage value at each sample. Each sampled voltage value is mapped to a numerical range (0-65535) corresponding to a 16-bit digital code, completing quantization. The quantized value is then converted into a binary digital signal through encoding logic, forming a discrete digital data stream. This data stream completely records the amplitude and frequency variation characteristics of the original electrical signal, providing a digital basis for subsequent viscosity calculations. After digitization, environmental noise needs to be removed, and the final output is continuous real-time viscosity data. Environmental noise refers to irrelevant interference signals mixed in during signal acquisition and conversion, including thermal noise from the circuit itself and noise generated by external electromagnetic interference. This noise can cause data distortion and needs to be eliminated through filtering. Real-time viscosity data refers to continuous values ​​that reflect the current viscosity state of the measured medium, output at a frequency of 10Hz to ensure dynamic tracking of medium viscosity changes. The specific noise reduction and output process is as follows: The digitized digital data stream is input into the sensor's built-in digital low-pass filter (cutoff frequency calibrated to 1kHz) to filter out high-frequency noise signals above 1kHz, retaining the effective signals related to probe vibration. For the filtered data stream, based on a preset vibration parameter-viscosity calibration curve (this curve is calibrated experimentally using standard liquids of different viscosities and stored in the sensor's built-in memory), the vibration amplitude and frequency changes corresponding to the digital signal are converted into specific viscosity values. Viscosity values ​​are continuously output at a frequency of 10Hz, forming continuous real-time viscosity data. Each data point is accompanied by a timestamp to ensure timing alignment with subsequent data. Finally, the real-time viscosity data is transmitted to the dynamic calibration coefficient correction model unit 4. The dynamic calibration coefficient correction model unit 4 is the core unit in the system responsible for integrating multiple parameters and calculating dynamic calibration coefficients. It has a built-in embedded processor and a pre-stored dynamic calibration model and needs to receive real-time data. Viscosity data is used for subsequent calibration calculations. Transmission is achieved using the industrially common SPI digital interface, which features high transmission speed and strong anti-interference capabilities. The specific transmission process is as follows: Viscosity sensor unit 1 establishes a communication connection with the embedded processor of dynamic calibration coefficient correction model unit 4 via the SPI interface, with the communication baud rate set to 1 Mbps. Real-time viscosity data with timestamps is packaged in a standardized format (including sensor identifier, data check bits, timestamp, and viscosity value) and transmitted frame by frame via the data bus. The embedded processor at the receiving end verifies the transmitted data (check bit comparison) to ensure error-free data transmission. If verification fails, a retransmission is requested. After successful verification, the real-time viscosity data is stored in the unit's built-in buffer and marked as valid for subsequent use by the dynamic calibration model. The entire transmission process ensures data integrity and synchronization, laying the foundation for multi-parameter fusion calculations.

[0020] While the viscosity sensor unit outputs real-time viscosity data, the pressure sensor unit 2 needs to simultaneously complete the acquisition, amplification, frequency analysis, and noise reduction of the pressure signal, extracting the fluctuation frequency data that reflects the dynamic characteristics of the pressure, providing key input for the calculation of the dynamic calibration coefficient. The specific implementation process is as follows: The pressure signal from pressure sensor unit 2 is first acquired by a strain gauge. A strain gauge is a core sensing component based on the strain resistance effect, typically a metal foil strain gauge bonded to the elastic body of the sensor. When the pressure of the measured medium acts on the elastic body, the elastic body undergoes slight deformation, causing the strain gauge to deform synchronously. This results in a linear change in the resistance value of the strain gauge with the degree of deformation, thus converting the physical quantity of pressure into a measurable resistance change signal. Since the resistance change signal output by the strain gauge is extremely weak (typically in the millivolt range) and easily affected by environmental interference, it needs to be amplified. A high-precision instrumentation amplifier is selected here, which features high input impedance, low noise, and high common-mode rejection ratio, effectively amplifying the useful signal and suppressing interference. The specific amplification process is as follows: The resistance change signal of the strain gauge is transmitted to the input of the amplifier through a shielded wire. The amplifier linearly amplifies the weak signal at a preset amplification factor (calibrated to 1000 times), boosting the millivolt-level signal to a volt-level signal to ensure the accuracy of subsequent signal analysis. During the amplification process, the amplifier's built-in zero-adjustment circuit cancels out zero-point drift in real time, avoiding signal distortion caused by zero-position deviation. Finally, amplified pressure signals with stable amplitude and improved signal-to-noise ratio are obtained. This signal is a continuously changing analog signal, which completely preserves the fluctuation characteristics of the original pressure. Subsequently, the frequency analysis module calls the Fast Fourier Transform algorithm to analyze the amplified pressure signal. The frequency analysis module is a signal processing module built into the pressure sensor unit 2. Its core is an embedded microprocessor, which is specifically responsible for converting time-domain signals into frequency-domain signals to extract frequency features. The Fast Fourier Transform algorithm is an efficient implementation of the Discrete Fourier Transform, which can quickly convert continuous time-domain pressure signals (amplitude curves that change with time) into frequency-domain spectrograms, greatly reducing computational complexity and adapting to real-time analysis requirements. The specific analysis process is as follows: The frequency analysis module performs discrete sampling on the amplified analog pressure signal at a preset sampling frequency (calibrated to 1kHz), converting the continuous signal into a discrete digital signal sequence. Then, the digital signal sequence is divided into multiple data windows of fixed length. A Hanning window function is applied to the data in each window to suppress spectral leakage. Finally, a fast Fourier transform algorithm is used to convert the time-domain data of each window into the corresponding frequency-domain data, resulting in a spectrum containing different frequency components and their corresponding amplitudes. The horizontal axis of the spectrum represents frequency, and the vertical axis represents the amplitude of the corresponding frequency component, visually presenting the frequency distribution of the pressure signal. Based on the generated spectrum, the dominant fluctuation frequency component is further identified. The dominant fluctuation frequency component refers to the frequency component with the largest amplitude in the spectrum. Its corresponding frequency reflects the main periodic characteristics of pressure fluctuations, distinguishing it from other secondary frequency components and noise. It is a core indicator characterizing dynamic changes in pressure. The specific identification process is as follows: The frequency analysis module iterates through the amplitude data of all frequency components in the spectrum, using a peak detection algorithm to filter out peaks with amplitudes greater than those of surrounding frequency components. If multiple peaks exist, the frequency corresponding to the peak with the largest amplitude is selected as the dominant fluctuation frequency component. For example, the amplitude of 5Hz in the spectrum is 3.2V, significantly higher than the amplitudes of other frequency components; therefore, 5Hz is the dominant fluctuation frequency component of the pressure signal. This process automatically excludes minor frequencies with excessively small amplitudes, ensuring that the extracted frequency components accurately reflect the main fluctuation patterns of the pressure. The identified dominant fluctuation frequency component is extracted as pressure fluctuation frequency data. This pressure fluctuation frequency data is numerical data that quantifies the dominant pressure fluctuation characteristics and is directly related to the response stability of the pressure sensor. Its numerical changes reflect the dynamic fluctuation intensity of the measured medium pressure. The specific extraction process is as follows: The frequency analysis module records the frequency value corresponding to the dominant fluctuation frequency component and verifies the reliability of the amplitude of the frequency component (if the amplitude is lower than the preset threshold, it is determined to be invalid and the spectrum analysis is performed again). After the verification is passed, the frequency value is determined as the pressure fluctuation frequency data. At the same time, the information of other secondary frequency components in the spectrum is removed, and only the core dominant frequency data is retained to ensure the simplicity and relevance of the data. The pressure fluctuation frequency data is output in numerical form. Numerical form means that the extracted frequency value is converted into a standardized digital signal, rather than a graphic or analog signal, so that the embedded processor of the subsequent dynamic calibration coefficient correction model unit 4 can directly read and calculate it. The specific output process is as follows: The frequency analysis module packages the pressure fluctuation frequency data according to a preset communication protocol, adds data check bits and sensor identifiers, and transmits it to the module's output buffer via a digital interface. The buffer updates and outputs data in real time at a frequency of 10Hz, ensuring that the output data is synchronized with the dynamic changes of the pressure signal, providing a time-consistent input for multi-parameter fusion calculation. Simultaneously with frequency analysis and data output, the frequency analysis module eliminates the influence of high-frequency noise through a built-in filter. This built-in filter is a digital low-pass filter, whose core function is to filter out high-frequency noise components in the spectrum that are higher than the set cutoff frequency. These high-frequency noises mostly come from irrelevant factors such as circuit thermal noise and external electromagnetic interference, which can cause frequency analysis distortion. The specific noise reduction process is as follows: The preset low-pass filter has a cutoff frequency of 100Hz (experimentally calibrated to cover the effective frequency range of pressure fluctuations while filtering high-frequency noise). The filter filters the frequency domain data after the fast Fourier transform, retaining the effective frequency components below 100Hz (including the dominant fluctuation frequency components) and setting the amplitude of high-frequency components above 100Hz to 0. Then, the processed frequency domain data is converted back to the time domain signal through the inverse fast Fourier transform for secondary verification in subsequent frequency analysis. At the same time, the filter's built-in adaptive noise suppression algorithm monitors the noise intensity in real time and dynamically adjusts the filtering parameters to ensure effective noise suppression under different interference environments. The final output pressure fluctuation frequency data is not affected by high-frequency noise, has high reliability and stability, and can be directly transmitted to the dynamic calibration coefficient correction model unit 4 to participate in the calculation of dynamic calibration coefficients.

[0021] While the viscosity sensor unit outputs real-time viscosity data and the pressure sensor unit outputs pressure fluctuation frequency data, the counter unit 3 needs to accurately record the cumulative usage frequency of the instruments, capture the core characteristics related to instrument aging, and provide key input for the dynamic calibration model to analyze the cumulative error caused by component aging. The specific implementation process is as follows: Counter unit 3 uses a built-in hardware timer and event counter to specifically record the instrument's runtime and startup count, respectively. The hardware timer is a physical timing component based on a high-precision crystal oscillator clock, with timing accuracy down to the millisecond level. It is unaffected by software operation and can stably output continuous cumulative time values. The event counter is a counting component specifically designed to respond to instrument startup events, triggering counting only when a valid startup signal is detected to avoid numerical distortion caused by false triggers. The instrument's runtime refers to the cumulative time the instrument remains in working condition from successful startup to normal shutdown, with the unit being hours, directly reflecting the instrument's actual workload. The startup count refers to the cumulative number of times the instrument switches from power-off or standby state to normal working state. Each complete startup (power on and system initialization completed) is counted as one, even if the instrument is shut down for a short period after startup. The specific recording process is as follows: The hardware timer starts counting immediately after the instrument's startup signal (stable power-on, system initialization complete) is triggered. It accumulates timing pulses at the crystal oscillator clock frequency (calibrated to 1MHz), automatically updating the runtime value every 3,600,000,000 pulses (corresponding to 1 hour), and adding it to the historical accumulated duration in real time. If the instrument loses power midway, the timer stops counting but retains the current accumulated value, continuing to accumulate from that value upon restarting. The event counter automatically increments its count each time it detects an instrument startup trigger signal (stable for more than 1 second, excluding false triggers caused by power fluctuations). The count starts from 0 and increments by 1 after each startup. The results are stored in real-time in non-volatile memory to prevent data loss during power outages. The runtime is accumulated starting from each startup, with the startup count incrementing with each activation. Accumulated runtime means the runtime record is not reset when the instrument is shut down; instead, it continuously adds to the historical accumulated runtime, ensuring a complete record of the instrument's entire lifecycle operating time. For example, if the last run was 2.5 hours and this startup runs for 3 hours, the accumulated runtime is updated to 5.5 hours. The activation increment count means the startup count is triggered only when a startup event is validly activated. The activation conditions are that the instrument power is on and the system has completed initialization, preventing invalid startups caused by brief power fluctuations or misoperation from being counted. The specific process is as follows: Each time the instrument starts, the system first verifies the validity of the startup signal (checking power supply voltage stability and initialization completion flag). Upon successful verification, the event counter increments by 1, and the hardware timer continues counting from the currently stored cumulative duration. If an unexpected power outage occurs during instrument operation, the hardware timer immediately stops counting, and the current cumulative duration and startup count are stored in non-volatile memory. Upon restarting, the event counter increments again, and the hardware timer continues accumulating based on the stored cumulative duration, ensuring continuous and accurate recording of runtime and startup counts without omissions or duplicates. Subsequently, the cumulative usage frequency data is output through a digital interface. This digital interface is the standardized interface for communication between counter unit 3 and dynamic calibration coefficient correction model unit 4. An I2C interface is selected here, featuring stable transmission rate, simple wiring, and strong anti-interference capabilities, making it suitable for short-distance data transmission in industrial environments. The cumulative usage frequency data is structured data containing startup counts, runtime, and a subsequently integrated comprehensive numerical index, used to transmit instrument usage intensity information to the dynamic calibration model. The specific output process is as follows: Counter unit 3 packages the currently recorded start count (16-bit integer format), runtime (32-bit floating-point format), and comprehensive numerical index (32-bit floating-point format) into a standardized data frame according to a preset output frequency (calibrated to 1Hz). The data frame includes a start bit, data bits, parity bit, and stop bit, where the parity bit is used to verify the integrity of data transmission. The data frame is transmitted to the input buffer of the dynamic calibration coefficient correction model unit 4 via the I2C digital interface. If the receiving end detects a data error during transmission, it will send a retransmission signal, and counter unit 3 will immediately retransmit the data until the receiving end confirms the error. To ensure the accuracy and reliability of the output data, the cumulative usage frequency data is integrated with the number of startups and runtime using a preset algorithm to generate a comprehensive numerical index. Integration involves converting the two parameters—startup count (discrete counting parameter) and runtime (continuous time parameter)—which have different dimensions and types, into a single dimensionless value through weighted calculation, thus eliminating analytical interference caused by dimensional differences. The comprehensive numerical index is the core quantitative indicator after integration, with a value range of 0-100. A higher value indicates a higher cumulative usage intensity of the instrument, indirectly reflecting a more severe degree of aging. The specific integration process is as follows: The counter unit 3 has a built-in integrated algorithm that first normalizes the number of starts and the runtime to the range of 0-50 (for example, if the maximum calibration value for the number of starts is 1000 times and the actual number of starts is 500, it will be normalized to 25; if the maximum calibration value for the runtime is 10000 hours and the actual runtime is 5000 hours, it will be normalized to 25). Then, it calculates the weighted sum of the two according to preset weights (0.4 for the number of starts and 0.6 for the runtime) to obtain a comprehensive numerical index (e.g., 25×0.4+25×0.6=25). This index value is updated in real time as the number of starts increases and the runtime accumulates, ensuring that the cumulative usage intensity of the instrument can be dynamically and accurately quantified. Through a timestamp matching mechanism, the comprehensive numerical index is directly linked to pressure fluctuation frequency data and real-time viscosity data. Direct linking means binding the comprehensive numerical index, pressure fluctuation frequency data, and real-time viscosity data at the same timestamp to form a structured data record containing multi-dimensional parameters. This ensures a one-to-one correspondence of multiple parameters in time series, providing synchronous input for the multi-factor coupling analysis of the dynamic calibration model. The timestamp matching mechanism means that all data is accompanied by a millisecond-level timestamp. Parameter binding is achieved through timestamp consistency verification, avoiding analysis errors caused by time series misalignment. The specific linking process is as follows: The comprehensive numerical index output by counter unit 3 includes a millisecond-level timestamp consistent with the pressure fluctuation frequency data and real-time viscosity data. After receiving the three types of data, the dynamic calibration coefficient correction model unit 4 compares the timestamps and binds the comprehensive numerical index, pressure fluctuation frequency, and real-time viscosity with completely consistent timestamps into a four-tuple data record (timestamp-comprehensive numerical index-pressure fluctuation frequency-real-time viscosity), storing it in an indexed database. When the model calls data later, it can quickly extract the multi-parameter data of the corresponding group through the timestamp, ensuring the temporal synchronization of multi-factor analysis. This comprehensive numerical index indirectly reflects the aging state of the instrument by quantifying the cumulative usage intensity of the instrument. The aging state of the instrument refers to the state of performance degradation of the internal core components due to wear, fatigue, oxidation, and other factors after long-term use. The higher the degree of aging, the greater the performance deviation of the components, and the more significant the measurement error may be. This reflects a positive correlation between the comprehensive numerical index and the degree of instrument aging. Based on the correspondence between the comprehensive numerical index and aging error in historical experimental data, the impact of aging on measurement accuracy can be indirectly quantified through the index value. The specific process is as follows: As the number of instrument startups increases and the runtime accumulates, the performance of internal sensing elements (such as the elastic modulus of strain gauges) and circuit components (such as the amplification factor of amplifiers) gradually degrades, leading to a gradual accumulation of systematic errors in pressure and viscosity measurements. The comprehensive numerical index, by integrating startup counts and runtime, accurately quantifies this aging accumulation effect. For example, a comprehensive numerical index of 80 corresponds to approximately 0.5% of the aging error in historical experimental data. After the dynamic calibration model calls this index, it can combine it with the preset index-aging error mapping relationship to accurately analyze the cumulative error caused by aging in the current instrument. This provides a key basis for the calculation of the aging dimension in subsequent dynamic calibration coefficients, ensuring that the calibration process can specifically compensate for measurement deviations caused by aging.

[0022] To enable the dynamic calibration coefficient correction model unit 4 to accurately calculate the dynamic calibration coefficient and effectively compensate for interference from multiple factors, the dynamic calibration model needs to be trained first. The core of the training is to rely on historical experimental datasets covering multiple operating conditions, and to optimize the parameter weights through algorithmic iteration to ultimately form a deployable embedded model. The specific implementation process is as follows: The dynamic calibration model primarily relies on historical experimental datasets during the training phase. These datasets are structured data sets collected and organized by the system under laboratory conditions, simulating various typical operating conditions of pressure gauges in actual operation. The core of these datasets includes real pressure values ​​and error data under different temperature, viscosity, pressure fluctuation frequencies, and instrument usage frequencies. Different operating conditions refer to temperature coverage of -20℃ to 80℃ (common industrial temperature range), viscosity coverage of 1-1000 mPa·s (common viscosity range of measured media), pressure fluctuation frequency coverage of 1-100 Hz (common range of dynamic pressure changes), and instrument usage frequency. The instrument's usage frequency covers 0-10,000 startups / 0-10,000 hours of operation (instrument lifecycle usage intensity). 100 sets of data are repeatedly collected for each operating condition combination to ensure data reliability. The true pressure value is the standard pressure value measured by a high-precision standard pressure gauge (accuracy class 0.01, far higher than the pressure gauge being measured) under the same operating conditions, serving as the benchmark for measuring error. The error data is the difference between the original measured value and the true pressure value of the pressure gauge under the same operating conditions (error = measured value - true value), directly reflecting the impact of different operating conditions on measurement accuracy. The specific dataset construction process is as follows: A working condition simulation platform was built in the laboratory. The ambient temperature was adjusted by a temperature control system, the viscosity of the measured medium was changed by a medium viscosity adjustment device, and a pressure fluctuation generator simulated pressure fluctuations at different frequencies. At the same time, the number of times the pressure gauge was started and the running time were controlled by a mechanical device. Under each working condition combination, the measured value of the measured pressure gauge and the actual pressure value of the standard pressure gauge were collected synchronously. After calculating the error data, the data was organized into structured data in the format of temperature-viscosity-pressure fluctuation frequency-instrument usage frequency-actual pressure value-error data. Finally, a historical experimental dataset containing more than 100,000 sets of valid data was formed. The dataset was divided into a training set (for model training) and a validation set (for verifying model accuracy) in a 7:3 ratio. The training process employs a machine learning regression algorithm, which iteratively optimizes and dynamically calibrates the model's parameter weights. Here, a gradient descent-optimized linear regression algorithm is selected to meet the error correction value prediction requirements of multi-input single-output models. This algorithm can efficiently establish a nonlinear mapping relationship between input parameters (temperature, viscosity, etc.) and output (error correction value). The parameter weights are quantified coefficients in the model that characterize the degree of influence of each input parameter on the error correction value (such as the weight corresponding to viscosity, the weight corresponding to pressure fluctuation frequency, etc.). The initial values ​​are randomly set according to a normal distribution and are continuously adjusted through iteration to make the model prediction more accurate. The specific optimization process is as follows: First, the input parameters (temperature, viscosity, pressure fluctuation frequency, and instrument usage frequency) in the training set data are normalized to unify the dimensions to the 0-1 range, avoiding imbalances in weight optimization due to differences in numerical ranges. Then, the normalized input data is fed into the dynamic calibration model, which calculates the predicted error correction value based on the current parameter weights. The predicted value is compared with the corresponding true error data in the training set to calculate the mean square error (reflecting the overall deviation between the predicted and true values). Error information is backpropagated using the gradient descent algorithm, and the weights of each input parameter are adjusted according to a preset learning rate (calibrated to 0.001). Parameters with a significant impact on the error have their weights increased, while parameters with a smaller impact have their weights decreased. This process is repeated for the input data... The process follows a sequence of prediction, error calculation, and weight adjustment. Each iteration uses the full training set data until the mean squared error (MSE) changes over 100 consecutive iterations is less than a preset threshold (e.g., 0.0001), and the MSE of the validation set is less than 1.2 times that of the training set (to avoid overfitting). At this point, the parameter weight optimization is considered complete, and the model possesses stable predictive capabilities. Through this iterative optimization, the dynamic calibration model can output an error correction value. This error correction value is a value predicted by the model based on the current input operating parameters (real-time viscosity, pressure fluctuation frequency, cumulative usage frequency) to offset the pressure gauge measurement error. Its magnitude and direction accurately match the actual error under the current operating conditions. This value can then be used to correct the original measurement value, improving measurement accuracy. The specific output process is as follows: After receiving the normalized input parameters, the model processes them using weighted summation (multiplying each parameter by its corresponding weight and then summing the results) and a nonlinear activation function to generate a single error correction value. The unit of this correction value is consistent with the pressure value (e.g., MPa). If the predicted error is positive (the measured value is too large), the correction value is negative; otherwise, it is positive, ensuring that it can directly offset the original measurement error. During training, the model continuously optimizes the output logic, gradually reducing the deviation between the correction value and the true error, ultimately achieving an accuracy requirement where the average absolute error between the predicted correction value and the true error is less than 0.01 MPa, thus meeting the needs of dynamic calibration. After training, the dynamic calibration model needs to be embedded into embedded code and stored in the memory of the dynamic calibration coefficient correction model unit 4. Embedding into embedded code means converting the trained model parameters, weights, input / output processing logic, normalization algorithms, etc., into low-level language code that the embedded processor can directly execute. This eliminates redundant computational logic from the training process, retains the core inference flow, and ensures the model runs efficiently in the embedded environment. The memory of the dynamic calibration coefficient correction model unit 4 is a built-in non-volatile memory with the characteristic of not losing data when power is off. It can store embedded code for a long time, ensuring that the model can be directly called after the instrument is powered on. The specific embedding process is as follows: First, the trained parameter weights (such as a weight of 0.35 for viscosity, 0.4 for pressure fluctuation frequency, and 0.25 for cumulative usage frequency) are written into the code as an array, clearly defining the index and corresponding weight of each parameter. Then, the core processes, including normalization logic, weighted summation calculation, and error correction value output, are written as independent functions in embedded C language. The function interfaces are adapted to the instruction set of the embedded processor (such as the ARM Cortex-M series) of the dynamic calibration coefficient correction model unit 4. The written code is compiled into machine code using a compiler, removing comments, redundant variables, and other irrelevant content, compressing the code size, and finally writing the machine code into the Flash memory of the dynamic calibration coefficient correction model unit 4 using a JTAG programmer. Simultaneously, memory read permissions are set to ensure the model code is not tampered with. After solidification, the instrument is powered on for testing to verify that the embedded model can normally read parameters, perform calculations, and output error correction values, ensuring that the model is successfully transferred from the training environment to the hardware unit, providing core support for the subsequent real-time calculation of dynamic calibration coefficients.

[0023] After acquiring the core data output from the viscosity sensor unit, pressure sensor unit, and counter unit, the embedded processor of the dynamic calibration coefficient correction model unit 4 needs to first receive and standardize the data before inputting it into the pre-trained dynamic calibration model for calculation, ensuring efficient fusion and synchronous analysis of multiple parameters. The specific implementation process is as follows: The embedded processor receives real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data via a data bus. The embedded processor is the core computing component of the dynamic calibration coefficient correction model unit 4, employing a low-power ARM Cortex-M series processor. It possesses fast data processing and interface communication capabilities, adapting to the real-time requirements of industrial scenarios. The data bus is a parallel communication line connecting each sensor unit to the embedded processor. Here, an SPI data bus is used, with a transmission rate calibrated to 1Mbps, capable of simultaneously receiving multiple data streams and exhibiting strong anti-interference capabilities. The real-time viscosity data is a continuous value output by the viscosity sensor unit, the pressure fluctuation frequency data is the dominant fluctuation frequency extracted by the pressure sensor unit, and the cumulative usage frequency data is a comprehensive numerical index generated by the counter unit. All three types of data are accompanied by millisecond-level timestamps to ensure timing consistency. The specific receiving process is as follows: The embedded processor selects each sensor unit via the chip select signal of the SPI interface and sequentially reads real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data according to a preset communication protocol. Each data stream includes a data header, valid values, a check bit, and a data tail. After receiving the data, the processor first verifies its integrity using the check bit. If the verification fails, it triggers a reread. After successful verification, the data is temporarily stored in the processor's built-in cache, and a corresponding timestamp is recorded to ensure that the three types of data are received without loss or distortion. Before inputting the data into the model, it needs to be normalized to unify the units. Normalization is a standardization operation that transforms raw data with different units and numerical ranges into the 0-1 range through linear mapping. Its core purpose is to eliminate the imbalance of model parameter weights caused by differences in data range. For example, real-time viscosity data may be thousands of mPa·s, while pressure fluctuation frequency data is only tens of Hz. Direct input will cause the model to be overly biased towards parameters with larger values. Unifying the units means that after normalization, all input data are in the same numerical range, and there is no need to consider the differences in the original units. They only participate in the model calculation based on their relative size. The specific processing is as follows: The embedded processor retrieves historical extreme values ​​of three types of data pre-stored in memory: real-time viscosity data (historical maximum value 1000 mPa·s, minimum value 1 mPa·s); pressure fluctuation frequency data (historical maximum value 100 Hz, minimum value 1 Hz); and cumulative usage frequency data (historical maximum value 100, minimum value 0). For each type of data, a normalization calculation is performed: the current data value is subtracted from the historical minimum value of that type, and then divided by the difference between the historical maximum and minimum values ​​to obtain the normalized value. For example, if the current real-time viscosity data is 500.5 mPa·s, the normalized value is (500.5 - 1) ÷ (1000 - 1) ≈ 0.5; the pressure fluctuation frequency data is 50 Hz, and the normalized value is (50 - 1) ÷ (100 - 1) ≈ 0.495; the cumulative usage frequency data is... 50, after normalization, becomes (50-0)÷(100-0)=0.5. All three types of data are unified to the 0-1 range to avoid interference from the difference in units on the model calculation. The normalized three types of data are input into the dynamic calibration model. After receiving the input, the model performs forward propagation calculation. The dynamic calibration model is a machine learning model that has been trained and optimized on historical experimental datasets and then solidified into embedded code. It has a simplified architecture with built-in input layer, hidden layer and output layer, and is specifically used to establish the mapping relationship between multiple input parameters and dynamic calibration coefficients. Forward propagation calculation refers to the process in which data enters the model from the input layer and is passed along the unidirectional path of input layer → hidden layer → output layer. Through the weighted operation and activation function processing of each layer, the output result is finally generated. It is different from the backpropagation in the training stage and only focuses on inference calculation without parameter update operation. The specific calculation process is as follows: The embedded processor inputs the normalized real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data into the input layer of the model in a fixed order. The neurons in the input layer convert the three types of data into digital signals that the model can process and transmit them synchronously to the hidden layer. The hidden layer contains 16 neurons, each of which performs a weighted summation of the input data (the weights are parameters that are fixed after training), and then processes it through a sigmoid activation function to fit the nonlinear correlation between multiple parameters and generate an intermediate feature vector. The intermediate feature vector is transmitted to the output layer, where the neurons in the output layer perform a linear weighted operation to initially generate intermediate values ​​of the dynamic calibration coefficients. The entire forward propagation process takes less than 1 millisecond, which meets the requirements of real-time calibration. During the calculation, parameter weights are used for data processing. These weights are quantized coefficients obtained during model training and correspond to real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data, respectively. They characterize the degree of influence of each type of data on the dynamic calibration coefficient (e.g., a weight of 0.3 for real-time viscosity, 0.4 for pressure fluctuation frequency, and 0.3 for cumulative usage frequency; specific values ​​are fixed through training). The larger the weight, the more significant the impact of the corresponding data on the output result. The specific processing procedure is as follows: During hidden layer computation, each neuron is loaded with pre-defined weights. The normalized data for the three classes is multiplied by its corresponding weight, and then the neuron's bias term (a training-fixed constant) is added to obtain a weighted sum. For example, the normalized value of real-time viscosity data (0.5 × weight 0.3) + the normalized value of pressure fluctuation frequency data (0.495 × weight 0.4) + the normalized value of cumulative usage frequency data (0.5 × weight 0.3) + the bias term (0.1) ≈ 0.448. This weighted sum is then mapped to the 0-1 interval using a sigmoid activation function to obtain the neuron's output value. The output values ​​of all hidden layer neurons are integrated into an intermediate feature vector, which is then fed into the output layer. After further weighted calculations, dynamic calibration coefficients are finally generated to ensure that the data processing accurately reflects the different degrees of influence of various parameters on error correction. The entire process ensures real-time fusion and synchronization of input data through timestamp synchronization and high-speed computation. Real-time fusion means binding parameters from three different sources to the same timestamp and integrating them into a unified input for the model, avoiding calibration deviations caused by isolated calculations of single parameters. Synchronization means that the entire process of data reception, normalization processing, and model calculation proceeds in millisecond-level time sequences, ensuring that the dynamic calibration coefficients output by the model can match the measurement data under the current working conditions in real time without significant delay. The specific implementation process is as follows: When receiving data, the embedded processor compares timestamps and only includes real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data with completely consistent timestamps in the calculation. If a certain type of data has a mismatched timestamp due to transmission delay, the data is temporarily stored and awaits the next set of synchronized data to avoid timing misalignment. At the same time, normalization processing and forward propagation calculation are both performed by the processor's hardware acceleration module, shortening the computation time and ensuring that the entire process from data reception to model output of dynamic calibration coefficients takes no more than 5 milliseconds. This enables real-time fusion and synchronous processing of input data, providing accurate and timely core basis for the dynamic calibration of subsequent pressure measurements.

[0024] After the embedded processor inputs the normalized real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data into the dynamic calibration model, the model analyzes the independent impact of the three types of parameters on the pressure measurement error, generates corresponding weighting factors, and then quantifies the collaborative interference between multiple parameters through a cross-correlation algorithm, providing an accurate basis for the fusion of dynamic calibration coefficients. The specific implementation process is as follows: In its internal processing, the dynamic calibration model first analyzes the nonlinear impact of the input real-time viscosity data on fluid resistance, quantifying this impact as the first weighting factor. The real-time viscosity data is a normalized value (0-1 range) reflecting the viscosity of the measured medium. Higher viscosity means stronger flow resistance to the pressure gauge's sensing components. Fluid resistance is the resistance force generated when the measured medium flows through the pressure gauge's sensing area; its magnitude directly affects the deformation of the sensing element, leading to pressure measurement deviation. The nonlinear impact refers to the fact that viscosity and fluid resistance are not a simple linear relationship. When viscosity exceeds a certain threshold, the rate of resistance increase changes significantly, leading to complex variations in measurement error. The first weighting factor is a coefficient (0-1 range) quantifying the degree of viscosity's influence on measurement error. A larger value indicates a more significant contribution of the current viscosity state to the error. The specific process is as follows: The model retrieves a pre-stored viscosity-fluid resistance-error 3D mapping table (generated from historical experimental data, recording the variation of fluid resistance and corresponding error contributions at different viscosities). Based on the current normalized real-time viscosity data, the model locates the corresponding fluid resistance variation curve in the mapping table. The slope of the curve indicates the strength of the nonlinear effect; a steeper slope indicates a more pronounced nonlinearity, and the first weighting factor is set higher. For example, if the current viscosity normalization value is 0.6, the corresponding curve slope is relatively large, indicating a significant nonlinear effect, and the first weighting factor can be set to 0.35. This value directly represents the weight of the viscosity factor on the current measurement error. Next, the coupling effect of pressure fluctuation frequency data on sensor response delay is analyzed and quantified as the second weighting factor. The rate data is a normalized value (0-1 range) reflecting the dynamic fluctuation period of pressure. The higher the frequency, the more frequent the pressure changes. The sensor response delay is the time difference between the pressure gauge sensing element sensing a pressure change and outputting a measurement signal. Ideally, the delay should be close to 0, but in reality, it is affected by the frequency of pressure changes. The coupling effect refers to the interaction between the pressure fluctuation frequency and the sensor response delay. That is, frequency changes directly lead to changes in response delay, and delay changes, in turn, affect the real-time performance and accuracy of pressure measurement, forming cooperative interference. The second weighting factor is a coefficient (0-1 range) that quantifies the degree of influence of this coupling effect on measurement error. The larger the value, the more prominent the impact of the response delay caused by the current pressure fluctuation frequency on the error. The specific process is as follows: The model incorporates a frequency-response delay correlation model. Based on the current normalized pressure fluctuation frequency data, it calculates the corresponding sensor response delay. When the frequency is too high, the sensing element cannot keep up with pressure changes in time, increasing the delay. When the frequency is too low, although the delay is shortened, it is more susceptible to environmental noise interference. Then, through the quantitative relationship between delay and error (based on historical experimental calibration), the value of the second weighting factor is determined. For example, if the current frequency normalization value is 0.45, the corresponding response delay is moderate, and the coupling effect is moderate. The second weighting factor can be set to 0.4, accurately reflecting the weight of this factor on the error. Subsequently, the cumulative error caused by component aging was analyzed based on the cumulative usage frequency data, quantified as a third weighting factor. The cumulative usage frequency data is a comprehensive numerical index (0-1 range) integrating the number of instrument starts and running time. The larger the value, the higher the instrument's usage intensity and the more severe the aging of its internal core components. Component aging refers to the phenomenon of performance degradation of pressure gauge sensing elements (such as strain gauges and vibration probes) and circuit components (such as amplifiers and converters) after long-term use due to factors such as wear, fatigue, and oxidation. For example, the elastic modulus of strain gauges decreases, and the charge conversion efficiency of piezoelectric converters decreases. Cumulative error is a systematic deviation formed by the gradual accumulation of measurement errors caused by performance degradation during component aging. The more severe the aging, the greater the cumulative error. The third weighting factor is a coefficient (0-1 range) that quantifies the degree of influence of cumulative error on the current measurement error. The larger the value, the more significant the contribution of instrument aging to the error. The specific process is as follows: The model calls the mapping relationship between usage frequency, aging level, and cumulative error (calibrated through historical full life cycle experiments). Based on the comprehensive numerical index of the current cumulative usage frequency, the aging level of the component is determined. The higher the index value, the higher the aging level and the greater the cumulative error. The magnitude of the cumulative error is then converted into a third weighting factor. For example, if the current comprehensive numerical index is 0.5, corresponding to a medium aging level and a moderate cumulative error, the third weighting factor can be 0.25, clearly representing the weight of the aging factor on the error. After obtaining the first, second, and third weighting factors, the degree of mutual influence among them is calculated using the cross-correlation algorithm built into the dynamic calibration model to quantify multi-factor coupling distortion. The cross-correlation algorithm is a mathematical algorithm used to analyze the degree of linear correlation between two or more variables. Here, it is optimized to adapt to the quantification of nonlinear correlations, accurately calculating the synergistic interference strength between the parameters corresponding to any two weighting factors. The degree of mutual influence refers to the fact that the three types of parameters (viscosity, pressure fluctuation frequency, and usage frequency) do not independently affect the measurement error, but rather have a synergistic effect. For example, under high viscosity conditions, the coupling effect of pressure fluctuation frequency on response delay is amplified; this superimposed effect is called interaction. Multi-factor coupling distortion refers to the phenomenon that the measurement error caused by the synergistic effect of multiple parameters deviates from the sum of errors under the independent influence of a single parameter. This distortion makes the error pattern more complex and must be accurately compensated through quantification. The specific process is as follows: The model inputs the original parameter data (normalized real-time viscosity, pressure fluctuation frequency, and cumulative usage frequency) corresponding to the first, second, and third weighting factors into the cross-correlation algorithm to calculate the correlation coefficient (in the range of -1 to -1) between each pair of parameters. The larger the absolute value of the correlation coefficient, the stronger the mutual influence between the two pairs of parameters. For example, the correlation coefficient between real-time viscosity data and pressure fluctuation frequency data is 0.3, indicating that high viscosity will enhance the coupling effect of pressure fluctuation frequency on response delay. This synergistic effect will make the actual error greater than the sum of the errors of a single factor. The algorithm integrates the three sets of correlation coefficients to generate a coupling distortion quantification value (in the range of 0 to 0.5). This value directly reflects the strength of multi-parameter synergistic interference. For example, a coupling distortion quantification value of 0.15 indicates that there is a moderate degree of coupling distortion. When the weighting factors are subsequently integrated to generate dynamic calibration coefficients, this distortion quantification value needs to be taken into consideration. The error deviation caused by synergistic interference is offset by weighted adjustment to ensure that the dynamic calibration coefficient can fully cover the independent influence of single factors and the coupled influence of multiple factors.

[0025] After quantifying the first, second, and third weighting factors, as well as the quantified values ​​of multi-factor coupling distortion, the dynamic calibration model needs to integrate the independent influence of single factors and the synergistic interference of multiple factors into a unified error correction index. This is achieved by fusing the three weighting factors according to the coefficient ratios obtained during training, ultimately generating dynamic calibration coefficients. The specific implementation process is as follows: First, the three weighting factors are fused according to the coefficient ratios obtained during training. These ratios are fixed parameters determined through iterative optimization during the training phase of the dynamic calibration model on historical experimental datasets. They are used to allocate the contribution weights of the three weighting factors in the fusion process. This ratio is based on the average influence intensity of the three types of factors on pressure measurement error, and is solidified through training to a ratio of 0.3 for the first weighting factor, 0.4 for the second weighting factor, and 0.3 for the third weighting factor, ensuring that the fusion result can preferentially cover the more significant error sources. Fusion refers to the process of combining the quantified values ​​of multi-factor coupling distortion, fine-tuning the three weighting factors, and then calculating a weighted sum according to the ratios to integrate the dispersed single-factor influence indicators into a unified comprehensive correction index. The specific fusion process is as follows: The dynamic calibration model first calls the previously calculated multi-factor coupling distortion quantization value (e.g., 0.15), multiplies each weight factor by (1 + coupling distortion quantization value), and performs collaborative interference compensation on the single-factor weights. For example, the first weight factor is 0.35 × (1 + 0.15) ≈ 0.4025, the second weight factor is 0.4 × (1 + 0.15) ≈ 0.46, and the third weight factor is 0.25 × (1 + 0.15) ≈ 0.2875. Then, according to the coefficient ratio obtained during training, the three fine-tuned weight factors are weighted and summed, i.e., 0.4025 × 0.3 + 0.46 × 0.4 + 0.2875 × 0. 3. Finally, a comprehensive quantitative value is obtained, which is the dynamic calibration coefficient. The dynamic calibration coefficient generated by the above fusion calculation is a unified correction index for pressure measurement error under the current operating conditions. The dynamic calibration coefficient is a single value that integrates the nonlinear influence of real-time viscosity, the coupling effect of pressure fluctuation frequency, the cumulative error of instrument aging, and the synergistic interference of multiple factors. Its magnitude directly corresponds to the error amplitude that needs to be corrected, and its sign reflects the error direction (if the coefficient is positive, it means that the original measured value is too small and needs to be positively supplemented; if it is negative, it means that the original measured value is too large and needs to be negatively offset). It is the core bridge connecting multi-parameter analysis and final pressure calibration. During the generation process, the model constrains the weighted summation values, limiting them to the range of -0.5 to 0.5 (historical experiments have shown that this range covers error correction requirements under most operating conditions), avoiding overcalibration due to extreme values. If the weighted sum exceeds this range, it is treated as a boundary value (e.g., if the result is 0.6, it is taken as 0.5; if it is -0.7, it is taken as -0.5), ensuring the rationality and safety of the dynamic calibration coefficients. The final generated coefficients not only fully preserve the influence of various error sources but also possess the characteristics of simplicity and direct computation. After the dynamic calibration coefficients are generated, they need to be temporarily stored in floating-point format in the embedded processor's registers for subsequent calibration operations. Registers are high-speed temporary storage units built into the processor, with nanosecond-level access speeds. They are specifically used to store critical data to be processed immediately, unlike ordinary memory, which can significantly improve the efficiency of subsequent data reading and processing. Floating-point format is a digital storage format that can accurately represent decimals. Here, 32-bit single-precision floating-point numbers are used, which ensures the decimal precision of the dynamic calibration coefficients (accurate to 6 decimal places) while controlling storage space usage, adapting to the resource limitations of embedded processors. Temporary storage refers to temporarily storing the dynamic calibration coefficients in registers until a complete stress calibration process is completed. The next fusion calculation generates new coefficients, which are then overwritten and updated to ensure the real-time performance and validity of the data. The specific temporary storage process is as follows: The dynamic calibration model transmits the generated dynamic calibration coefficient (e.g., 0.1856) to the embedded processor's arithmetic cache. The processor's built-in format conversion module converts the decimal value into a 32-bit single-precision floating-point binary code. Subsequently, the processor writes the binary code into a preset dedicated control register via the internal data bus. During the writing process, a parity check mechanism is activated to verify the integrity of the data transmission through the parity bit. If the check finds a data error, it is immediately re-converted and rewritten until the check passes. The register locks the floating-point data to prevent interference from other operations and simultaneously sends a data ready signal to the processor, indicating that the coefficient can be called at any time to perform subsequent multiplication operations with the pressure fluctuation frequency data, providing reliable data support for fast and accurate pressure calibration.

[0026] After the dynamic calibration coefficients are temporarily stored in the embedded processor registers in floating-point format, the final pressure value calibration execution stage can begin. By calculating the calibration coefficients with key parameters and then combining them with the original measurement values, dynamic compensation for pressure measurement errors is achieved. The specific implementation process is as follows: The embedded processor first obtains the dynamic calibration coefficients from a dedicated register and simultaneously reads the pressure fluctuation frequency data from the data buffer. It then performs a multiplication operation to combine the two. As the core computing unit of the dynamic calibration coefficient correction model unit 4, the embedded processor has a hardware floating-point unit (FPU) capable of efficiently handling floating-point multiplication and avoiding precision loss caused by software simulation. Obtaining the dynamic calibration coefficients involves the processor accessing the control register of the temporary coefficients via the internal bus, reading the 32-bit single-precision floating-point format dynamic calibration coefficients at nanosecond speeds. These coefficients contain comprehensive correction information for all error sources. The pressure fluctuation frequency data is the dominant fluctuation frequency value output by the pressure sensor unit 2 after frequency analysis and noise reduction. It has been previously timestamped with the real-time viscosity data and cumulative usage frequency data and stored in the processor's input buffer to ensure timing consistency with the dynamic calibration coefficients. The multiplication operation is a floating-point multiplication performed by the processor through the FPU, retaining 6 decimal places of precision during the operation to avoid distortion of the correction amount. The specific process is as follows: The processor first verifies the data readiness flag of the dynamic calibration coefficient in the register. After confirming its validity, it reads the coefficient and simultaneously reads the pressure fluctuation frequency data at the corresponding time from the buffer according to the timestamp index. The two data are then passed to the FPU for multiplication. During the operation, the processor monitors for data overflow. If the result exceeds the preset range, it is truncated at the boundary value to ensure operational safety. The final product is the temporary calibration value. The temporary calibration value generated by the above multiplication operation is an intermediate error correction quantity adapted to the current pressure fluctuation characteristics. The temporary calibration value is the product of the dynamic calibration coefficient and the pressure fluctuation frequency data, with the unit being MPa (pressure unit). Its magnitude and sign directly correspond to the magnitude and direction of the correction to the original pressure measurement value. For example, if the coefficient and frequency are positive, the temporary calibration value is positive, indicating that the original measurement value is too small and needs to be positively supplemented. If the coefficient is negative, the temporary calibration value is negative and needs to be subtracted from the original measurement value. During the specific generation process, after the multiplication operation is completed, the processor stores the result in a temporary arithmetic register and performs a validity check. It compares the temporary calibration value with the correction range under similar historical operating conditions. If the deviation exceeds ±30%, it is judged as abnormal, and the dynamic calibration coefficient and pressure fluctuation frequency data are read again for recalculation. If the check passes, the temporary calibration value is confirmed to be valid, which prepares for the subsequent superposition with the original pressure measurement value and ensures that the correction amount conforms to the actual operating conditions.Subsequently, the processor retrieves the original pressure measurement value, adds the temporary calibration value to the original pressure measurement value, and finally outputs the calibrated pressure value. The original pressure measurement value is the raw pressure data collected by the strain element of pressure sensor unit 2, amplified by the amplifier, and has not undergone any error correction. It directly reflects the pressure physical quantity of the measured medium, but includes systematic errors caused by viscosity, frequency fluctuations, and instrument aging. Addition refers to performing a floating-point addition operation on the temporary calibration value and the original pressure measurement value to cancel or supplement the error. The addition process is also performed by the FPU to ensure that the accuracy is not lost. The calibrated pressure value is an accurate pressure value after eliminating multi-factor systematic errors. It can truly reflect the actual pressure of the measured medium and is the core data that the pressure gauge ultimately displays, transmits, or stores. The specific process is as follows: The processor reads the raw pressure measurement value from the output buffer of pressure sensor unit 2, confirms that the timestamp of the value is completely consistent with the dynamic calibration coefficient and pressure fluctuation frequency data, avoiding calibration deviation caused by timing misalignment. The processor adds the raw pressure measurement value to the temporary calibration value in the temporary arithmetic register. After the calculation is completed, the processor constrains the result to the range of the pressure gauge. If the range is exceeded, the range boundary value is output and an alarm is marked. Finally, the calibrated pressure value is converted into a standardized format and displayed in real time through the instrument's display module. At the same time, it is transmitted to the remote monitoring system via the digital interface and stored in non-volatile memory for traceability. The entire calibration process takes no more than 10 milliseconds, realizing real-time dynamic compensation of pressure measurement and ensuring that accurate pressure data can be output under different operating conditions.

[0027] like Figure 2 As shown, a second objective of this invention is to provide a method for implementing a remote intelligent diagnostic and dynamic calibration system for a pressure gauge, comprising the following steps: S1. The vibrating probe contacts the medium being tested and generates a mechanical vibration signal based on the change in viscous resistance. This signal is then converted into an electrical signal by a piezoelectric converter, and further processed digitally by an analog-to-digital converter to filter out environmental noise and output real-time viscosity data. S2. The pressure signal is collected by the strain gauge and amplified by the amplifier. The spectrum of the amplified signal is analyzed by the Fast Fourier Transform algorithm to identify the dominant fluctuation frequency component. After eliminating high-frequency noise by the filter, the pressure fluctuation frequency data is output. S3. The instrument running time is accumulated using a hardware timer, and the number of startups is recorded by an event counter. The two are integrated into a cumulative usage frequency data in the form of a comprehensive numerical index, stored in non-volatile memory, and output through a digital interface. S4. The embedded processor normalizes the input real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data, and inputs them into the pre-trained dynamic calibration model. S5. The dynamic calibration model analyzes the nonlinear effect of viscosity on fluid resistance, the coupling effect of pressure fluctuation frequency on response delay, and the cumulative error of component aging due to usage frequency through cross-correlation algorithms. S6. Based on the coefficient ratio obtained from training, the three factors are fused to generate dynamic calibration coefficients in floating-point format. The dynamic calibration coefficients are multiplied by the pressure fluctuation frequency data to generate a temporary calibration value. Finally, the calibrated pressure value is added to the original pressure measurement value to perform dynamic compensation.

[0028] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A remote intelligent diagnostic and dynamic calibration system for pressure gauges, characterized in that, include: The viscosity sensor unit (1) measures the viscosity of the medium being measured in real time and outputs real-time viscosity data through signal conversion; The pressure sensor unit (2) monitors the pressure signal and extracts and outputs the pressure fluctuation frequency data from the pressure signal through the built-in frequency analysis module; The counter unit (3) records the cumulative usage frequency of the instrument, including the number of starts and the running time, and outputs the cumulative usage frequency data; The dynamic calibration coefficient correction model unit (4) calculates the dynamic calibration coefficient based on real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data, and applies the dynamic calibration coefficient to correct the pressure measurement value. The dynamic calibration coefficient correction model unit (4) includes an embedded processor and a pre-stored dynamic calibration model. The dynamic calibration model is established by training with historical experimental data. The embedded processor takes real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data as inputs and inputs them into the dynamic calibration model. The dynamic calibration model is a regression algorithm based on machine learning. By analyzing the influence of viscosity change on fluid resistance, the coupling effect of pressure fluctuation frequency on sensor response stability, and the cumulative error of cumulative usage frequency on instrument aging, it performs multi-parameter weight adjustment calculation internally and outputs the dynamic calibration coefficient. Then, the dynamic calibration coefficient is multiplied by the pressure fluctuation frequency to generate the calibrated pressure value. The real-time viscosity of the measured medium, the pressure fluctuation frequency, and the cumulative usage frequency of the instrument are dynamically compensated to achieve calibration.

2. The pressure gauge remote intelligent diagnostic and dynamic calibration system according to claim 1, characterized in that: The viscosity sensor unit (1) uses a vibrating probe to contact the medium being measured. The vibrating probe generates a mechanical vibration signal based on the change in the viscosity resistance of the medium. The mechanical vibration signal is converted into an electrical signal by a built-in piezoelectric converter, and the electrical signal is digitized by an analog-to-digital converter. After removing environmental noise, the output is continuous real-time viscosity data, which is then transmitted to the dynamic calibration coefficient correction model unit (4).

3. The pressure gauge remote intelligent diagnostic and dynamic calibration system according to claim 2, characterized in that, The pressure signal of the pressure sensor unit (2) is acquired by the strain element and amplified by the amplifier to obtain the amplified pressure signal. The frequency analysis module uses the fast Fourier transform algorithm to analyze the amplified pressure signal, identify the dominant fluctuation frequency component in the spectrum of the amplified pressure signal, and extract the dominant fluctuation frequency component as pressure fluctuation frequency data. The pressure fluctuation frequency data is output in numerical form. At the same time, the frequency analysis module eliminates the influence of high frequency noise through the built-in filter.

4. The pressure gauge remote intelligent diagnostic and dynamic calibration system according to claim 3, characterized in that: The counter unit (3) records the running time and the number of starts of the instrument through a hardware timer and an event counter, respectively. The running time is accumulated from each start, and the number of starts is incremented each time it is activated. The accumulated usage frequency data is output through a digital interface. The accumulated usage frequency data integrates the number of starts and the running time into a comprehensive numerical index, which is directly associated with the pressure fluctuation frequency data and the real-time viscosity data to reflect the aging status of the instrument.

5. The pressure gauge remote intelligent diagnostic and dynamic calibration system according to claim 4, characterized in that: The dynamic calibration model uses historical experimental datasets during the training phase. The historical experimental datasets contain real pressure values ​​and error data collected under different temperature, viscosity, pressure fluctuation frequency and instrument usage frequency conditions. The training process uses machine learning regression algorithm to iteratively optimize the parameter weights of the dynamic calibration model so that the dynamic calibration model outputs error correction values. After training, the dynamic calibration model is solidified into embedded code and stored in the memory of the dynamic calibration coefficient correction model unit (4).

6. The pressure gauge remote intelligent diagnostic and dynamic calibration system according to claim 5, characterized in that: The embedded processor receives real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data via a data bus. Before input, the data is normalized to unify the dimensions. The data is then input into the dynamic calibration model. After receiving the input, the dynamic calibration model performs forward propagation calculations and uses parameter weights for data processing to ensure real-time fusion and synchronization of the input data.

7. The pressure gauge remote intelligent diagnostic and dynamic calibration system according to claim 1, characterized in that: In its internal processing, the dynamic calibration model uses the nonlinear effect of real-time viscosity data on fluid resistance as the first weighting factor, the coupling effect of pressure fluctuation frequency data on sensor response delay as the second weighting factor, and the cumulative error of component aging as the cumulative usage frequency data as the third weighting factor. The first, second, and third weighting factors are used to calculate the degree of mutual influence through the cross-correlation algorithm built into the dynamic calibration model to quantify the multi-factor coupling distortion.

8. The pressure gauge remote intelligent diagnostic and dynamic calibration system according to claim 7, characterized in that: The first weighting factor, the second weighting factor, and the third weighting factor are fused according to the coefficient ratio obtained during training to generate a dynamic calibration coefficient. The dynamic calibration coefficient represents the error correction value in the current state. When output, it is temporarily stored as a floating-point number in the register of the embedded processor for subsequent calibration operations.

9. The pressure gauge remote intelligent diagnostic and dynamic calibration system according to claim 1, characterized in that: The embedded processor acquires the dynamic calibration coefficient and reads the pressure fluctuation frequency data. It performs a multiplication operation to multiply the dynamic calibration coefficient by the pressure fluctuation frequency to generate a temporary calibration value. Then, it adds the temporary calibration value to the original pressure measurement value and outputs the calibrated pressure value.

10. A method for implementing a remote intelligent diagnostic and dynamic calibration system for a pressure gauge as described in any one of claims 1-9, characterized in that: Includes the following steps: S1. The vibrating probe contacts the medium being measured and generates a mechanical vibration signal based on the change in viscous resistance. This signal is then converted into an electrical signal by a piezoelectric converter and digitally processed by an analog-to-digital converter to filter out environmental noise and output real-time viscosity data. S2. The pressure signal is collected by the strain gauge and amplified by the amplifier. The spectrum of the amplified signal is analyzed by the Fast Fourier Transform algorithm to identify the dominant fluctuation frequency component. After eliminating high-frequency noise by the filter, the pressure fluctuation frequency data is output. S3. The instrument running time is accumulated using a hardware timer, and the number of times the event counter is recorded. The two are integrated into a cumulative usage frequency data in the form of a comprehensive numerical index, stored in non-volatile memory and output through a digital interface. S4. The embedded processor normalizes the input real-time viscosity data, pressure fluctuation frequency data, and cumulative usage frequency data, and inputs them into the pre-trained dynamic calibration model. S5. The dynamic calibration model analyzes the nonlinear effect of viscosity on fluid resistance, the coupling effect of pressure fluctuation frequency on response delay, and the cumulative error of component aging due to usage frequency through cross-correlation algorithms. S6. Based on the coefficient ratio obtained from training, the three factors are fused to generate dynamic calibration coefficients in floating-point format. The dynamic calibration coefficients are multiplied by the pressure fluctuation frequency data to generate a temporary calibration value. Finally, the calibrated pressure value is added to the original pressure measurement value to perform dynamic compensation.

Images