An intelligent sensor data calibration method for reactor condition awareness

Abstract

The present application relates to the field of sensor data calibration, and particularly relates to a kind of intelligent sensor data calibration method for reactor state perception, power voltage is mapped to three engineering states of stability, drift and distortion based on analog device physical characteristics, and non-symmetrical hysteresis comparison logic is introduced to prevent critical point state oscillation.In stable state, baseline clamping and dead zone control are executed to filter background noise;In drift state, based on real-time voltage index preset lookup table, sensor gain attenuation is compensated using fixed-point displacement algorithm;In distortion state, start logic fuse mechanism to shield invalid signals.In addition, through zero-order hold strategy, low-speed power data and high-speed partial discharge signal are time-aligned.The present application effectively solves the problem of signal distortion and false alarm caused by power supply voltage fluctuation, significantly improves the robustness and monitoring accuracy of sensor in complex electromagnetic environment.

Description

Technical Field

[0001] This invention relates to the field of sensor data calibration, and more specifically to an intelligent sensor data calibration method for reactor state sensing. Background Technology

[0002] Intelligent sensors play a crucial role in reactor condition monitoring, primarily assessing the insulation health of the equipment by collecting high-frequency partial discharge signals. However, in actual operating conditions, these sensors are often situated in complex electromagnetic environments and are frequently constrained by power supply conditions, leading to fluctuations in their power supply voltage. Due to the physical characteristics of analog front-end circuits, these voltage fluctuations directly cause the device's operating point to shift, resulting in unexpected drift in signal gain and affecting the accuracy of the monitoring data.

[0003] To eliminate the effects of voltage fluctuations, existing technologies typically employ calibration algorithms based on digital signal processing. The main process involves the processor acquiring power supply voltage values ​​in real time, calculating the current gain deviation according to a preset linear model, and performing amplitude compensation on the acquired raw signal using floating-point multiplication and division operations to attempt to restore the true physical signal.

[0004] However, the above methods have obvious drawbacks in practical applications: on the one hand, on low-cost and computationally limited microcontrollers, performing floating-point correction operations on all sampled data throughout the entire time period will consume a lot of computing resources and reduce the system's response speed; on the other hand, existing algorithms usually lack judgment on the physical state of the circuit. When the power supply voltage drops severely and causes the analog device to enter the nonlinear region, simple linear ratio correction will not only fail to recover the distorted signal, but will also amplify the pseudo signal or noise floor generated by the circuit truncation, causing the back-end system to misjudge it as a high-amplitude partial discharge pulse, thereby triggering a false alarm. Summary of the Invention

[0005] To address the problem of false alarms caused by amplifying spurious signals or background noise errors resulting from circuit truncation in the aforementioned backend system, this invention proposes a smart sensor data calibration method for reactor state sensing. The method includes: real-time monitoring of the analog circuit supply voltage of the smart sensor and acquisition of raw sensor data; constructing a state machine containing multiple discrete voltage ranges based on the analog circuit supply voltage, wherein the state machine includes: a stable state, triggered in response to the supply voltage being within a preset range of the rated voltage; a drift state, triggered in response to the supply voltage deviating from the preset range but exceeding a set minimum operating voltage; and a distortion state, triggered in response to the supply voltage being below the minimum operating voltage; in response to the state machine being in a stable state, performing dead-zone control on the raw data and outputting it, wherein the dead-zone control includes forcibly setting data whose absolute difference from a set zero point value is less than a noise threshold to zero; in response to the state machine being in a drift state, obtaining a correction coefficient from a preset lookup table based on the current supply voltage, and multiplying the raw data with the correction coefficient to obtain corrected data; and in response to the state machine being in a distortion state, forcing the current output data to a preset invalid value and adding an invalid flag.

[0006] Compared to traditional sensor data processing methods that often ignore the impact of power supply voltage fluctuations on measurement accuracy, this invention monitors the analog circuit power supply voltage in real time and constructs a state machine that includes stable, drift, and distorted states. This allows for adaptive selection of data processing strategies based on voltage quality. Specifically, when the voltage is stable, dead-time control filters out noise; when the voltage drifts but is still usable, a lookup table corrects and compensates for the deviation; and when the voltage is too low and distorted, it is forcibly marked as invalid. This effectively solves the problem of measurement data distortion caused by power supply fluctuations, ensuring the accuracy and reliability of the sensor's output under different power supply qualities.

[0007] Furthermore, the state transition of the state machine includes hysteresis comparison logic, including: setting a high threshold to determine the state transition from high to low voltage; setting a low threshold to determine the state transition from low to high voltage; the difference between the high threshold and the low threshold constitutes the hysteresis.

[0008] Furthermore, the hysteresis value ranges from 50mV to 100mV.

[0009] Furthermore, obtaining the power supply voltage of the analog circuit includes: applying a sliding window mean filter to the N continuously collected power supply voltage values; and using the filtered mean as the current voltage value to trigger the state machine.

[0010] By applying sliding window mean filtering to the continuously acquired power supply voltage values, the voltage signal can be smoothed, effectively eliminating the interference of high-frequency power supply noise or transient spikes on voltage detection, preventing false triggering of the state machine due to single sampling errors, and thus providing a reliable voltage reference value for accurate determination of the state machine.

[0011] Furthermore, the triggering of the state machine also includes a trend prediction mechanism: calculating the slope of the change in the power supply voltage of the analog circuit, and when the absolute value of the slope exceeds a preset threshold, switching the state from the stable state to the drift state in advance.

[0012] By introducing a trend prediction mechanism based on the slope of voltage change, the system can identify the trend of power quality deterioration in advance and switch to the drift state for correction when the supply voltage drops rapidly but has not yet completely left the stable range. This improves the dynamic response speed of the system and avoids data errors during the voltage change transition period.

[0013] Furthermore, the corrected data is obtained by multiplying the original data with the correction coefficient, including using a fixed-point calculation method: ;in The corrected data; The original data; The correction coefficient is stored in integer form. This is the preset fixed-point displacement.

[0014] Using a fixed-point computation method that includes shift operations to replace floating-point operations for data correction significantly reduces the computational resource consumption and processing time of the microcontroller (MCU), improves the real-time performance of data calibration, and is particularly suitable for embedded intelligent sensor platforms with limited computing power.

[0015] Furthermore, in the initial stage of switching from the stable state to the drift state, the correction coefficients obtained from the lookup table are subjected to first-order low-pass filtering to smoothly transition to the target value over multiple sampling periods.

[0016] Within a preset number of sampling periods from the steady state to the drift state, the correction coefficient is processed by first-order low-pass filtering to avoid step or sudden changes in the output data caused by the sudden intervention of the correction coefficient, thus ensuring the smoothness and continuity of the sensor output data during the state transition process.

[0017] Furthermore, the method for obtaining the lookup table includes: injecting a standard reference signal with a constant amplitude into the signal input terminal of the smart sensor; using a programmable power supply to perform step scanning of the power supply voltage of the sensor within the voltage range corresponding to the drift state; at each voltage scanning point, acquiring the amplitude of the sensor's output signal, and calculating the ratio of the amplitude of the standard reference signal to the acquired amplitude of the output signal as a correction coefficient for the corresponding voltage point.

[0018] Furthermore, the lookup table is a two-dimensional lookup table, with the sensor's operating temperature as another index dimension; the method for obtaining the lookup table further includes repeatedly performing voltage scanning at multiple preset temperature points to obtain correction coefficients at different temperatures.

[0019] Furthermore, the current state code of the state machine is time-aligned with the corrected data using a zero-order hold method, and then encapsulated together into a unified data frame format containing a timestamp and a data validity flag.

[0020] The technical effects of this invention are as follows:

[0021] This invention constructs a state machine encompassing stable, drift, and distorted states to monitor the power supply voltage in real time and accordingly perform dead-zone denoising, lookup table coefficient correction, or forced invalidation processing on the raw data. Combined with hysteresis switching, trend prediction, and fixed-point calculation optimization, it effectively solves the data drift and distortion problems caused by power supply fluctuations. Without increasing hardware costs, it significantly improves the measurement accuracy and reliability of the reactor's intelligent sensor in complex electrical environments. Attached Figure Description

[0022] Figure 1 This is a schematic flowchart illustrating an intelligent sensor data calibration method for reactor state sensing according to an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram illustrating the original waveform of the sensor power supply voltage changing over time and the threshold determination in an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram illustrating the voltage state switching logic based on a hysteresis comparator in an embodiment of the present invention;

[0025] Figure 4 This is a timing diagram illustrating the change of system status codes over time in an embodiment of the present invention;

[0026] Figure 5 This is a schematic diagram illustrating the characteristic curve of the voltage correction coefficient K varying with the supply voltage in an embodiment of the present invention;

[0027] Figure 6 This is a schematic diagram illustrating the discrete mapping of the correction coefficients calculated using the fixed-point algorithm in an embodiment of the present invention as a function of the supply voltage.

[0028] Figure 7 This is a schematic diagram illustrating the structure of an output data frame containing calibration data and status information in an embodiment of the present invention. Detailed Implementation

[0029] 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, not all, of the embodiments of the present invention. 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.

[0030] The technical solution of this embodiment operates in an embedded microprocessor system for high-frequency partial discharge monitoring. The system architecture includes a main control unit, a sensing layer circuit, a storage unit, and a communication interface.

[0031] Specifically, the main control unit adopts a low-power microcontroller based on the ARM Cortex-M7 core, with its main frequency preferably set between 400MHz and 600MHz to meet the real-time requirements of edge computing. The sensing layer includes two independent analog-to-digital conversion channels. The first channel is connected to a high-frequency partial discharge sensor and configured as a high-speed acquisition channel with a sampling rate of no less than 10MSps to acquire the raw partial discharge pulse waveform. The second channel is connected to the power supply bus of the analog front-end and configured as a low-speed monitoring channel with a sampling rate between 1kSps and 10kSps to monitor the power supply voltage in real time. The fluctuations.

[0032] The storage unit includes on-chip Flash and external EEPROM, used to store preset calibration coefficient lookup tables and historical status data. The data transmission interface adopts a CAN-FD bus or RS485 differential bus with high anti-interference capability, responsible for uploading the processed data packet containing validity flags to the remote monitoring backend.

[0033] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0034] An embodiment of a smart sensor data calibration method for reactor state sensing:

[0035] like Figure 1 As shown, the intelligent sensor data calibration method for reactor state sensing according to the present invention includes:

[0036] S101. Based on the linear operating characteristics of analog devices, the continuously fluctuating power supply voltage is mapped to discrete engineering states and hysteresis comparison logic is constructed.

[0037] In this embodiment, the main control unit collects the power supply voltage of the analog circuit in real time through the low-speed monitoring channel, and constructs a voltage state observer with hysteresis characteristics based on the physical characteristics of analog devices, such as operational amplifiers.

[0038] First, the system defines three discrete logic intervals to characterize the operating state of the circuit.

[0039] The first state is the steady state, corresponding to a voltage within a preset window near the rated value. In this state, the analog circuit is considered to be operating in the linear region with constant gain. For example, in a 3.3V power supply system, this range is set to... .

[0040] The second state is the drift state, which corresponds to the voltage deviating from the rated range but still being higher than the minimum operating voltage of the device. In this case, it is determined that the analog circuit is not distorted but the gain has physically shifted.

[0041] The third state is the distortion state, corresponding to a severe voltage drop or surge. In this state, the analog device is considered to have entered the saturation or cutoff region. For example, the operational amplifier's nonlinear inflection point... Set the boundary threshold for aberration.

[0042] Furthermore, to prevent state machine oscillations caused by minute fluctuations in the power supply voltage near the critical point, this embodiment introduces an asymmetric hysteresis comparison mechanism. The system sets a high threshold. State switching used to determine the direction of voltage degradation, setting a low threshold. State switching used to determine the direction of voltage recovery.

[0043] Hysteresis Defined as the difference between the high threshold and the low threshold, i.e.: ;

[0044] In this embodiment, The preferred setting is 50mV to 100mV. The physical meaning of this hysteresis range is that it only applies when the voltage fluctuation exceeds [a certain value]. Only when the system confirms the change in state will it demonstrate the action characteristics of the Schmitt trigger and ensure the decisiveness of state switching.

[0045] It should be noted that, to further enhance the system's robustness in complex electromagnetic environments, a sliding window averaging filter is configured in the processor before state determination. This filter uses a window size of N=10 and averages the voltage values ​​from the most recent 10 samples to filter out nanosecond-level transient glitches. In other embodiments where computing power allows, a trend prediction algorithm, i.e., calculating the slope of voltage change, can also be used. If the slope exceeds the preset threshold of 0.5V drop per millisecond, the state will be switched to the drift state in advance, even if the current absolute value of the voltage is still in the stable region, in order to compensate for the physical delay caused by the discharge of the capacitor in the hardware circuit.

[0046] Combination Figures 2 to 4 The voltage state monitoring and hysteresis logic of this embodiment will be described in detail below:

[0047] like Figure 2As shown, the system continuously collects the supply voltage through a low-speed channel. The figure mainly contains three types of signal characteristics: the solid line, which fluctuates continuously, represents the real-time acquired power supply voltage waveform; the two horizontal dashed lines in the middle area represent the upper and lower thresholds of the steady state, respectively, and the band they enclose is the steady-state interval; the horizontal dashed line at the bottom represents the distortion threshold, such as 2.7V. Figure 2 As can be seen, over time, the power supply voltage may drop and cross the aforementioned threshold line.

[0048] To avoid frequent state transitions caused by voltage fluctuations near the threshold, this embodiment employs the following... Figure 3 The hysteresis comparison logic is shown. Figure 3 The horizontal axis represents the power supply voltage, and the vertical axis represents the system status code.

[0049] The path with the right arrow in the diagram represents the logic for determining the voltage drop process. When the voltage drops from a high potential, it must fall below a lower threshold to trigger a state switch. The path with the left arrow in the diagram represents the logic for determining the voltage recovery process. When the voltage recovers from a low potential, it must rise above a higher threshold to restore the previous state. The rectangular area between the two vertical dashed lines in the diagram is the defined hysteresis interval. Within this range, the system maintains its original state, thereby enhancing the system's anti-interference capability.

[0050] Figure 4 Showing Figure 2 The voltage waveform shown passes through Figure 3 The image shows the final output after hysteresis logic processing. The vertical axis represents discrete engineering state codes, where 0 represents a stable state, 1 represents a drift state, and 2 represents a distorted state.

[0051] In the 0.00s to 0.35s interval, despite slight voltage fluctuations, the status code remains at 0 (stable state); in the 0.35s to 0.85s interval, the voltage drops and exceeds the hysteresis lower limit, and the status code flips to 1 (drift state); in the 0.85s to 1.20s interval, the voltage drops significantly below the distortion threshold, and the status code jumps to 2 (distortion state), triggering a logic circuit breaker.

[0052] The voltage then recovered, and the status codes were restored sequentially. Figure 4 The results showed that the hysteresis logic effectively filtered out transient noise and output a stable state control signal.

[0053] S102. In response to the state machine being in a stable state, baseline clamping and dead-time control are performed on the acquired high-frequency signal to filter out background noise in the circuit.

[0054] When the voltage state observer determines that the system is in a stable state, the system assumes that the ADC reference voltage is accurate and the analog front-end gain has not shifted, and then enters the transparent transmission mode.

[0055] In this mode, the processor does not perform multiplication correction operations, but instead executes dead-time control logic. The system first calculates the currently sampled signal. With zero point The absolute value of the deviation, and compared with the preset background noise threshold. Compare them.

[0056] like If the condition is met, the output data will be forced to be set to zero; otherwise, the original data will be output.

[0057] Its beneficial effects are that, under ideal operating conditions with stable power supply, unnecessary floating-point or fixed-point operations are avoided, saving the MCU's computing power resources to the greatest extent, while effectively suppressing the static noise floor of the sensor through dead-time logic.

[0058] S103. In response to the state machine being in a drift state, a lookup table is preset based on the current voltage index, and signal ratio correction is performed using a fixed-point displacement algorithm.

[0059] When the supply voltage fluctuates and enters a drift state, the ADC quantization reference shifts accordingly. At this point, the processor initiates a ratio lookup table correction strategy.

[0060] First, the system uses the currently collected power supply voltage. As the index key, the corresponding correction coefficient is retrieved from the lookup table (LUT) pre-stored in Flash. .

[0061] Next, the retrieved coefficients are used to analyze the original partial discharge signal. Corrections are made. To achieve efficient computation on a low-cost MCU, this embodiment uses fixed-point calculation formulas instead of floating-point division:

[0062] ;

[0063] in, For the calibrated output data, This is the fixed-point displacement.

[0064] In this embodiment, The value is 12. The above fixed-point calculation method can amplify the floating-point coefficients. The value is stored as an integer and then restored during operation by right shifting by 12 bits (i.e., dividing by 4096).

[0065] From the above formula, it can be seen that when the input signal... If the coefficients in the lookup table remain unchanged, The voltage drop will be increased accordingly, resulting in the final output. The gain will be increased to compensate for the sensor gain reduction caused by the drop in power supply voltage.

[0066] Furthermore, at the instant of switching from a steady state to a drift state, in order to avoid a step discontinuity in the output waveform, the system extracts... Perform a first-order low-pass filter to smoothly transition the gain coefficient to the target value over 3 to 5 consecutive sampling periods.

[0067] Combination Figure 5 and Figure 6 The lookup table correction strategy of this embodiment will be explained as follows:

[0068] Figure 5 This visually illustrates the physical layer's gain correction logic. The horizontal axis represents the supply voltage, and the vertical axis represents the correction coefficient. .

[0069] The horizontal dashed line in the figure represents The baseline, i.e., the uncorrected state, is shown in the graph. The solid line curve with data points illustrates the monotonically increasing characteristic of the correction coefficient as the voltage decreases. It can be seen that when the voltage is at the rated value of 3.3V, the coefficient is 1.0; as the voltage decreases to the left, the correction coefficient increases to compensate for the gain attenuation of the analog circuit. The gain gradually increases. The shaded area in the diagram schematically represents the gain difference that needs to be compensated.

[0070] Figure 6 Further demonstration Figure 5 The engineering implementation in an embedded processor. To adapt to low-cost MCUs that do not support hardware floating-point operations, this embodiment uses fixed-point conversion technology.

[0071] The vertical axis represents the path through which... The integer coefficients after displacement processing, i.e. As shown by the dotted curve in the figure, the original decimal coefficients are mapped to an integer sequence between 4096 and 5000. For example, when the voltage is 3.3V, the corresponding fixed-point value is 4096; when the voltage drops to 2.7V, the corresponding fixed-point value rises to approximately 5000. Through this mapping, the processor only needs to perform integer lookup table operations and shift operations to complete high-precision signal calibration.

[0072] S104. In response to the state machine being in an abnormal state, the logic circuit breaker mechanism is executed to shield invalid physical signals and prevent false alarms.

[0073] When the supply voltage deteriorates further and enters a distorted state, the analog front-end circuit has entered the nonlinear region, and the physical signal undergoes irreversible distortion. At this point, the system executes an active gating shielding strategy.

[0074] Specifically, the processor forces the validity flag (Flag_Valid) of the current data frame to 0 and outputs the data. Forced to be pulled low to zero or marked as invalid (NaN), thus forming a soft circuit breaker mechanism, that is, cutting off the flow of data to the upper layer algorithm when the physical layer signal is unreliable.

[0075] This embodiment uses a logic gate to prevent the output signal from being truncated or oscillating due to insufficient power supply to the operational amplifier, which is then misjudged as a high-amplitude partial discharge pulse by the back-end algorithm. This solves the problem of false alarms caused by voltage drops in traditional passive sensors.

[0076] S105. Construct an automated calibration platform based on a precision power supply and a standard signal generator to generate a high-precision voltage correction coefficient lookup table.

[0077] To support accurate calibration in step S103, a lookup table needs to be pre-built during the sensor manufacturing process. This embodiment constructs an automated calibration platform that includes a programmable precision power supply, a standard signal generator, and a constant temperature chamber.

[0078] The calibration process is as follows:

[0079] First, a standard sine wave signal with a constant amplitude is injected into the sensor signal input terminal. .

[0080] Secondly, control the programmable power supply to bring the sensor's supply voltage down from the drift state lower limit. Initially, gradually increase the voltage to the rated value in 10mV increments. .

[0081] At each voltage step point Collect raw values ​​output by the sensor. And calculate the correction factor. : It is important to note that before performing the above operations, a judgment must be made. Whether it is a non-zero valid value; if it is zero, mark it as a fault or skip the point.

[0082] Finally, the floating-point coefficients enlarge The result is multiplied and rounded down, then stored in the specified address area of ​​the sensor's Flash memory.

[0083] It should be noted that if the sensor application scenario involves a large temperature difference, such as -40℃ to 85℃, this lookup table can be expanded into a two-dimensional array, i.e. At this point, the calibration process needs to repeat the voltage scanning steps at different temperature points in the constant temperature chamber to further eliminate the influence of temperature drift on the sensor gain.

[0084] S106. The low-speed power state data and high-speed partial discharge signal data are time-aligned and encapsulated using the zero-order hold method.

[0085] Since the frequency of the power supply sampling channel is much lower than that of the partial discharge signal sampling channel, the system adopts a zero-order hold strategy for data alignment in order to ensure data consistency.

[0086] Specifically, the system locks the latest power status code and calibration coefficients and extends their application to all subsequent high-frequency signal frames until the next power sampling update. Finally, the processor encapsulates the calibrated waveform data payload, the current voltage status code, and the validity flags into a unified data frame format.

[0087] The data frame structure typically includes: a header preamble, a 32-bit timestamp, a 1024-point waveform data payload, an 8-bit status code (indicating stability / drift / distortion), a 1-bit validity flag, and a CRC checksum. This structure ensures that the backend monitoring system can not only obtain the waveform but also trace the current power supply health status.

[0088] like Figure 7 As shown, this illustrates the final output data frame structure constructed in this embodiment. The data packet is divided into five functional domains from left to right:

[0089] Header preamble: Located at the beginning of the frame, used for data stream synchronization and identification, example value is 0xAA55;

[0090] Timestamp: Used to record the moment of data collection, ensuring that the backend system can perform time-series alignment;

[0091] Waveform data payload: The largest data field, containing 1024 points of calibrated partial discharge waveform data;

[0092] Status code and valid bit: The status code (8 bits) records the power state corresponding to the data frame, and the valid bit (1 bit) is used to indicate whether the data is available, for example, set to 0 in the distorted state;

[0093] CRC check: Located at the end of the frame, it is used to verify the integrity of data during transmission.

[0094] This structure ensures that the stored and transmitted data not only includes physical waveforms but also carries power supply health records, facilitating subsequent fault tracing and analysis.

Claims

1. A method for calibrating intelligent sensor data for reactor state sensing, characterized in that, The method includes: real-time monitoring of the analog circuit supply voltage of the smart sensor and acquisition of raw sensor data; and construction of a state machine containing multiple discrete voltage ranges based on the analog circuit supply voltage. The state machine includes: Stable state, triggered when the supply voltage is within the preset range of the rated voltage; Drift state, triggered when the supply voltage deviates from the preset range but is higher than the set minimum operating voltage; Distortion state, triggered when the supply voltage is lower than the minimum operating voltage. In response to the state machine being in a stable state, the output is performed after dead-time control on the original data. Dead-time control includes forcibly setting data whose absolute difference from the set zero point value is less than the noise threshold to zero. In response to the state machine being in a drift state, a correction coefficient is obtained from a preset lookup table based on the current supply voltage, and the original data and the correction coefficient are multiplied to obtain the corrected data; obtaining the correction coefficient from the preset lookup table includes: injecting a standard reference signal with a constant amplitude into the signal input terminal of the smart sensor; using a programmable power supply to perform step scanning of the sensor's supply voltage within the voltage range corresponding to the drift state; at each voltage scanning point, acquiring the amplitude of the sensor's output signal, and calculating the ratio of the standard reference signal amplitude to the acquired output signal amplitude as the correction coefficient for the corresponding voltage point; In response to the state machine being in a distorted state, the current output data is forced to a preset invalid value and an invalid flag is added.

2. The intelligent sensor data calibration method for reactor state sensing according to claim 1, characterized in that, The state machine's state transitions include hysteresis comparison logic, including: A high threshold is set to determine the voltage transition from high to low. A low threshold is set to determine the voltage transition from low to high. The difference between the high threshold and the low threshold constitutes the hysteresis.

3. The intelligent sensor data calibration method for reactor state sensing according to claim 2, characterized in that, The hysteresis value ranges from 50mV to 100mV.

4. The intelligent sensor data calibration method for reactor state sensing according to claim 1, characterized in that, Obtaining the power supply voltage of the analog circuit includes: Apply sliding window mean filtering to the N continuously collected power supply voltage values; The filtered mean value is used as the current voltage value that triggers the state machine.

5. The intelligent sensor data calibration method for reactor state sensing according to claim 1, characterized in that, The triggering of the state machine also includes a trend prediction mechanism: Calculate the slope of the change in the power supply voltage of the analog circuit. When the absolute value of the slope exceeds a preset threshold, switch the state from the steady state to the drift state in advance.

6. The intelligent sensor data calibration method for reactor state sensing according to claim 1, characterized in that, The corrected data is obtained by multiplying the original data with the correction coefficient, including by using a fixed-point calculation method: ； in The corrected data; The original data; The correction coefficient is stored in integer form. This is the preset fixed-point displacement.

7. The intelligent sensor data calibration method for reactor state sensing according to claim 1, characterized in that, Within a preset number of sampling periods from the stable state to the drift state, the correction coefficients obtained from the lookup table are subjected to first-order low-pass filtering to smoothly transition to the target value over multiple sampling periods.

8. The intelligent sensor data calibration method for reactor state sensing according to claim 1, characterized in that, The lookup table is a two-dimensional lookup table, with the sensor's operating temperature as another index dimension; The method for obtaining the lookup table further includes repeatedly performing voltage scans at multiple preset temperature points to obtain correction coefficients at different temperatures.

9. The intelligent sensor data calibration method for reactor state sensing according to claim 1, characterized in that, The current state code of the state machine is time-aligned with the corrected data using a zero-order hold method, and then encapsulated together into a unified data frame format that includes a timestamp and a data validity flag.

Images