Method and system for determining abnormality of metrological equipment based on big data evaluation

Abstract

The application discloses a metrological equipment abnormality determination method and system based on big data evaluation, and belongs to the technical field of computer data processing and system monitoring, which comprises the following steps: obtaining structure characteristic parameters and initial electrical performance calibration data of a to-be-monitored metrological equipment and establishing a performance digital twin model; obtaining multi-dimensional environmental stress data of the metrological equipment under a transportation environment, and injecting the performance digital twin model to simulate an internal parameter drift process of the metrological equipment and generate a predicted performance degradation sequence; synchronously collecting real-time physical feedback signals of the metrological equipment in the transportation process, generating an actual performance state sequence, and comparing the actual performance state sequence with the predicted performance degradation sequence to calculate a performance mismatch degree; and when the performance mismatch degree exceeds a stability threshold, generating an equipment metrological performance abnormality early warning signal. The method of constructing a performance digital twin model to simulate and predict performance degradation and comparing the actual state data collected to calculate can accurately determine the equipment performance abnormality and trace the cause.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computer data processing and system monitoring technology, and in particular to a metrological equipment abnormality determination method and system based on big data evaluation. BACKGROUND

[0002] As a key terminal for measurement, data collection and transmission, the accuracy and stability of metrological equipment are the basis for ensuring the normal operation of trade settlement, energy management, process control and other fields. The internal of such equipment usually integrates high-precision analog circuits and sensitive electronic components, whose performance is extremely susceptible to various environmental stresses during transportation, installation and use. Using data processing technology to monitor and evaluate the health status of metrological equipment throughout its life cycle is an important means to ensure its accurate and reliable value.

[0003] Currently, the state monitoring of metrological equipment in transportation or use usually adopts discrete or simplified technical means. For example, irreversible impact labels or temperature and humidity indicator cards are placed in the equipment package to roughly judge whether the equipment has been exposed to extreme environments exceeding the preset threshold. Or, after installation, periodic on-site verification or functional self-checking is performed to determine whether the equipment is qualified by comparing the measurement results at a specific time point with the standard value. Some advanced solutions integrate a single sensor such as an accelerometer to record impact event data during transportation.

[0004] However, the existing technical means have obvious technical defects. The threshold-based scheme cannot reflect the cumulative damage process of the equipment under continuous stress below the threshold or the coupling effect of multiple stresses, and is prone to false negatives. The periodic on-site verification method has a long detection interval and cannot timely capture the gradual or sudden changes in performance occurring between two verifications, which may cause the equipment to operate with defects and result in losses. In addition, most existing monitoring methods can only provide qualitative information on whether an event has occurred, lack quantitative analysis capabilities for the causal relationship between environmental stress and internal performance degradation of the equipment, and cannot effectively distinguish between normal performance drift and real physical damage, nor support accurate fault diagnosis and root cause tracing. SUMMARY

[0005] To solve the above problems, the present application provides a metrological equipment abnormality determination method and system based on big data evaluation, which adopts a method of constructing a performance digital twin model to simulate and predict performance degradation, and comparing and calculating with the actual state data collected, which can accurately determine the performance abnormality of the equipment and trace its causes before the metrological function is substantially disabled.

[0006] The above-mentioned object can be achieved by the following scheme:

[0007] The method for determining abnormality of metrological equipment based on big data evaluation comprises the following steps: obtaining structural characteristic parameters and initial electrical performance calibration data of a metrological equipment to be monitored, and establishing a performance digital twin model reflecting stress sensitivity of internal components of the equipment according to the structural characteristic parameters; obtaining multi-dimensional environmental stress data of the metrological equipment in a transportation environment in real time, and inputting the multi-dimensional environmental stress data as input variables into the performance digital twin model; driving the performance digital twin model to run, simulating internal parameter drift process of the metrological equipment under current cumulative environmental stress, and generating a predicted performance degradation sequence; synchronously collecting real-time physical feedback signals of the metrological equipment in the transportation process, and generating an actual performance state sequence; comparing the predicted performance degradation sequence and the actual performance state sequence in a unified performance evaluation space, and calculating a performance mismatch degree; when the performance mismatch degree exceeds a stability threshold set according to equipment precision level, generating an equipment metrological performance abnormality warning signal.

[0008] Optionally, the performance digital twin model comprises the following steps: extracting a list of components and spatial layout information of a key measurement loop inside the metrological equipment; calling physical sensitivity matrices of each component to vibration acceleration, temperature and humidity change and electromagnetic interference from a preset sensitivity database; and constructing a stress and parameter mapping operator based on the physical sensitivity matrices, so as to represent the conduction mechanism of external physical stress to internal electrical parameter drift.

[0009] Optionally, the multi-dimensional environmental stress data comprises the following: instantaneous impact and continuous vibration acceleration data obtained by a three-axis acceleration sensor; environmental stress gradient change data obtained by a temperature and humidity monitoring unit; and spatial electromagnetic field intensity monitoring data reflecting stability of a microenvironment where the equipment is located.

[0010] Optionally, the generation of the predicted performance degradation sequence comprises the following: applying an incremental step evolution algorithm to calculate electrical performance offset values of internal components caused by multi-dimensional environmental stress in each time slice, wherein the offset values are obtained by accumulating and calculating various environmental stress components and corresponding stress damage weights; and accumulating the offset values in time sequence to obtain a metrological value drift curve of the metrological equipment, as the predicted performance degradation sequence.

[0011] Optionally, the generation of the actual performance state sequence comprises the following: synchronously obtaining internal reference level signals or self-check pulse frequency characteristics of the metrological equipment; associating and binding the internal reference signals with physical state data measured by external sensors according to time stamps to form an enhanced actual state sequence; and when the warning signal is generated, synchronously outputting internal signal slices corresponding to an abnormal period to assist in judging a physical damage degree of the metrological function of the equipment.

[0012] Optionally, the computing performance mismatch degree comprises: extracting a spatial distance between the actual performance state characteristic value and the predicted performance offset value on a time slice by time slice basis; introducing a weight coefficient reflecting the importance of different measurement accuracy levels to perform a weighted average calculation on the spatial distance to generate the performance mismatch degree.

[0013] Optionally, the method further comprises: identifying a time point at which the performance mismatch degree first exceeds a threshold, and determining a mismatch starting state grid corresponding thereto; extracting a triggering key variable injected before the time point from a running log of the digital twin model, wherein the triggering key variable comprises an external environmental stress value that has a numerical jump; inputting the triggering key variable into a causal analysis model embedded with a physical topology logic to perform deduction and generate a chain impact path of the stress inside the device.

[0014] Optionally, the generating of the chain impact path of the stress inside the device comprises: quantifying a state change amplitude of each circuit node on each conduction path; calculating a change ratio of a node electrical parameter value after being affected by the stress relative to a normal working condition reference value; and screening out a main path with a change ratio exceeding a preset impact threshold as the chain impact path causing the device anomaly.

[0015] Optionally, the method further comprises: storing an abnormal early warning signal, a performance mismatch degree, a triggering key variable and a chain impact path as a structured case in a case database; using a machine learning algorithm to mine historical cases to analyze a correlation rule between an environmental stress combination and a device electrical parameter degradation; and using the correlation rule to update an internal parameter matrix and a physical response logic of the performance digital twin model.

[0016] Based on the same inventive concept, the present application also provides a metrological equipment anomaly determination system based on big data evaluation, which comprises:

[0017] a digital twin modeling module configured to obtain structural characteristic parameters and initial electrical performance calibration data of a metrological equipment to be monitored, and to establish a performance digital twin model reflecting the stress sensitivity of internal components of the equipment according to the structural characteristic parameters;

[0018] a multi-dimensional stress acquisition module configured to acquire multi-dimensional environmental stress data of the metrological equipment in a transportation environment in real time, and to inject the multi-dimensional environmental stress data as input variables into the performance digital twin model;

[0019] a performance prediction evolution module configured to drive the performance digital twin model to run, simulate an internal parameter drift process of the metrological equipment under a current cumulative environmental stress, and generate a predicted performance degradation sequence;

[0020] Actual status monitoring module: used to synchronously collect real-time physical feedback signals of the metering equipment during transportation and generate an actual performance status sequence;

[0021] Performance mismatch assessment module: used to compare the predicted performance degradation sequence with the actual performance state sequence within a unified performance evaluation space, and calculate the performance mismatch degree;

[0022] Anomaly detection and early warning module: used to generate an early warning signal for abnormal equipment metering performance when the performance mismatch exceeds the stability threshold set according to the equipment accuracy level.

[0023] Compared with the prior art, the present invention has the following advantages:

[0024] 1. By constructing a performance digital twin model reflecting the physical characteristics of the equipment and injecting real-time environmental stress data into the model to generate a predictive performance degradation sequence, predictive assessment of the equipment's performance status is achieved. Compared to traditional post-fault diagnosis or periodic testing, this method can issue early warnings by comparing the deviation between predictions and actual values ​​before the equipment's metering function suffers substantial failure. This transforms equipment condition management from reactive to proactive, gaining a valuable time window to ensure equipment integrity.

[0025] 2. It enables precise attribution and root cause tracing of metering equipment anomalies. When performance mismatch exceeds limits, it not only issues an early warning but also automatically pinpoints the instantaneous time of the anomaly and the triggering key variables. Utilizing a causal analysis model with embedded physical topology logic, it deduces the chain reaction path of external stress within the equipment. This provides technicians with intuitive and physically meaningful decision support for quickly locating the root cause of faults, assessing the extent of physical damage, and developing effective maintenance strategies.

[0026] 3. It endows the evaluation model with the ability to self-learn and continuously evolve, ensuring its effectiveness and accuracy in long-term application. By storing each confirmed anomaly as a structured case in the database and using machine learning algorithms to mine historical data, it can automatically parse out deeper correlation rules between environmental stress combinations and equipment degradation, and use these rules to update and optimize the internal parameters and response logic of the performance digital twin model, enabling the model to continuously improve itself with the accumulation of empirical data.

[0027] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a flowchart illustrating the method for determining anomalies in metering equipment based on big data assessment, according to an embodiment of the present invention.

[0030] Figure 2 This is a schematic diagram illustrating the performance degradation prediction and actual measurement comparison, as well as the mismatch evolution of the metering equipment anomaly determination method and system based on big data evaluation, according to an embodiment of the present invention.

[0031] Figure 3 This is a schematic diagram illustrating the chain reaction path node quantification screening of the metering equipment anomaly determination method and system based on big data assessment, according to an embodiment of the present invention.

[0032] Figure 4 This is a schematic diagram of the structure of the metering equipment anomaly determination system based on big data assessment according to an embodiment of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0034] Reference Figure 1 One embodiment of the present invention proposes a method for determining anomalies in metering equipment based on big data assessment. By constructing a performance digital twin model to simulate and predict performance degradation and comparing it with the collected actual state data, the method can accurately determine the performance anomalies of the equipment and trace their causes before the metering function suffers substantial failure.

[0035] The method described in this embodiment specifically includes:

[0036] Obtain the structural characteristic parameters and initial electrical performance calibration data of the metering equipment to be monitored, and establish a performance digital twin model reflecting the stress sensitivity of the internal components of the equipment based on the structural characteristic parameters;

[0037] Real-time acquisition of multi-dimensional environmental stress data of metering equipment under transportation environment, and injection of multi-dimensional environmental stress data as input variables into the performance digital twin model;

[0038] The performance digital twin model is driven to run, simulating the internal parameter drift process of the metering equipment under the current cumulative environmental stress, and generating a predicted performance degradation sequence;

[0039] The real-time physical feedback signals of the metering equipment during transportation are collected synchronously to generate an actual performance status sequence;

[0040] Within a unified performance evaluation space, the predicted performance degradation sequence is compared with the actual performance state sequence to calculate the performance mismatch.

[0041] When the performance mismatch exceeds the stability threshold set according to the equipment accuracy level, an abnormal equipment measurement performance warning signal is generated.

[0042] Specifically, a virtual benchmark for equipment performance degradation was created by constructing a performance digital twin model reflecting the physical structure and stress-sensitive characteristics of the metering equipment. Multi-dimensional environmental stress data from real-time monitoring under transportation conditions was used as the driving input to dynamically simulate and deduce the expected performance drift process of the equipment under the current cumulative stress, thereby generating a predicted performance degradation sequence. Simultaneously, the method synchronously collects the equipment's own real-time physical feedback signals, forming an actual performance state sequence reflecting its true operating condition. The core principle is to continuously compare the performance trajectory predicted by the virtual model with the measured state trajectory in the physical world under the same evaluation system. By quantifying the degree of mismatch between the two, it is determined whether the equipment's operating state deviates from its normal physical laws under the current environmental stress. By comparing the expected performance degradation predicted by the model with the actual measured state of the equipment, normal performance drift caused by known environmental stress can be effectively distinguished from abnormal performance degradation caused by unexpected impacts, potential defects, or unknown factors. This avoids false alarms caused by normal drift due to environmental changes and improves the accuracy and reliability of anomaly detection. It can detect early signs of performance deviation from the normal evolution trajectory before the equipment's metering function suffers substantial failure, thus providing key decision-making basis for taking preventive maintenance, replacing equipment, or investigating the cause of the accident, ensuring the integrity of the metering equipment and the accuracy of the value transmission.

[0043] Optionally, the performance digital twin model includes:

[0044] Extract the component list and spatial layout information of the key measurement circuits inside the metering equipment;

[0045] Retrieve the physical sensitivity matrix of each component for vibration acceleration, temperature and humidity changes, and electromagnetic interference from the preset sensitivity database;

[0046] A stress-parameter mapping operator is constructed based on the physical sensitivity matrix to characterize the transmission mechanism of external physical stress drifting to internal electrical parameters.

[0047] Specifically, the target metrology equipment is deconstructed and analyzed. By reviewing design drawings, bill of materials (BOM), and circuit diagrams, the component list and 3D spatial layout information of its internal key measurement loops are extracted. Key measurement loops typically refer to parts that directly affect metrological accuracy, such as signal sampling front-ends, amplification and filtering circuits, and analog-to-digital conversion units. This list and layout information constitute the static geometric topology of the model. Dynamic behavioral attributes are assigned to each key component by calling a pre-defined sensitivity database. This database stores massive amounts of component performance drift data under specific stresses, derived from accelerated life testing, supplier specifications, or finite element analysis. A unique physical sensitivity matrix is ​​extracted for each component. This matrix is ​​a multi-dimensional array that quantifies a component's electrical parameters, such as resistance, capacitance, operational amplifier offset voltage, sensitivity to different types of environmental stresses, such as vibration acceleration in the range of 0 to 10g, temperature and humidity gradients in the range of -40 to 85 degrees Celsius, and electromagnetic interference field strength of 0.1 to 10 volts per meter at specific frequency bands. Based on the acquired component list, spatial layout information, and physical sensitivity matrix, a core stress-parameter mapping operator is constructed. This operator is a mathematical expression of the mechanism for transmitting external physical stress to internal electrical parameters; that is, a transformation function. Its input is multi-dimensional environmental stress data collected in real time by external sensors, and its output is the predicted drift values ​​of the electrical parameters of each key component inside the equipment. This mapping operator is calculated using the following formula:

[0048] ;

[0049] in, This represents the total offset of the electrical parameters of the k-th component inside the device. It is a dimensionless relative rate of change or an absolute change value with specific units. This represents the j-th type of external environmental stress data detected by the sensor, such as vibration acceleration values. It is an element in the physical sensitivity matrix, representing the sensitivity coefficient of the k-th component to the j-th type of stress. Its unit is determined according to the stress type and parameter type to ensure that the physical meaning of the product is correct, such as ohms per unit of gravitational acceleration. This is the structural conduction coefficient, calculated based on the spatial layout information of the k-th component within the device. It reflects the attenuation or enhancement effect of external stress as it is transmitted to the component's location, and its value is obtained through structural mechanics or electromagnetic field simulation. Using this operator, the performance digital twin model can transform changes in the external environment into parameter degradation at the internal circuit level, thereby simulating the performance evolution of the entire device.

[0050] For example, taking the precision sampling resistor component k in a single-phase electronic energy meter as an example, its initial resistance value is First, through deconstructive analysis, it was determined that the resistor is located at the edge of the PCB board and is susceptible to mechanical vibration. Then, by accessing a sensitivity database, the physical sensitivity matrix elements of the resistor to vibration stress (stress j) were extracted. This coefficient is set based on the principle of thin-film resistance strain effect. That is, each increase Acceleration resistance offset Structural conductivity Based on finite element simulation, considering that the resistor is located in the vibration amplification zone, it is set to a dimensionless value of 1.2, indicating that the stress transmitted to this location is amplified by a factor of 1.2. If the external vibration stress collected by the sensor... for According to the formula Perform verification: Here The absolute change value represents the change in resistance due to vibration. The resistance drift is calculated. This operator transforms multidimensional stress into electrical parameter shifts through the principle of linear superposition, with clear physical meaning and a reproducible closed-loop calculation process.

[0051] Optionally, the multidimensional environmental stress data includes:

[0052] Instantaneous impact and continuous vibration acceleration data acquired through a triaxial accelerometer;

[0053] Environmental stress gradient change data obtained through temperature and humidity monitoring unit;

[0054] Spatial electromagnetic field intensity monitoring data reflecting the stability of the microenvironment in which the equipment is located.

[0055] Specifically, to acquire mechanical stress data, a high-precision triaxial accelerometer is deployed on the outer casing or internal rigid structure of the measuring equipment. This sensor operates continuously at a sampling rate of at least 1 kHz to capture two key mechanical events: instantaneous impact and continuous vibration. The system identifies and records the peak value, duration, and timing of impact events exceeding a set acceleration threshold, such as 10 times the force of gravity. The acquired acceleration time-domain signal is processed, typically using a Fast Fourier Transform algorithm to convert it into a power spectral density map in the frequency domain, or by calculating the root mean square value of the acceleration within each 1-second time window, thereby quantifying the distribution and overall intensity of vibration energy at different frequencies. To acquire temperature and humidity environmental stress data, a highly sensitive temperature and humidity monitoring unit is integrated near the key measurement loop inside the equipment. This monitoring unit acquires temperature and humidity readings at a frequency of, for example, once every 10 seconds, calculates the difference between the current reading and the previous sampling point reading in real time, and then divides it by the sampling time interval to obtain the temperature change rate (in degrees Celsius per minute) and the humidity change rate. This quantification of gradient changes allows for a more precise characterization of material fatigue and electrical parameter drift caused by thermal expansion and contraction and moisture penetration. This is especially important for precision components made of materials with different coefficients of thermal expansion, where rapid temperature changes are a major cause of performance degradation. To obtain electromagnetic environment stability data, a spatial electromagnetic field strength monitoring module is deployed inside or outside the equipment. This module, consisting of a broadband antenna and radio frequency detection circuitry, assesses the electromagnetic interference level of the microenvironment in which the metrology equipment is located. The module covers key frequency bands from tens of hertz to thousands of hertz, outputting the comprehensive field strength value within that band in real time, in volts per meter. When the metrology equipment passes near strong electromagnetic radiation sources, such as high-voltage lines, radio base stations, or large motors, the monitoring data will show significant jumps. These data are directly related to induced noise and signal distortion that may occur in the internal analog circuitry, providing a direct basis for evaluating the metrology performance stability of the equipment in complex electromagnetic environments. These three sets of processed time-series data—impact and vibration acceleration data, temperature and humidity gradient change data, and spatial electromagnetic field strength data—together constitute the multidimensional environmental stress data injected into the performance digital twin model.

[0056] For example, accelerometers are used for real-time monitoring via an accelerometer deployed on the meter casing. For continuous vibration, the system uses a root mean square (RMS) algorithm to quantify energy, as shown in the formula: ,in The instantaneous acceleration at a sampling rate of 1kHz N The total number of sampling points within the time window. If the average of the squares of 1000 data points within 1 second is 4.0, then the RMS is 2g. This algorithm can effectively filter out high-frequency noise and reflect the vibration intensity benchmark. For temperature and humidity gradients, the monitoring unit... Temperature measured at time ,exist Time measured The calculation formula is: The rate of temperature change was obtained. ,Right now , The temperature gradient is the rate of change of temperature. End of sampling time The measured temperature value, The starting sampling time The measured temperature value. This gradient value of 3.0 directly characterizes the severity of thermal stress, based on the thermodynamic principle that the thermal stress of a material is proportional to the rate of temperature change. Finally, the space electromagnetic field monitoring module converts the induced voltage into field strength through a radio frequency detection circuit. If the output is... This, together with the impact, vibration, and temperature and humidity gradient data, forms a multidimensional stress vector injection model.

[0057] Optionally, generating the predicted performance degradation sequence includes:

[0058] An incremental step evolution algorithm is applied to calculate the electrical performance deviation of internal components caused by multidimensional environmental stress in each time segment. The deviation value is obtained by accumulating various environmental stress components and their corresponding stress failure weights.

[0059] By accumulating the offset values ​​over time, the expected measurement value drift curve of the metering equipment is obtained, which serves as a predictive performance degradation sequence.

[0060] Specifically, an incremental step evolution algorithm is employed, a computational method that discretizes continuous-time processes. The transportation process is divided into a series of equal-length time segments, for example, each segment being 1 second. At the end of each time segment, a performance degradation calculation is performed. Within each independent time segment, the algorithm first calculates the incremental offset value of the internal component electrical performance caused by the multidimensional environmental stresses during that period. This offset value is calculated using a weighted summation model, which measures the destructive effects of different stresses using a uniform performance degradation scale. Specifically, this offset value is obtained according to the following formula:

[0061] ;

[0062] in, This represents the incremental drift of the quantity that the measuring device is expected to measure within the current time segment. Its unit is typically parts per million (ppm) or a tiny change in a specific unit of measurement. Δt is the duration of this time segment, for example, 1 second. This is the average or equivalent value of the j-th type of environmental stress component within this time segment. For example, it can be the equivalent value of continuous vibration obtained by calculating the root mean square value of acceleration data within 1 second, or the average value of the temperature change rate within this period. This represents the stress failure weight corresponding to the j-th type of environmental stress. It is a key parameter that quantifies the drift rate of a unit intensity of the j-th type of stress on the final measurement result per unit time, for example, in ppm per gravitational acceleration per hour. This weight... The performance drift is calculated using a digital twin model based on the physical sensitivity matrix of components and the structural conduction characteristics of the equipment, and corrected through pre-shipment calibration experiments. These incremental offset values ​​are accumulated over time to generate a complete predicted performance degradation sequence. The total drift variable is maintained. Iterate and update in the following manner:

[0063] ;

[0064] in, This represents the cumulative drift up to the current time point T. This represents the drift of the cumulative value from the previous time point. This is achieved by continuously recording the value at each time point. The system then generates a curve depicting the drift of the expected measurement value of the metering device over time. This data sequence, consisting of a series of timestamps and corresponding predicted drift values, serves as the predictive performance degradation sequence for subsequent comparisons.

[0065] For example, an incremental step evolution algorithm is used, and the step size is set. Assume the equivalent vibration stress monitored during the current time period is... Average temperature change rate Preset stress failure weight (Vibration) is The basis for this is the average error deviation of the meter under accelerated aging test in the laboratory; (Temperature change) According to the formula Perform incremental verification: The cumulative drift amount maintained. Assuming it has accumulated to Then the updated Through this iterative accumulation, the system generates a performance prediction curve with a resolution of 1 second, which clearly describes the evolution of measurement accuracy with transportation time and stress intensity.

[0066] Optionally, generating the actual performance state sequence includes:

[0067] Synchronously acquire the internal reference level signal or self-test pulse frequency characteristics of the metering device;

[0068] The internal reference signal is associated and bound with the physical state data measured by external sensors based on the timestamp to form an enhanced real-state sequence;

[0069] When the warning signal is generated, the internal signal slice corresponding to the abnormal period is output synchronously to help determine the degree of physical damage to the equipment's metering function.

[0070] Specifically, the system monitors the built-in internal reference level signal or self-test pulse frequency characteristics. For metering equipment with analog circuitry at its core, the output level of a highly stable reference voltage source is typically acquired, such as a 1.25V DC voltage generated by a precision bandgap reference source. The analog-to-digital converter (ADC) samples this level at a frequency of, for example, 100Hz. For digital metering equipment, the frequency characteristics of the self-test pulse generated by its internal clock system are measured, such as a standard pulse output driven by a high-precision crystal oscillator. The precise frequency value is obtained by measuring the number of pulses per unit time using an internal counter. These real-time acquired voltage or frequency value sequences constitute the basic data reflecting the internal physical state of the equipment. A uniform, high-precision timestamp is applied to each data source, including the internal reference signal sampling point and readings from external triaxial accelerometers, temperature and humidity monitoring units, and spatial electromagnetic field strength monitoring units. The data processing unit aligns and aggregates all data acquired at the same time based on these timestamps. For example, a data tuple is generated at a certain moment, containing the internal reference level value at that moment, as well as physical state data such as instantaneous impact acceleration, environmental stress gradient, and electromagnetic field strength. By concatenating these data tuples in chronological order, an enhanced real-state sequence is formed, where each data point simultaneously records both the external cause and the internal effect. When the performance mismatch exceeds a threshold, a warning signal for abnormal equipment metering performance is generated, which simultaneously triggers data extraction. The time point of the warning signal generation is immediately located, and a preset time window is extracted, such as 5 seconds before the anomaly occurs to 2 seconds after. This extracted data segment is called an internal signal slice, and it is a high-sampling-rate raw waveform or data point. By analyzing this slice, technicians can observe the details of instantaneous signal changes, such as sudden voltage drops, sharp increases in noise, or frequency drift and jitter. These features can directly point to specific physical damage modes, such as component burnout, poor solder joints, or shielding failure.

[0071] For example, a 0.1-class precision meter internally uses a high-stability bandgap reference source such as the ADR1399 to output a 1.25V reference voltage. Sampling is performed using a 16-bit precision ADC with a resolution of [missing information]. At time t, the actual voltage value acquired by the ADC is 1.24998V. Simultaneously, environmental stress data at that time, such as vibration (2g) and electromagnetic field (3V / m), are correlated to form a data tuple [t, 1.24998, 2, 3]. When the performance mismatch is determined to be excessive, a slicing mechanism is triggered, capturing the original waveform sequence of the 5 seconds prior to the anomaly point. For example, if a momentary voltage drop to 1.20V occurs at t+1s, the slice will record all 100 original sampling points within that 1s. By observing the slope and recovery pattern of the voltage drop, technicians can determine that the fault is caused by a poor solder joint instantaneous open circuit under high stress, rather than slow drift due to component aging.

[0072] Optionally, the computational performance mismatch includes:

[0073] Extract the spatial distance between the actual performance status feature values ​​and the predicted performance offset values ​​at each time segment;

[0074] By introducing weighting coefficients that reflect the importance of different measurement accuracy levels, the spatial distance is weighted and averaged to generate the performance mismatch degree.

[0075] Specifically, at each time segment, the actual performance state characteristic values ​​are first extracted. The internal reference level signal value acquired at the current time point is calculated and compared with the initially calibrated stress-free reference value to obtain a relative change reflecting the true physical drift. For example, if a voltage signal is acquired, this characteristic value is the difference between the current voltage and the reference voltage divided by the reference voltage. This value represents the actual performance drift of the equipment and is converted to a unit consistent with the predicted value, such as parts per million (ppm). The predicted performance offset value output by the performance digital twin model within the same time segment is extracted. The so-called spatial distance is the absolute difference between these two unitized values. The instantaneous deviation values ​​are weighted to generate the final performance mismatch index. Weighting coefficients reflecting the importance of different metering accuracy levels are introduced. This coefficient is a dimensionless parameter, and its value is directly linked to the accuracy class of the equipment. For high-precision equipment such as class 0.01, this weighting coefficient is set to a higher value, such as 1.2 to 1.5, to enhance its sensitivity to deviations; while for ordinary-precision equipment such as class 1.0, its set value is lower, such as 0.8 to 1.0. This coefficient is used to adjust the influence of deviations in the final evaluation results. The performance mismatch is generated by weighting the spatial distance, and the calculation formula is as follows:

[0076] ;

[0077] Where M is the final generated performance mismatch, a dimensionless value used directly for comparison with the subsequently set stability threshold. W is the weighting coefficient, which is obtained from the system configuration table based on the accuracy level calibrated at the device's factory. This refers to the actual performance status feature value extracted within the current time segment t, which is the normalized amount of the actual performance drift of the device. This refers to the predicted performance offset value output by the digital twin model of performance within the same time segment. The performance mismatch calculated by this formula accurately reflects the degree of deviation between the model prediction and the actual state of the equipment. Figure 2 As shown, the figure is divided into two related panels. The upper panel presents a comparison curve between the dashed line of the predicted performance degradation sequence generated by the incremental step evolution algorithm and the solid line of the actual performance state sequence extracted based on the internal reference signal, with the unit uniformly set to parts per million. This clearly demonstrates the deviation between the expected drift trajectory of the measurement values ​​under cumulative stress and the actual physical state. The lower panel shows the evolution process of the performance mismatch calculated according to the formula, and marks the preset stability threshold. When the mismatch index exceeds the threshold, the system's anomaly warning and data slicing mechanism are triggered.

[0078] For example, the predicted and measured values ​​are normalized. Assume that at time t, the performance deviation of the model prediction is... The value is 1.207 ppm. Meanwhile, the measured internal reference voltage is 1.249981 V. Given that the stress-free reference value is 1.250000 V, the actual performance condition characteristic value is... Regarding the device's 0.01 level of accuracy, a weighting coefficient W=1.5 was selected from the configuration table. The rationale for this setting is that high-precision equipment has extremely low tolerance for minute deviations, necessitating the amplification of mismatch performance. Based on the formula... Perform verification: The obtained mismatch degree of 20.9895 is far higher than the set threshold such as 5.0. Based on this, it is determined that the digital twin model can no longer accurately reflect the current physical damage, and root cause tracing must be initiated.

[0079] Optionally, the method further includes:

[0080] Identify the instantaneous time point that causes the performance mismatch to first exceed the threshold, and determine its corresponding mismatch initiation state grid;

[0081] Based on the stated time point, extract the triggering key variables injected before that time point from the operation log of the digital twin model, wherein the triggering key variables include the external environmental stress values ​​that cause numerical jumps;

[0082] By inputting triggering key variables into a causal analysis model with embedded physical topology logic, the chain reaction path of stress within the equipment is generated.

[0083] Specifically, the calculated performance mismatch is continuously compared in real time with a stability threshold set according to the device's accuracy level. Once the instantaneous point in time when the performance mismatch first exceeds the threshold is identified, this point is immediately captured, and the complete state at that moment is simultaneously solidified, forming a mismatch initiation state grid. This grid is a multi-dimensional data snapshot containing all key information within the time segment that caused the mismatch to exceed the limit, such as actual performance state characteristic values, predicted performance offset values, and all multi-dimensional environmental stress data injected into the digital twin model. Using the newly captured instantaneous point in time as a benchmark, the performance digital twin model's operating log is queried back to a very short time window before that moment, such as 100 milliseconds. This log records all input environmental stress values ​​in the form of a high-frequency time series. The data within this time window is scanned using a rate-of-change detection algorithm, specifically looking for external environmental stress values ​​that experience numerical jumps. If a stress value, such as the reading of a certain axis on a triaxial accelerometer, has a rate of change exceeding a preset jump threshold—for example, a value surging by more than 5 gravitational accelerations within 10 milliseconds—then that stress value and its occurrence time are identified as triggering key variables. The identified triggering key variables, such as impact events with specific amplitude and direction, are used as the sole input to a causal analysis model with embedded physical topology logic. This model integrates the complete circuit topology of the metering equipment, the three-dimensional spatial layout of components, and the mechanical properties of materials. The model begins its simulation, first demonstrating how the impact stress propagates within the equipment structure, for example, from the outer casing to the PCB board. Then, based on the deformation of the PCB board, it calculates the mechanical stress experienced by specific components mounted on it, such as precision resistors or crystal oscillators. Next, it calls upon the physical sensitivity matrix of the components to convert the mechanical stress into instantaneous drift of electrical parameters. Finally, it substitutes this parameter drift into the circuit simulation logic to calculate how this drift causes abnormal fluctuations in the voltage or frequency of key measurement nodes, ultimately forming a causal chain with clear physical meaning from external physical impact to internal metering function failure. This path represents the chain reaction path of the generated stress within the equipment.

[0084] For example, once the performance mismatch M > 5.0, a stability threshold is set based on 20% of the maximum permissible error of a 0.01-class meter, locking in the mismatch initiation time. Reviewing the operation log for the 100ms prior to this moment, it was found that the Z-axis reading of the triaxial accelerometer surged from 1g to 15g within 10ms. The rate of change detection algorithm calculated... The impact exceeded the preset threshold of 500g / s (set according to the logistics drop standard), thus identifying the impact as the triggering key variable. The 15g impact force was then incorporated into the causal analysis model. Based on the circuit topology logic, the expert system identified that the impact directly acted on the PCB board support, causing the resulting micro-strain to be transmitted to the pins of the voltage reference chip, resulting in a momentary fluctuation in the reference voltage from 1.25V. This process formed a complete causal path of external impact—structural strain—pin stress—reference level fluctuation, enabling automatic fault location.

[0085] Optionally, the chain reaction path of the generated stress within the device includes:

[0086] Quantify the state change amplitude of each circuit node on each conduction path;

[0087] Calculate the percentage change of nodal electrical parameters under stress relative to the baseline values ​​under normal operating conditions.

[0088] The main paths whose change rate exceeds the preset impact threshold are selected as the chain reaction paths that lead to equipment malfunctions.

[0089] Specifically, in the initial chain reaction path generated by the root cause tracing process, each circuit node along the path is traversed. In engineering, a circuit node refers to a component pin, a test point on a circuit board, or a connection point between different functional modules. For each node, the causal analysis model has simulated and calculated the magnitude of the state change at that node under the influence of the triggering key variable, i.e., the node's electrical parameter value after being affected by stress, such as the voltage value of the operational amplifier output pin, the frequency value of the crystal oscillator output, or the resistance value across a precision resistor. The absolute changes of different nodes are converted into dimensionless relative change ratios to conduct a uniform-scale importance assessment, avoiding misjudgments due to differences in the dimensions or benchmark values ​​of the parameters of each node. The change ratio of each node is calculated using the following formula:

[0090] ;

[0091] in, The calculated percentage change is a dimensionless pure number. This represents the nodal electrical parameter value after being affected by stress, which is derived from the causal analysis model. This represents the baseline value of the node under normal operating conditions. This baseline value is derived from the equipment design specifications or initial calibration database and represents the standard electrical parameter value of the node under no external abnormal stress. Through this calculation, the state changes of all nodes are normalized to the percentage deviation from their normal values. The calculated change percentage of each node is compared with a preset impact threshold. This impact threshold is set according to the criticality of the node and the functional sensitivity of the circuit to which it belongs. For example, for a reference voltage source node, the threshold might be set to 0.01%, while for a node in a general filter circuit, the threshold might be relaxed to 1%. All paths with change percentages exceeding their corresponding impact thresholds are filtered out. On any conduction path, if the change percentage of one or more nodes exceeds the limit, the entire path is determined to be the main path. This set of filtered paths is ultimately output as the cascading impact paths leading to equipment abnormalities, providing accurate fault location data for subsequent maintenance and design improvements. Figure 3 As shown in the figure, this chart uses a bar chart to quantitatively display the performance changes of several key circuit nodes, such as precision resistors, crystal oscillators, and operational amplifiers, involved in the root cause tracing process. The bars in the chart represent the relative change ratio of each node after being affected by stress, while the horizontal dashed lines represent preset impact thresholds for different circuit function sensitivities. By comparison, it can be seen that when the change ratio of a node, such as the crystal oscillator output frequency or the AD reference voltage, exceeds the corresponding threshold, that path is identified as the main conduction link leading to the metering anomaly.

[0092] For example, the causal analysis model evaluates a key node in the path—the output pin of a voltage amplifier. Under a 15g impact, the model simulates the electrical parameters of this node after stress. The baseline value under normal operating conditions According to the formula Perform verification: The functional sensitivity threshold for this precision amplifier node is known to be set at 0.01%. Since 0.16% > 0.01%, this node is deemed abnormal, and its conduction path is identified as the primary path. In contrast, if the offset of another ordinary filter capacitor node is only 0.05%, less than its 1% relaxation threshold, then that path is excluded. Through this quantitative comparison, the critical conduction links most likely to cause measurement accuracy deviations are screened from a massive number of analog paths, providing data support for design improvements.

[0093] Optionally, the method further includes:

[0094] Anomaly warning signals, performance mismatch, triggering key variables, and chain reaction paths are stored as structured cases in the case database;

[0095] Machine learning algorithms were used to mine historical cases and analyze the correlation rules between environmental stress combinations and equipment electrical parameter degradation.

[0096] The internal parameter matrix and physical response logic of the performance digital twin model are updated using the association rules.

[0097] Specifically, when a device's metering performance anomaly warning signal is triggered and subsequently confirmed by analysis or verification, the entire set of data for this event is automatically packaged. The anomaly warning signal identifier, the final calculated peak performance mismatch, previously identified triggering key variables (e.g., a 20g instantaneous impact), and the cascading impact path generated by the root cause tracing process are all integrated into a structured case. This case is stored in a dedicated case database in a standard data format, such as JSON or XML. Each case has a unique ID and timestamp for easy batch querying and analysis. For example, after accumulating more than 100 new cases in the database, a data mining task is automatically initiated. This task uses machine learning algorithms, such as Apriori or FP-Growth association rule mining algorithms, to analyze all historical cases in the case database. The goal of the algorithm is to identify the strong correlation between specific environmental stress combinations and the degradation of specific electrical parameters within the device. For example, the algorithm might discover a rule that when the root mean square value of vibration acceleration is between 3g and 5g and the rate of change of ambient temperature exceeds 5 degrees Celsius per minute, the probability of the drift amplitude of the internal reference voltage source output exceeding 0.05% reaches 85%. These resolved association rules quantify the equipment degradation behavior under various stress coupling effects. The association rules are input into the model update module. This module automatically adjusts the internal parameter matrix and physical response logic of the performance digital twin model based on the rule content. For example, for the aforementioned discovered association rules, the stress failure weights corresponding to the reference voltage source in the model are updated using the following formula:

[0098] ;

[0099] in, It is the updated stress failure weight. These are the original weights. α is a learning rate factor, typically a small value between 0.01 and 0.1, used to control the update step size and prevent overfitting. C is the confidence level of the association rule, for example, 85%. This is a confidence threshold, such as 70%. Only rules exceeding this threshold are used for updates. In this way, the model's physical sensitivity matrix is ​​dynamically corrected, making it more reflective of the multi-stress coupling effects under actual working conditions. Simultaneously, if new, high-confidence cascading effects are discovered, these paths are added to the causal analysis model's logic library, thus achieving a dual update of the model's internal parameter matrix and physical response logic.

[0100] For example, after the root cause analysis is completed, structured cases are generated and stored in the database. After accumulating 100 cases, data mining is initiated, revealing the rule: vibration > 10g and temperature change. At this point, the reference voltage weight becomes invalid. It is determined to update the stress failure weight. Let the original weight be... Learning rate factor The setting principle is to balance the model update speed and stability, with a rule confidence level C=0.85 and a confidence threshold. According to the formula Perform verification: The updated weights By re-solidifying the performance digital twin model and fine-tuning the weight coefficients, the model can predict performance degradation values ​​that are closer to actual measurements when faced with similar multi-stress coupling conditions in the future, thus achieving autonomous optimization and evolution of the model.

[0101] Reference Figure 4 Based on the same inventive concept, the present invention also provides a system for determining anomalies in metering equipment based on big data assessment, the system comprising:

[0102] Digital twin modeling module: used to acquire the structural characteristic parameters and initial electrical performance calibration data of the metering equipment to be monitored, and to establish a performance digital twin model reflecting the stress sensitivity of the internal components of the equipment based on the structural characteristic parameters;

[0103] Multidimensional stress acquisition module: used to acquire multidimensional environmental stress data of the metering equipment in the transportation environment in real time, and inject the multidimensional environmental stress data as input variables into the performance digital twin model;

[0104] Performance prediction evolution module: used to drive the performance digital twin model to run, simulate the internal parameter drift process of the metering equipment under the current cumulative environmental stress, and generate a predicted performance degradation sequence;

[0105] Actual status monitoring module: used to synchronously collect real-time physical feedback signals of the metering equipment during transportation and generate an actual performance status sequence;

[0106] Performance mismatch assessment module: used to compare the predicted performance degradation sequence with the actual performance state sequence within a unified performance evaluation space, and calculate the performance mismatch degree;

[0107] Anomaly detection and early warning module: used to generate an early warning signal for abnormal equipment metering performance when the performance mismatch exceeds the stability threshold set according to the equipment accuracy level.

[0108] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0109] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for determining anomalies in metering equipment based on big data assessment, characterized in that, The method includes: The structural characteristic parameters and initial electrical performance calibration data of the metering equipment to be monitored are obtained. Based on the structural characteristic parameters, a performance digital twin model reflecting the stress sensitivity of the internal components of the equipment is established. The performance digital twin model includes: extracting the component list and spatial layout information of the key measurement circuits inside the metering equipment; retrieving the physical sensitivity matrices of each component corresponding to vibration acceleration, temperature and humidity changes, and electromagnetic interference from a preset sensitivity database; and constructing a stress-parameter mapping operator based on the physical sensitivity matrix to characterize the transmission mechanism of external physical stress to internal electrical parameters. The stress-parameter mapping operator is calculated using the following formula: ； in, This represents the total offset of the electrical parameters of the k-th component inside the device. This represents the j-th type of external environmental stress data detected by the sensor. These are elements in the physical sensitivity matrix, representing the sensitivity coefficient of the k-th component to the j-th type of stress. It is the structural conductivity coefficient; Real-time acquisition of multi-dimensional environmental stress data of metering equipment under transportation environment, and injection of multi-dimensional environmental stress data as input variables into the performance digital twin model; The performance digital twin model is driven to run, simulating the internal parameter drift process of the metering equipment under the current cumulative environmental stress, and generating a predicted performance degradation sequence; The real-time physical feedback signals of the metering equipment during transportation are collected synchronously to generate an actual performance status sequence; Within a unified performance evaluation space, the predicted performance degradation sequence is compared with the actual performance state sequence to calculate the performance mismatch: ； Where M is the final generated performance mismatch degree, and W is the weighting coefficient. These are the actual performance state feature values ​​extracted within the current time segment t. The predicted performance offset value output by the digital twin model of performance within the same time segment; When the performance mismatch exceeds the stability threshold set according to the equipment accuracy level, an abnormal equipment measurement performance warning signal is generated. Identify the instantaneous time point that causes the performance mismatch to first exceed the threshold, and determine its corresponding mismatch initiation state grid; extract triggering key variables injected before the instantaneous time point from the operation log of the digital twin model based on the time point, wherein the triggering key variables include the external environmental stress value that causes a numerical jump; input the triggering key variables into a causal analysis model with embedded physical topology logic for deduction, and generate the chain reaction path of stress within the equipment. The generation of the chain reaction path of stress within the equipment includes: quantifying the state change amplitude of each circuit node on each conduction path; calculating the change ratio of the node electrical parameter values ​​after being affected by stress relative to the normal operating condition baseline value: ； in, For the calculated percentage change, This represents the nodal electrical parameter values ​​after being affected by stress. This represents the baseline value of the node under normal operating conditions; The main paths whose change rate exceeds the preset impact threshold are selected as the chain reaction paths that lead to equipment malfunctions.

2. The method for determining anomalies in metering equipment based on big data assessment according to claim 1, characterized in that, The multidimensional environmental stress data includes: Instantaneous impact and continuous vibration acceleration data acquired through a triaxial accelerometer; Environmental stress gradient change data obtained through temperature and humidity monitoring unit; Spatial electromagnetic field intensity monitoring data reflecting the stability of the microenvironment in which the equipment is located.

3. The method for determining anomalies in metering equipment based on big data assessment according to claim 1, characterized in that, The generated prediction performance degradation sequence includes: An incremental step evolution algorithm is applied to calculate the electrical performance deviation of internal components caused by multidimensional environmental stress in each time segment. The deviation value is obtained by accumulating various environmental stress components and their corresponding stress failure weights. By accumulating the offset values ​​over time, the expected measurement value drift curve of the metering equipment is obtained, which serves as a predictive performance degradation sequence.

4. The method for determining anomalies in metering equipment based on big data assessment according to claim 1, characterized in that, The generation of the actual performance state sequence includes: Synchronously acquire the internal reference level signal or self-test pulse frequency characteristics of the metering device; The internal reference level signal or self-test pulse frequency characteristics are associated and bound with the physical state data measured by external sensors based on the timestamp to form an enhanced real state sequence; When the warning signal is generated, the internal signal slice corresponding to the abnormal period is output synchronously to help determine the degree of physical damage to the equipment's metering function.

5. The method for determining anomalies in metering equipment based on big data assessment according to claim 1, characterized in that, The computational performance mismatch includes: Extract the spatial distance between the actual performance status feature values ​​and the predicted performance offset values ​​at each time segment; By introducing weighting coefficients that reflect the importance of different measurement accuracy levels, the spatial distance is weighted and averaged to generate the performance mismatch degree.

6. The method for determining anomalies in metering equipment based on big data assessment according to claim 1, characterized in that, The method further includes: Anomaly warning signals, performance mismatch, triggering key variables, and chain reaction paths are stored as structured cases in the case database; Machine learning algorithms were used to mine historical cases and analyze the correlation rules between environmental stress combinations and equipment electrical parameter degradation. The internal parameter matrix and physical response logic of the performance digital twin model are updated using the association rules.

7. A system for determining anomalies in metering equipment based on big data assessment, applied to the method for determining anomalies in metering equipment based on big data assessment as described in any one of claims 1-6, characterized in that, The system includes: Digital twin modeling module: used to acquire the structural characteristic parameters and initial electrical performance calibration data of the metering equipment to be monitored, and to establish a performance digital twin model reflecting the stress sensitivity of the internal components of the equipment based on the structural characteristic parameters; Multidimensional stress acquisition module: used to acquire multidimensional environmental stress data of the metering equipment in the transportation environment in real time, and inject the multidimensional environmental stress data as input variables into the performance digital twin model; Performance prediction evolution module: used to drive the performance digital twin model to run, simulate the internal parameter drift process of the metering equipment under the current cumulative environmental stress, and generate a predicted performance degradation sequence; Actual status monitoring module: used to synchronously collect real-time physical feedback signals of the metering equipment during transportation and generate an actual performance status sequence; Performance mismatch assessment module: used to compare the predicted performance degradation sequence with the actual performance state sequence within a unified performance evaluation space, and calculate the performance mismatch degree; Anomaly detection and early warning module: used to generate an early warning signal for abnormal equipment metering performance when the performance mismatch exceeds the stability threshold set according to the equipment accuracy level.

Images