Gravity measurement data field quality evaluation method

Abstract

The invention provides a gravity measurement data field quality evaluation method, which belongs to the technical field of gravity quality control, and comprises the following steps: establishing an initial instrument parameter database and an environmental parameter monitoring system, and tracking the state change of an instrument in real time by using a Kalman filtering drift monitoring model; a drift correction equation set and a time-frequency domain filtering algorithm are adopted to carry out real-time preprocessing and error compensation on original gravity data, a multi-dimensional index-based quality evaluation system is constructed to quantify a data quality grade, and an anomaly detection model of a Transform architecture is applied to identify abnormal data and generate quality early warning information. Finally, an on-site quality evaluation report containing data quality distribution, abnormal positions and processing suggestions is formed, real-time quality monitoring and accurate anomaly detection of the gravity measurement data are achieved, and the technical problems that gravity measurement data quality evaluation lags behind and is insufficient in accuracy are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of gravity quality control, and in particular relates to a method for on-site quality assessment of gravity measurement data. Background Art

[0002] Gravity measurement, as an important tool in geophysical exploration, is widely used in fields such as oil and gas exploration, mineral resource surveys, engineering geological surveys, and geological structure research. Traditional gravity measurement data quality control relies primarily on post-processing and manual judgment, assessing data reliability by comparing repeated measurements of benchmark points and statistically analyzing measurement errors. In existing technologies, gravity instruments experience zero drift and scale factor changes over long-term operation. Environmental factors such as temperature changes, vibration interference, and air pressure fluctuations can introduce systematic errors, resulting in a large amount of noise and outliers in the measurement data. Traditional quality control methods are unable to achieve real-time monitoring and accurate assessment. In other words, existing technologies suffer from the technical problem of delayed and insufficiently accurate gravity measurement data quality assessment. Summary of the Invention

[0003] In view of this, the present invention provides a method for on-site quality assessment of gravity measurement data, which can solve the technical problems in the prior art of delayed quality assessment of gravity measurement data and insufficient accuracy.

[0004] The present invention is implemented as follows: The present invention provides a method for field quality assessment of gravity measurement data, comprising: performing initial calibration on a gravimeter to obtain a zero offset and a scale factor to establish an initial instrument parameter database; deploying environmental parameter monitoring sensors in a measurement area to collect temperature, humidity, air pressure, and vibration acceleration parameter data in real time to form an environmental parameter time series; synchronously starting a drift monitoring model when starting a gravity measurement operation, and calculating the gravimeter zero drift and scale factor drift in real time by repeatedly measuring changes in gravity values ​​and environmental parameters at a reference point; performing real-time preprocessing on the collected raw gravity measurement data, using a drift compensation function to eliminate instrument systematic errors, simultaneously using a drift correction equation group to calculate an accurate drift compensation value, and using a time-frequency domain filtering algorithm to remove high-frequency noise interference; constructing a gravity data quality assessment index system to calculate data accuracy, consistency, stability, and reliability indicators for each measuring point, and quantifying the data quality level using a multi-dimensional quality comprehensive evaluation model; using a gravity anomaly detection model to identify outliers in the preprocessed gravity data, marking abnormal data points that exceed a quality threshold, and generating quality warning information; and generating a field quality assessment report for the gravity measurement data.

[0005] Among them, the drift correction equation group specifically includes a zero drift correction equation and a scale factor drift correction equation; the zero drift correction equation is used to calculate the time variation characteristics and environmental response characteristics of the gravimeter zero offset, and the input includes the reference point gravity value sequence, the environmental parameter time series, the measurement time interval, the instrument initial calibration parameters and the zero offset, and the output is the zero drift correction coefficient; the scale factor drift correction equation is used to calculate the nonlinear variation characteristics and temperature response characteristics of the gravimeter scale factor, and the input includes the scale factor drift, the temperature change gradient, the measurement time interval, the instrument initial calibration parameters and the environmental parameter time series, and the output is the scale factor drift correction coefficient.

[0006] The drift compensation function is specifically used to eliminate the zero offset and scale factor drift errors generated by the gravimeter during long-term operation. The input includes the reference point gravity value sequence, the environmental parameter time series, the measurement time interval and the instrument's initial calibration parameters. The output is the drift-compensated gravity measurement value.

[0007] Among them, the gravity anomaly detection model is specifically a sequence analysis model based on the Transformer architecture, which includes a multi-head attention mechanism to capture the spatiotemporal correlation of gravity data. The number of attention heads is dynamically determined based on three parameters: the complexity of the measurement area, the data sampling density, and the environmental noise level. When the geological structure of the measurement area is complex, the data sampling density is high, and the environmental noise level is low, the number of attention heads is increased to improve the accuracy of anomaly detection.

[0008] Among them, the steps for establishing the training data set of the gravity anomaly detection model are to collect gravity measurement data under different geological environments as normal samples, artificially inject various types of instrument failures, environmental interference and measurement errors to generate abnormal samples, label and classify all samples to form a training data set containing normal data and abnormal data, and expand the number of training samples through data enhancement technology to improve the generalization ability of the model.

[0009] The environmental parameter monitoring sensor refers to a multi-type sensor array installed in the measurement area, which is used to monitor the changes in environmental factors that affect the accuracy of gravity measurement in real time.

[0010] The drift monitoring model refers to a state estimation model based on the Kalman filter algorithm, which is used to estimate and predict the drift state of the gravimeter in real time.

[0011] The multi-dimensional quality comprehensive evaluation model refers to a comprehensive evaluation algorithm that integrates multiple quality indicators and outputs data quality grade classification results.

[0012] Among them, the attention head number adjustment function is specifically used to adjust the multi-head attention mechanism parameters of the gravity anomaly detection model. The attention adjustment value is calculated based on three data: measurement area complexity, data sampling density, and environmental noise level. When the attention adjustment value is in the range of 0 to 0.25, 4 attention heads are used and the attention range of each attention head is 256 time steps. When the attention adjustment value is in the range of 0.25 to 0.5, 6 attention heads are used and the attention range of each attention head is 128 time steps. When the attention adjustment value is in the range of 0.5 to 0.75, 8 attention heads are used and the attention range of each attention head is 64 time steps. When the attention adjustment value is in the range of 0.75 to 1, 12 attention heads are used and the attention range of each attention head is 32 time steps to adjust the multi-head attention mechanism parameters of the model.

[0013] Among them, the attention weight threshold of each attention head is dynamically set according to the variance of the gravity data sequence. When the variance of the gravity data sequence is less than 0.01mGal, the attention weight threshold is set to 0.8; when the variance of the gravity data sequence is between 0.01 and 0.05mGal, the attention weight threshold is set to 0.6; when the variance of the gravity data sequence is greater than 0.05mGal, the attention weight threshold is set to 0.4. The attention head uses a sliding window attention method to process the gravity data time series and screens important features according to the attention weight threshold.

[0014] Among them, the gravity anomaly detection model training steps specifically include using supervised learning methods to train model parameters, using the cross-entropy loss function to measure the difference between the model prediction results and the true labels, optimizing the model weight parameters through the backpropagation algorithm, using the validation set to evaluate the model performance and adjust the hyperparameters, and finally obtaining a trained model for gravity data anomaly detection.

[0015] The time-frequency domain filtering algorithm refers to a composite filtering method that combines low-pass filtering and band-stop filtering, and is used to remove high-frequency noise and frequency band interference in gravity data.

[0016] Among them, gravity measurement data includes three main categories: normal gravity data, boundary gravity data and abnormal gravity data. Normal gravity data is obtained through statistical analysis of measurement value distribution, accounting for 70% to 80% of the total data volume; boundary gravity data is obtained by setting quality threshold range identification, accounting for 15% to 20% of the total data volume; abnormal gravity data is obtained through gravity anomaly detection model identification, accounting for 5% to 10% of the total data volume.

[0017] Among them, the change in the proportion of different classified data will directly affect the accuracy and reliability of the final gravity field interpretation. When the proportion of abnormal gravity data exceeds 10%, the overall quality of the gravity measurement will be significantly reduced. Real-time quality monitoring and dynamic data screening mechanisms are used to reduce the negative impact of abnormal gravity data on the final results. At the same time, multiple verification and cross-checking methods are used to improve the control accuracy of the conversion rate and delay rate of boundary gravity data.

[0018] Among them, the zero drift correction coefficient refers to the correction parameter used to compensate for the time change of the gravimeter's zero offset; the scale factor drift correction coefficient refers to the correction parameter used to compensate for the nonlinear change of the gravimeter's scale factor drift; and the temperature change gradient refers to the rate of change parameter of the ambient temperature within the measurement time interval.

[0019] Among them, the field quality assessment report of gravity measurement data includes data quality level distribution, abnormal data location, recommended processing measures and comprehensive evaluation results of measurement quality.

[0020] The present invention establishes a drift monitoring model and a gravity anomaly detection model, combines real-time monitoring of environmental parameters with a multi-dimensional quality assessment index system, and achieves on-site real-time quality assessment and precise anomaly detection of gravity measurement data. The present invention uses a drift monitoring model based on Kalman filtering to track instrument status changes in real time, uses an anomaly detection model based on the Transformer architecture to accurately identify data anomalies, and effectively eliminates systematic errors and noise interference through a set of drift correction equations and a time-frequency domain filtering algorithm, significantly improving the timeliness and accuracy of data quality assessment. In summary, the present invention solves the technical problems of delayed and insufficiently accurate gravity measurement data quality assessment mentioned in the background technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0023] like Figure 1 FIG. 1 is a flow chart of a method for field quality assessment of gravity measurement data provided by the present invention, and the method comprises the following steps: S01. Before starting gravity measurement, the gravimeter is initially calibrated by measuring the reference point with known gravity value to obtain the gravimeter zero offset and scale factor, and establish the initial instrument parameter database; S02. Evenly distribute environmental parameter monitoring sensors in the measurement area to collect temperature, humidity, air pressure, and vibration acceleration parameter data in real time to form an environmental parameter time series; S03. When starting the gravity measurement operation, the drift monitoring model is started simultaneously. By repeatedly measuring the changes in the gravity value of the reference point and the changes in environmental parameters, the gravimeter zero drift and scale factor drift are calculated in real time. S04. Perform real-time preprocessing on the collected raw gravity measurement data, apply the drift compensation function to eliminate the instrument systematic error, calculate the accurate drift compensation value using the drift correction equation group, and use the time-frequency domain filtering algorithm to remove high-frequency noise interference; S05. Build a gravity data quality assessment index system, calculate the data accuracy, consistency, stability and reliability indicators of each measuring point, and quantify the data quality level through a multi-dimensional quality comprehensive evaluation model; S06. Use the gravity anomaly detection model to identify outliers in the pre-processed gravity data, mark abnormal data points that exceed the quality threshold, and generate quality warning information; S07. Generate an on-site quality assessment report for gravity measurement data, including data quality level distribution, abnormal data locations, recommended processing measures, and comprehensive measurement quality evaluation results.

[0024] The drift correction equation group includes a zero drift correction equation and a scale factor drift correction equation; the zero drift correction equation is used to calculate the time variation characteristics and environmental response characteristics of the gravimeter zero offset, and the input includes the reference point gravity value sequence, the environmental parameter time sequence, the measurement time interval, the instrument initial calibration parameters and the zero offset, and the output is the zero drift correction coefficient; the scale factor drift correction equation is used to calculate the nonlinear variation characteristics and temperature response characteristics of the gravimeter scale factor, and the input includes the scale factor drift, the temperature change gradient, the measurement time interval, the instrument initial calibration parameters and the environmental parameter time sequence, and the output is the scale factor drift correction coefficient.

[0025] The drift compensation function is used to eliminate zero offset and scale factor drift errors incurred by the gravimeter during long-term operation. Its inputs include a sequence of reference point gravity values, a time series of environmental parameters, a measurement interval, and initial instrument calibration parameters. The output is a drift-compensated gravity measurement. The gravity anomaly detection model is a sequence analysis model based on the Transformer architecture, incorporating a multi-head attention mechanism to capture the spatiotemporal correlations of gravity data. The number of attention heads is dynamically determined based on three parameters: the complexity of the measurement area, the data sampling density, and the ambient noise level. When the measurement area has complex geological structures, high data sampling density, and low ambient noise levels, the number of attention heads is increased to improve anomaly detection accuracy. The training dataset for the gravity anomaly detection model specifically involves collecting gravity measurement data from different geological environments as normal samples, artificially injecting various types of instrument failures, environmental interference, and measurement errors to generate abnormal samples, annotating and classifying all samples to form a training dataset containing both normal and abnormal data, and expanding the number of training samples through data augmentation techniques to improve model generalization. The gravity anomaly detection model training steps specifically include using a supervised learning method to train model parameters, using a cross-entropy loss function to measure the difference between the model prediction results and the true labels, optimizing the model weight parameters through a backpropagation algorithm, using a validation set to evaluate the model performance and adjust the hyperparameters, and finally obtaining a trained model for gravity data anomaly detection.

[0026] The environmental parameter monitoring sensor is a multi-type sensor array installed within the measurement area, used to monitor in real time changes in environmental factors that affect gravity measurement accuracy. The drift monitoring model is a state estimation model based on the Kalman filter algorithm, used to estimate and predict the gravimeter drift state in real time. The time-frequency domain filtering algorithm is a composite filtering method that combines low-pass filtering and band-stop filtering to remove high-frequency noise and frequency band interference from gravity data. The multi-dimensional quality comprehensive evaluation model is a comprehensive evaluation algorithm that integrates multiple quality indicators and outputs data quality classification results. The zero drift correction coefficient is a correction parameter used to compensate for time changes in the gravimeter zero offset. The scale factor drift correction coefficient is a correction parameter used to compensate for nonlinear changes in the gravimeter scale factor drift. The temperature change gradient is a parameter that measures the rate of change of the ambient temperature within a measurement time interval.

[0027] The attention head number adjustment function is used to adjust the multi-head attention mechanism parameters of the gravity anomaly detection model. The attention adjustment value is calculated based on three data: measurement area complexity, data sampling density, and environmental noise level. When the attention adjustment value is in the range of 0 to 0.25, 4 attention heads are used and the attention range of each attention head is 256 time steps. When the attention adjustment value is in the range of 0.25 to 0.5, 6 attention heads are used and the attention range of each attention head is 128 time steps. When the attention adjustment value is in the range of 0.5 to 0.75, 8 attention heads are used and the attention range of each attention head is 64 time steps. When the attention adjustment value is in the range of When the gravity data variance is in the range of 0.75 to 1, 12 attention heads are used and the attention range of each attention head is 32 time steps. The multi-head attention mechanism parameters of the model are adjusted. The attention weight threshold of each attention head is dynamically set according to the variance of the gravity data sequence. When the variance of the gravity data sequence is less than 0.01mGal, the attention weight threshold is set to 0.8. When the variance of the gravity data sequence is between 0.01 and 0.05mGal, the attention weight threshold is set to 0.6. When the variance of the gravity data sequence is greater than 0.05mGal, the attention weight threshold is set to 0.4. The attention head uses a sliding window attention method to process the gravity data time series and screens important features according to the attention weight threshold.

[0028] Gravity measurement data consists of three main categories: normal gravity data, boundary gravity data, and anomalous gravity data. Normal gravity data, acquired through statistical analysis of measurement value distribution, accounts for 70% to 80% of the total data volume and has a stable positive impact on measurement results. Boundary gravity data, acquired through the use of a set quality threshold range, accounts for 15% to 20% of the total data volume, but its contribution fluctuates with environmental conditions. Anomalous gravity data, acquired through the use of gravity anomaly detection models, accounts for 5% to 10% of the total data volume and can negatively impact measurement accuracy. The proportion of data from different categories directly affects the accuracy and reliability of the final gravity field interpretation. When the proportion of anomalous gravity data exceeds 10%, the overall quality of the gravity measurement is significantly reduced. Real-time quality monitoring and dynamic data screening mechanisms are used to mitigate the negative impact of anomalous gravity data on the final results. Multiple validation and cross-verification methods are also used to improve the control accuracy of the transition rate and delay rate of boundary gravity data.

[0029] The specific implementation of the above steps is described in detail below. The specific implementation of step S01 is to establish the instrument basic parameter database through the precise control of the initial calibration program. First, select no less than 3 national or regional reference points with known gravity values. The accuracy of the reference point gravity value should reach m / The distance between the benchmark points should be kept above 1000m to ensure spatial representativeness. Gravity values ​​are measured continuously for no less than 10 times at each benchmark point, with the measurement interval set to 5 minutes. The least squares method is used to fit the linear relationship between the gravimeter reading and the true gravity value. The zero offset is calculated by regression analysis, which reflects the systematic deviation of the instrument. The scale factor is determined by the slope coefficient, which reflects the sensitivity characteristics of the instrument. The established initial instrument parameter database contains key parameters such as zero offset, scale factor, linearity error, repeatability error, etc., which provide a benchmark reference for subsequent drift correction. The purpose of this step is to eliminate the inherent systematic error of the instrument, establish an accurate measurement benchmark, and ensure the reliability and traceability of subsequent gravity measurement data.

[0030] The specific implementation method of step S02 is to construct a multi-parameter environmental monitoring network to realize real-time environmental data collection. The environmental parameter monitoring sensor array is evenly deployed in the measurement area according to the grid layout principle. The sensor spacing is set to 200m to 500m, and adjusted according to the complexity of the terrain in the measurement area. The temperature sensor adopts a platinum resistance thermometer, the measurement accuracy should reach 0.01℃, and the sampling frequency is set to 1Hz. The humidity sensor adopts a capacitive hygrometer, the measurement accuracy should reach 0.5% relative humidity, and the sampling frequency is set to 0.5Hz. The pressure sensor adopts a piezoresistive barometer, the measurement accuracy should reach 0.1hPa, and the sampling frequency is set to 0.2Hz. The vibration acceleration sensor adopts a piezoelectric accelerometer, and the measurement accuracy should reach m / The sampling frequency is set to 100 Hz to capture high-frequency vibration signals. All sensor data is collected in real time via a wireless transmission network to a data acquisition center, forming a time series of synchronized environmental parameters. This step aims to comprehensively monitor changes in environmental factors that affect gravity measurement accuracy and provide environmental context for drift correction and anomaly detection.

[0031] The specific implementation of step S03 is to construct a dynamic drift monitoring model based on the Kalman filter algorithm to achieve real-time drift estimation. The drift monitoring model uses state-space equations to describe the gravimeter's drift behavior. The state variables include zero drift, scale factor drift, and their rate of change. The model prediction equation is established based on the instrument's physical characteristics, and the observation equation is established using repeated measurements of benchmark points. The Kalman filter process includes a prediction step and an update step. The prediction step predicts the current drift state based on the previous state and the system model. The update step uses the benchmark point measurements to correct the prediction results. The benchmark point repeated measurement frequency is set to once every 30 minutes, with the average of five consecutive measurements taken each time. The process noise covariance matrix is ​​set according to the instrument's technical specifications, and the observation noise covariance matrix is ​​set according to the benchmark point measurement accuracy. The optimal estimates of zero drift and scale factor drift and their uncertainties are obtained through recursive calculation. The purpose of this step is to track changes in the instrument's drift state in real time and provide accurate correction parameters for drift compensation.

[0032] The specific implementation of step S04 utilizes a multi-level data processing algorithm to eliminate systematic errors and random noise. The drift compensation function, based on the drift parameters obtained in step S03, uses linear interpolation to calculate the zero offset compensation value and scale factor compensation value at each measurement point. The zero drift correction equation inputs include the reference point gravity value sequence, the environmental parameter time series, the measurement interval, and the instrument's initial calibration parameters. A multivariate regression analysis is used to establish a relationship model between the zero offset and time and environmental parameters, outputting the zero drift correction coefficient. The scale factor drift correction equation prioritizes temperature response characteristics, establishing a nonlinear relationship model between the scale factor and the temperature gradient, and outputting the scale factor drift correction coefficient. The time-frequency domain filtering algorithm first utilizes a Butterworth low-pass filter to remove high-frequency noise above 0.1 Hz, with the cutoff frequency set based on the gravity signal's spectral characteristics. A notch filter is then used to remove 50 Hz power frequency interference and its harmonics. The filter parameters are determined through spectral analysis and signal-to-noise ratio optimization to ensure noise removal while maintaining the integrity of the effective gravity signal. The goal of this step is to obtain high-quality gravity measurement data, providing a reliable data foundation for quality assessment.

[0033] The specific implementation of step S05 is to build a multi-dimensional quality assessment system to quantify the data quality level. The data accuracy index is calculated by repeated measurement standard deviation, and the accuracy threshold is set to 0.02mGal. Data exceeding this threshold is marked as low-precision data. The consistency index is evaluated by the gradient change of the gravity value of adjacent measuring points. Data with a gradient change of more than 0.05mGal / m are marked as consistency anomalies. The stability index is evaluated by time series variance analysis. The variance exceeds 0.01 Data that is not reliable is marked as unstable. Reliability indicators are evaluated by comparing the results of multiple independent measurements. Data with a difference of more than 0.03 mGal is marked as unreliable. The multi-dimensional quality comprehensive evaluation model uses a weighted comprehensive evaluation method. The weight of each indicator is determined by the hierarchical analysis method, with a weight of 0.4 for the accuracy indicator, 0.3 for the consistency indicator, 0.2 for the stability indicator, and 0.1 for the reliability indicator. After calculation, the comprehensive quality score is divided into four levels: excellent, good, qualified, and unqualified, with score thresholds of 90, 80, and 70 points, respectively. The purpose of this step is to comprehensively evaluate the quality level of gravity data and provide a quantitative basis for data screening and anomaly identification.

[0034] The specific implementation of step S06 involves identifying anomalous data using a gravity anomaly detection model based on the Transformer architecture. The anomaly detection model employs an encoder-decoder structure. The encoder comprises a multi-head attention mechanism and a feedforward neural network, while the decoder outputs anomaly probabilities. The multi-head attention mechanism is dynamically adjusted using a function to adjust the number of attention heads. The complexity of the measurement area is quantified by geological structure density and topographic relief. The data sampling density is calculated by the number of measurement points per unit area, and the ambient noise level is assessed by the standard deviation of environmental parameters. After normalizing these three parameters, an attention adjustment value is calculated. The number of attention heads and the attention range are determined based on the adjustment value range. The anomaly detection threshold is determined by optimizing the receiver operating characteristic curve, with a false positive rate of less than 5% and a true positive rate of at least 95%. Detected anomalous data points undergo secondary verification, with a comprehensive assessment based on environmental parameter anomalies and instrument status anomalies. Quality warning information includes the location of the anomalous data, the type of anomaly, the severity of the anomaly, and recommended actions. The purpose of this step is to accurately identify anomalous data that impacts measurement quality, providing intelligent technical support for data quality control.

[0035] The specific implementation method of step S07 is to generate a comprehensive quality assessment report to provide decision support information. The report generation module automatically summarizes the analysis results of the aforementioned steps, and the data quality level distribution is obtained by statistical analysis to obtain the proportion of data of each level and the spatial distribution characteristics. The location of abnormal data is visualized through the geographic information system, including the spatial coordinates of the abnormal data, the abnormal type identification and the abnormal degree classification. Recommended treatment measures are automatically generated according to the type and degree of abnormal data, including re-measurement suggestions, instrument inspection suggestions, environmental condition improvement suggestions, etc. The comprehensive evaluation results of measurement quality are calculated by a weighted scoring method, taking into account factors such as data integrity, accuracy level, and abnormality ratio. The report format adopts a standardized template, which includes an executive summary, detailed analysis, chart display, conclusion and suggestions, etc. After the report is generated, it is automatically stored and can be exported to multiple formats for subsequent use. The purpose of this step is to provide comprehensive quality assessment information to support measurement operation quality control and data processing decisions.

[0036] The gravity anomaly detection model utilizes a deep learning model based on the Transformer architecture. Its structure consists of an input embedding layer, a position encoding layer, a multi-layer encoder, and a classification output layer. The input embedding layer converts the gravity data sequence into a high-dimensional vector representation, with an embedding dimension of 512. The position encoding layer uses sine-cosine positional encoding to add temporal position information to the input sequence. The encoder layer has six layers, each consisting of a multi-head attention sublayer and a feedforward neural network sublayer, with residual connections and layer normalization used between the sublayers. In the multi-head attention mechanism, the number of attention heads is dynamically determined by an attention head number adjustment function, with each attention head having a dimension of 64. The feedforward neural network consists of two linear transformation layers, with the ReLU function used as the intermediate activation function, and the hidden layer dimension is 2048. The classification output layer uses a fully connected layer with a sigmoid activation function to output anomaly probability values.

[0037] The training dataset construction involves four main steps: normal sample collection, anomalous sample generation, sample annotation, and data augmentation. Normal samples were collected from historical gravity surveys in diverse geological settings, encompassing gravity data from sedimentary, igneous, and metamorphic rock regions, ensuring geological representativeness. Normal samples were required to achieve a measurement accuracy of at least 0.01 mGal, data integrity exceeding 95%, and be quality-verified to be free of significant anomalies. Anomalous samples were generated through artificial injection, with various anomaly patterns including instrument zero-point mutation, scale factor anomaly, temperature shock response, vibration interference, and electromagnetic interference. Zero-point mutation anomalies were simulated by superimposing a step signal on the normal data, with the mutation amplitude set between 0.05 mGal and 0.5 mGal. Scale factor anomalies were simulated by varying the data scaling, with the anomaly severity set between 5% and 20%. Environmental interference anomalies were simulated by superimposing sinusoidal signals of varying frequencies and amplitudes. Sample annotation utilized expert experience combined with statistical analysis to establish anomaly determination criteria, and all samples were labeled using binary classification. Data enhancement technology includes time series transformation, noise injection, data shift and other methods, which expand the number of training samples to more than 5 times the original samples and improve the model's adaptability to different measurement conditions.

[0038] The key technical concepts of this invention are mainly reflected in three aspects: real-time drift monitoring and compensation, Transformer-based anomaly detection, and a multi-dimensional quality assessment system. The real-time drift monitoring and compensation technology uses the Kalman filter algorithm to establish a dynamic state estimation model, capable of tracking the time-varying characteristics of the gravimeter zero offset and scale factor in real time. Compared with traditional linear drift assumption methods, this technology can capture the nonlinear variation of instrument drift and environmental response characteristics, significantly improving the accuracy and real-time performance of drift compensation, and effectively reducing the systematic errors caused by instrument drift. The Transformer-based anomaly detection technology uses a multi-head attention mechanism to capture the spatiotemporal correlations of gravity data. By dynamically adjusting the number of attention heads, it adapts to measurement environments of varying complexity. Compared with traditional statistical threshold detection methods, this technology has stronger feature extraction and pattern recognition capabilities, can identify complex anomaly patterns and potential data quality issues, and significantly improve the accuracy and recall of anomaly detection. The multi-dimensional quality assessment system comprehensively evaluates data quality from multiple perspectives: accuracy, consistency, stability, and reliability. Compared with single-metric evaluation methods, this system can more comprehensively reflect the data quality status and provide a more scientific basis for data screening and processing decisions.

[0039] The synergy of these three key technical approaches forms a complete solution for field quality assessment of gravity measurement data. Real-time drift monitoring provides accurate baseline data for anomaly detection, and anomaly detection results provide important input for quality assessment. The quality assessment results, in turn, guide the optimization and adjustment of drift monitoring parameters. These three elements form a closed-loop feedback system. Compared to traditional offline post-processing methods, this collaborative technical solution enables real-time monitoring and dynamic optimization of gravity measurement data quality, significantly improving the efficiency and data quality of gravity measurement operations and providing advanced technical support for high-precision gravity exploration.

[0040] It should be noted that the present invention also solves the following technical problem: the technical problem of inaccurate instrument drift compensation during gravity measurement. In traditional gravity measurement, instrument drift compensation mainly relies on simple linear interpolation methods or empirical formulas, which cannot accurately reflect the complex change rules of zero offset and scale factor. In particular, in the long-term continuous measurement process, instrument drift presents nonlinear characteristics and is significantly affected by environmental factors. The present invention establishes a zero drift correction equation and a scale factor drift correction equation, combines the environmental parameter time series and the instrument initial calibration parameters, and can accurately calculate the drift compensation coefficient to achieve accurate correction of the instrument's systematic error. At the same time, the drift compensation function comprehensively considers the changes in the reference point gravity value sequence and the environmental response characteristics. Through real-time calculation and dynamic adjustment, it significantly improves the accuracy and adaptability of drift compensation, effectively ensuring the reliability and consistency of gravity measurement data.

[0041] Specifically, the principle behind this invention is that the fundamental reason it can address the lagging and inaccurate quality assessment of gravity measurement data lies in the establishment of a complete real-time quality monitoring system and precise data processing algorithms. First, by deploying an array of environmental parameter monitoring sensors throughout the measurement area, key environmental factors influencing gravity measurements are collected in real time, providing accurate environmental context information for drift compensation and anomaly detection. Second, a drift monitoring model constructed based on the Kalman filter algorithm dynamically tracks the temporal variations of the gravimeter's zero offset and scale factor, enabling real-time monitoring and compensation of instrument drift through state estimation and prediction mechanisms. Third, an anomaly detection model based on a Transformer architecture utilizes a multi-head attention mechanism to capture the spatiotemporal correlations of gravity data and identifies complex anomaly patterns through deep learning methods, demonstrating superior nonlinear feature extraction capabilities compared to traditional statistical methods. Furthermore, a set of drift correction equations, combining zero drift correction and scale factor drift correction, accurately calculates compensation coefficients for various types of drift, effectively eliminating the impact of systematic errors on measurement accuracy. Finally, the multi-dimensional comprehensive quality evaluation model integrates multiple indicators such as data accuracy, consistency, stability and reliability, and achieves objective grading of data quality through quantitative evaluation methods to ensure the scientificity and accuracy of the quality evaluation results.

[0042] A specific embodiment 1 of the present invention is provided below. The specific implementation of each step in this embodiment 1 is described in detail as follows.

[0043] The specific implementation of step S01 is to establish the instrument basic parameter database through the precise control of the initial calibration procedure. First, select no less than 3 national or regional reference points with known gravity values. The accuracy of the reference point gravity value should reach m / The distance between the reference points should be kept at least 1000m to ensure spatial representativeness. Gravity values ​​should be measured continuously for no less than 10 times at each reference point, with a measurement interval of 5 minutes. The linear relationship between the gravimeter reading and the true gravity value is obtained by least square fitting. The zero offset is calculated by regression analysis. , the offset reflects the systematic deviation of the instrument, the unit is mGal, the scale factor The slope coefficient is determined to reflect the sensitivity of the instrument and is a dimensionless parameter. The initial instrument parameter database established includes zero offset, scale factor, linearity error, Repeatability error And other key parameters, among which and The unit is mGal, which provides a reference for subsequent drift correction. The purpose of this step is to eliminate the inherent systematic errors of the instrument, establish an accurate measurement benchmark, and ensure the reliability and traceability of subsequent gravity measurement data.

[0044] The specific implementation of step S02 is to build a multi-parameter environmental monitoring network to realize real-time environmental data collection. The environmental parameter monitoring sensor array is evenly deployed in the measurement area according to the grid layout principle, and the sensor spacing is The range is set to 200m to 500m and can be adjusted according to the complexity of the terrain in the measurement area. The temperature sensor uses a platinum resistance thermometer with a measurement accuracy of 0.01°C and a sampling frequency of The humidity sensor uses a capacitive hygrometer, and the measurement accuracy should reach 0.5% relative humidity. The sampling frequency Set to 0.5Hz. The pressure sensor uses a piezoresistive barometer, the measurement accuracy should reach 0.1hPa, and the sampling frequency Set to 0.2Hz. The vibration acceleration sensor uses a piezoelectric accelerometer, and the measurement accuracy should reach m / , sampling frequency The frequency is set to 100 Hz to capture high-frequency vibration signals. All sensor data is collected in real time via a wireless transmission network to a data acquisition center, forming a synchronized time series of environmental parameters. This step aims to comprehensively monitor changes in environmental factors that affect gravity measurement accuracy, providing environmental context for drift correction and anomaly detection.

[0045] The specific implementation of step S03 is to build a dynamic drift monitoring model based on the Kalman filter algorithm to achieve real-time drift estimation. The drift monitoring model uses state space equations to describe the drift behavior of the gravimeter. The state variables include the zero drift amount , scale factor drift and its rate of change and The model prediction equation is established based on the physical characteristics of the instrument, and the observation equation is established through repeated measurement data of the reference point, where the zero drift The unit is mGal, the scale factor drift is a dimensionless parameter. The Kalman filter process includes a prediction step and an update step. The prediction step predicts the current drift state based on the previous state and the system model. The update step uses the reference point measurement value to correct the prediction result. The reference point remeasurement frequency is set to once every 30 minutes, and each time the average value is taken for 5 consecutive measurements. The process noise covariance matrix According to the instrument technical specifications, the observation noise covariance matrix According to the measurement accuracy of the reference point, and All are symmetric positive definite matrices. The optimal estimates of the zero drift and scale factor drift, along with their uncertainties, are obtained through recursive calculation. The purpose of this step is to track instrument drift state changes in real time and provide accurate correction parameters for drift compensation.

[0046] The specific implementation of step S04 is to use a multi-level data processing algorithm to eliminate systematic errors and random noise. The drift compensation function uses a linear interpolation method to calculate the zero offset compensation value and scale factor compensation value at each measuring point based on the drift parameters obtained in step S03. The zero drift correction equation is expressed as ,in is the zero drift at time t, in mGal. is the constant term regression coefficient, the unit is mGal, is the time term regression coefficient, in mGal / h, is the temperature term regression coefficient, the unit is mGal / ℃, is the regression coefficient of the air pressure term, in mGal / hPa, is the regression coefficient of humidity term, unit is mGal / %. is the measurement time, in h, is the temperature in °C, is the air pressure, in hPa, The unit is humidity, %. The relationship model between zero offset and time and environmental parameters is established through multiple regression analysis, and the zero drift correction coefficient is output. , which is a dimensionless parameter. The scale factor drift correction equation is expressed as ,in is the scale factor drift at time t, which is a dimensionless parameter. is the constant term regression coefficient, dimensionless, is the temperature gradient regression coefficient, in h / ℃, is the temperature square regression coefficient, the unit is 1 / ℃², is the time term regression coefficient, the unit is 1 / h, is the temperature change gradient, in ℃ / h. Focus on the temperature response characteristics, establish a nonlinear relationship model between the scale factor and the temperature gradient, and output the scale factor drift correction coefficient , which is a dimensionless parameter. The drift compensation function is expressed as ,in is the gravity measurement value after drift compensation, is the raw gravity measurement value, all in milligal. This function is used to eliminate zero offset and scale factor drift errors generated by the gravimeter during long-term operation. The time-frequency domain filtering algorithm first uses a Butterworth low-pass filter to remove high-frequency noise above 0.1 Hz. The cutoff frequency is set based on the spectral characteristics of the gravity signal. A notch filter is then used to remove 50 Hz power frequency interference and its harmonic components. The filter parameters are determined through spectrum analysis and signal-to-noise ratio optimization to ensure that the integrity of the effective gravity signal is maintained while removing noise. The purpose of this step is to obtain high-quality gravity measurement data and provide a reliable data foundation for quality assessment.

[0047] The specific implementation of step S05 is to build a multi-dimensional quality assessment system to quantify the data quality level. The data accuracy index is measured by repeated measurement standard deviation. Calculation, precision threshold The value is set to 0.02mGal. Data exceeding this threshold is marked as low-precision data. The consistency index is calculated by the gradient change of gravity values ​​at adjacent measuring points. Evaluation, the data with gradient changes exceeding 0.05mGal / m are marked as consistency anomalies, where The unit is mGal / m, the consistency threshold The stability index is 0.05mGal / m. Analysis and evaluation, variance exceeds 0.01 The data is marked as unstable data, and the stability threshold 0.01 The reliability index is evaluated by comparing the results of multiple independent measurements. Data exceeding 0.03mGal is marked as unreliable data, and the reliability threshold The multi-dimensional quality comprehensive evaluation model adopts a weighted comprehensive evaluation method, and the comprehensive quality score Expressed as ,in to is the weight coefficient of each indicator, and , 、 、 、 They are the scores of accuracy, consistency, stability and reliability, with the value range of 0 to 100. The weight of each indicator is determined by the hierarchical analysis method. , consistency index weight , stability index weight , reliability index weight After calculating the comprehensive quality score, the data is divided into four levels: excellent, good, qualified, and unqualified, with score thresholds of 90, 80, and 70 points, respectively. The purpose of this step is to comprehensively evaluate the quality level of gravity data and provide a quantitative basis for data screening and anomaly identification.

[0048] The specific implementation of step S06 is to use the gravity anomaly detection model based on the Transformer architecture to identify abnormal data. The anomaly detection model adopts an encoder-decoder structure. The encoder part contains a multi-head attention mechanism and a feedforward neural network, and the decoder part outputs the abnormal probability. , the probability value ranges from 0 to 1, indicating the possibility of data anomaly. The multi-head attention mechanism is dynamically adjusted by the attention head number adjustment function, and the attention adjustment value Expressed as ,in 、 、 is the weight coefficient, and , To measure the regional complexity index, is the data sampling density index, is the environmental noise level index. All three parameters are dimensionless normalized indices with a value range of 0 to 1. The complexity of the measurement area is quantified by the geological structure density and terrain relief. The data sampling density is calculated by the number of measurement points per unit area. The environmental noise level is evaluated by the standard deviation of the environmental parameters. After the three parameters are normalized, the attention adjustment value is calculated, and the number of attention heads is determined according to the adjustment value range. and attention span ,in is a positive integer, is the number of time steps. Anomaly detection threshold Through receiver operating characteristic curve optimization, the false positive rate is controlled within 5%, and the true positive rate is required to reach above 95%. Detected abnormal data points undergo secondary verification, and a comprehensive assessment is made based on abnormal environmental parameters and instrument status. Quality warning information includes the abnormal data location, type, severity, and action recommendations. This step aims to accurately identify abnormal data that affects measurement quality and provide intelligent technical support for data quality control.

[0049] The specific implementation of step S07 is to generate a comprehensive quality assessment report to provide decision support information. The report generation module automatically summarizes the analysis results of the above steps, and the data quality level distribution is obtained through statistical analysis to obtain the proportion of data of each level and spatial distribution characteristics. The location of abnormal data is visualized through the geographic information system, including the spatial coordinates of the abnormal data. , exception type identification and abnormality level grading ,in is the coordinate of the measuring point, Encode the exception type, The abnormality level is 1 to 5. Suggested treatment measures are automatically generated based on the type and degree of abnormal data, including re-measurement suggestions, instrument inspection suggestions, environmental condition improvement suggestions, etc. The comprehensive evaluation results of measurement quality are calculated using a weighted scoring method, taking into account data integrity. , precision level , abnormal proportion Among these factors and is a percentage value, The report format uses a standardized template and includes an executive summary, detailed analysis, graphical presentation, and conclusions. Once generated, the report is automatically stored and can be exported to various formats for subsequent use. This step aims to provide comprehensive quality assessment information to support measurement quality control and data processing decisions.

[0050] The gravity anomaly detection model uses a deep learning model based on the Transformer architecture. The specific structure includes an input embedding layer, a position encoding layer, a multi-layer encoder, and a classification output layer. The input embedding layer converts the gravity data sequence into a high-dimensional vector representation, and the embedding dimension is set to 512 dimensions. The position encoding uses the sine-cosine function, and the position encoding function is expressed as and ,in is the position index, ranging from 0 to the sequence length minus 1, Is the dimension index, ranging from 0 to minus 1, is the embedding dimension, where , adding temporal position information to the input sequence. The number of encoder layers is set to 6, each layer contains a multi-head attention sublayer and a feedforward neural network sublayer, and residual connections and layer normalization are used between sublayers. In the multi-head attention mechanism, the attention weight calculation function is expressed as ,in 、 、 They are query matrix, key matrix and value matrix respectively, and their dimensions are all sequence length multiplied by , is the key vector dimension, where The number of attention heads is dynamically determined by the attention head number adjustment function, and the dimension of each attention head is 64. The feedforward neural network contains two linear transformation layers, the intermediate activation function uses the ReLU function, and the hidden layer dimension is 2048. The classification output layer uses a fully connected layer and a Sigmoid activation function to output anomaly probability values. The feedforward neural network contains two linear transformation layers, the intermediate activation function uses the ReLU function, and the hidden layer dimension is 2048. The classification output layer uses a fully connected layer and a Sigmoid activation function to output anomaly probability values.

[0051] The establishment of the training data set includes four main steps: normal sample collection, abnormal sample generation, sample annotation and data enhancement. Normal sample collection is obtained from historical gravity measurement projects in different geological environments, including gravity data of various geological backgrounds such as sedimentary rock areas, igneous rock areas, metamorphic rock areas, etc., to ensure the geological representativeness of the samples. The collected normal samples require a measurement accuracy of more than 0.01mGal, data integrity of more than 95%, and quality inspection to confirm that there are no obvious abnormalities. Abnormal samples are generated by artificial injection, and the types of injected anomalies include instrument zero-position mutation, scale factor anomaly, temperature shock response, vibration interference, electromagnetic interference and other abnormal modes. The zero-position mutation anomaly is simulated by superimposing a step signal on the normal data, and the mutation amplitude is The scale factor is set to 0.05mGal to 0.5mGal. The scale factor anomaly is simulated by changing the data scaling ratio. Set to 5% to 20%. Environmental interference anomalies are generated by superimposing different frequencies. and amplitude The sine wave signal simulation is The range is 0.01Hz to 10Hz, The range is 0.01mGal to 0.1mGal. Sample labeling uses expert experience combined with statistical analysis to establish anomaly determination criteria and assign binary classification labels to all samples. Data augmentation techniques, including time series transformation, noise injection, and data shifting, expand the number of training samples to more than five times the original number, improving the model's adaptability to varying measurement conditions.

[0052] In order to better understand and implement the present invention, the following provides a specific application scenario of the present invention, Example 2: A technical team performs a seabed gravity measurement task in the northern slope area of ​​the South China Sea, and the measurement area is about 500 The water depth ranges from 200m to 1500m, and the seabed topography is complex and varied, including various geomorphic units such as submarine canyons, seamounts, and sedimentary plains. The geological structure in this area is active, with multiple fault zones and magmatic intrusions, which have a significant impact on the gravity field. The technical team used the field quality assessment method for gravity measurement data of this invention and deployed 120 gravity measurement points on the seabed with a spacing of 2000m. The team planned to complete a 30-day continuous gravity observation operation.

[0053] Before the gravity measurement began, the technical team selected three seabed benchmarks with known gravity values ​​for initial calibration. The accuracy of the benchmark gravity values ​​reached m / The distance between the reference points is kept above 1200m. The gravity value is measured 15 times continuously at each reference point, with the measurement interval set to 5 minutes. The linear relationship between the gravimeter reading and the actual gravity value is obtained by least square fitting. The initial zero offset is calculated 0.025mGal, scale factor The linearity error is 0.9987. The repeatability error is 0.008mGal. The initial instrument parameter database established provides an accurate benchmark reference for subsequent drift correction.

[0054] The technical team deployed 75 environmental parameter monitoring sensors at 400m intervals within the measurement area to build a multi-parameter environmental monitoring network. The temperature sensor measurement accuracy reached 0.008℃, and the sampling frequency Set to 1.2Hz. The humidity sensor measurement accuracy reaches 0.3% relative humidity, the sampling frequency Set to 0.6Hz. The pressure sensor measurement accuracy reaches 0.08hPa, and the sampling frequency Set to 0.25Hz. The vibration acceleration sensor measurement accuracy reaches m / , sampling frequency The frequency is set to 120 Hz. All sensor data are transmitted in real time to the sea surface data collection center via submarine optical cables, forming a time series of environmental parameters that are synchronized.

[0055] During the formal measurement process, the technical team launched a drift monitoring model based on the Kalman filter algorithm. The model state variables include zero drift , scale factor drift and its rate of change and The frequency of repeated measurements of the benchmark points was set to once every 25 minutes, and the average value was obtained by 6 consecutive measurements each time. and the observation noise covariance matrix Determined based on the instrument's technical specifications and the benchmark's measurement accuracy. After 48 hours of continuous monitoring, real-time estimates of zero drift and scale factor drift were obtained, as shown in Table 1: Table 1 Gravity meter drift parameter monitoring results

[0056] The technical team uses a multi-level data processing algorithm to pre-process the raw gravity measurement data in real time. The regression coefficient of the zero drift correction equation is determined by multivariate regression analysis, where 0.018mGal, for mGal / h, for mGal / ℃, for mGal / hPa, for mGal / %. In the regression coefficient of the scale factor drift correction equation, for , for h / ℃, for / ℃², for The drift compensation function eliminates the systematic error of the instrument. A Butterworth low-pass filter is used to remove high-frequency noise above 0.12 Hz, and a notch filter is used to remove 50 Hz power frequency interference.

[0057] The technical team has built a multi-dimensional quality assessment system to calculate the data quality index of each measurement point. The data accuracy index is calculated by repeated measurement standard deviation. Calculation, precision threshold Set to 0.018mGal. The consistency index is calculated by the gradient change of gravity values ​​at adjacent measuring points. Evaluation, consistency threshold Set to 0.055mGal / m. The stability index is calculated by time series variance. Analysis, stability threshold Set to 0.012 The reliability index is compared with multiple independent measurement results, and the reliability threshold The value was set to 0.025 mGal. The quality assessment results are shown in Table 2: Table 2 Statistical results of gravity data quality assessment

[0058] The technical team uses a gravity anomaly detection model based on the Transformer architecture to identify abnormal data. The attention adjustment value is calculated based on three parameters: the complexity of the geological structure of the measurement area, the data sampling density, and the environmental noise level. is 0.68, determining the number of attention heads There are 8, and the attention range of each attention head is The model inputs a 512-dimensional gravity data sequence and outputs anomaly probability after processing it through a 6-layer encoder. Anomaly detection threshold The receiver operating characteristic curve was optimized to 0.75, the false positive rate was controlled at 4.2%, and the true positive rate reached 96.8%. The anomaly detection results are shown in Table 3: Table 3 Gravity anomaly detection model recognition results

[0059] The technical team generated a comprehensive quality assessment report, which included data quality level distribution, abnormal data location, recommended treatment measures and comprehensive measurement quality evaluation results. , exception type identification and abnormality level grading Visualization is performed through a geographic information system. Comprehensive evaluation of measurement quality takes into account data integrity. The accuracy level is 97.5%. Excellent grade, abnormal ratio The overall quality score is 2.5%. It scored 92.3 points and the overall measurement quality was rated as excellent.

[0060] Throughout the 30-day survey, gravity data classification statistics showed that normal gravity data accounted for 74.2% of the total data volume, boundary gravity data accounted for 23.3%, and abnormal gravity data accounted for 2.5%. This abnormal gravity data ratio was well below the 10% warning line, indicating effective measurement quality control. Real-time quality monitoring and dynamic data screening mechanisms effectively reduced the negative impact of abnormal gravity data on the final results. Multiple verification and cross-checking methods were used to improve the accuracy of control over the conversion rate and latency of boundary gravity data.

[0061] It should be noted that the detailed explanations of the variables involved in the present invention are shown in Tables 4 and 5.

[0062] Table 4 Variable Explanation Table (Part 1)

[0063] Table 5 Variable Explanation Table (Part II)

[0064] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A method for field quality assessment of gravity measurement data, characterized in that: include: Perform initial calibration on the gravimeter to obtain the zero offset and scale factor to establish the initial instrument parameter database; deploy environmental parameter monitoring sensors in the measurement area to collect temperature, humidity, air pressure, and vibration acceleration parameter data in real time to form an environmental parameter time series; When starting a gravity measurement operation, the drift monitoring model is simultaneously activated. The gravimeter zero drift and scale factor drift are calculated in real time by repeatedly measuring changes in the gravity value of the benchmark point and changes in environmental parameters. The collected raw gravity measurement data are preprocessed in real time, and the drift compensation function is used to eliminate the instrument's systematic errors. At the same time, the drift correction equation group is used to calculate the accurate drift compensation value, and the time-frequency domain filtering algorithm is used to remove high-frequency noise interference. A gravity data quality assessment index system is constructed to calculate the data accuracy, consistency, stability and reliability indicators of each measuring point, and the data quality level is quantified through a multi-dimensional quality comprehensive evaluation model. Use the gravity anomaly detection model to identify outliers in the preprocessed gravity data, mark abnormal data points that exceed the quality threshold, and generate quality warning information; Generate field quality assessment reports for gravity measurement data.

2. The method for on-site quality assessment of gravity measurement data according to claim 1, characterized in that: The drift correction equation group specifically includes a zero drift correction equation and a scale factor drift correction equation; the zero drift correction equation is used to calculate the time variation characteristics and environmental response characteristics of the gravimeter zero offset, and the input includes a reference point gravity value sequence, an environmental parameter time sequence, a measurement time interval, an instrument initial calibration parameter, and a zero offset; the output is a zero drift correction coefficient; The scale factor drift correction equation is used to calculate the nonlinear variation characteristics and temperature response characteristics of the gravimeter scale factor. The input includes the scale factor drift, temperature change gradient, measurement time interval, instrument initial calibration parameters and environmental parameter time series. The output is the scale factor drift correction coefficient.

3. The method for on-site quality assessment of gravity measurement data according to claim 2, characterized in that: The drift compensation function is specifically used to eliminate the zero offset and scale factor drift errors generated by the gravimeter during long-term operation. The input includes the reference point gravity value sequence, the environmental parameter time series, the measurement time interval and the instrument initial calibration parameters. The output is the drift-compensated gravity measurement value.

4. The method for on-site quality assessment of gravity measurement data according to claim 3, characterized in that: The gravity anomaly detection model is specifically a sequence analysis model based on the Transformer architecture, which includes a multi-head attention mechanism to capture the spatiotemporal correlation of gravity data. The number of attention heads is dynamically determined based on three parameters: the complexity of the measurement area, the data sampling density, and the environmental noise level. When the measurement area has complex geological structures, high data sampling density, and low environmental noise levels, the number of attention heads is increased to improve anomaly detection accuracy.

5. The method for on-site quality assessment of gravity measurement data according to claim 4, characterized in that: The training data set establishment step of the gravity anomaly detection model specifically involves collecting gravity measurement data under different geological environments as normal samples, artificially injecting various types of instrument failures, environmental interference and measurement errors to generate abnormal samples, labeling and classifying all samples to form a training data set containing normal data and abnormal data, and expanding the number of training samples through data augmentation technology to improve the generalization ability of the model.

6. The method for on-site quality assessment of gravity measurement data according to claim 5, characterized in that: The environmental parameter monitoring sensor refers to a multi-type sensor array installed in the measurement area, which is used to monitor the changes of environmental factors that affect the accuracy of gravity measurement in real time.

7. The method for on-site quality assessment of gravity measurement data according to claim 6, characterized in that: The drift monitoring model refers to a state estimation model based on the Kalman filter algorithm, which is used to estimate and predict the drift state of the gravimeter in real time.

8. The method for on-site quality assessment of gravity measurement data according to claim 7, characterized in that: The multi-dimensional quality comprehensive evaluation model refers to a comprehensive evaluation algorithm that integrates multiple quality indicators and outputs data quality grade classification results.

9. The method for on-site quality assessment of gravity measurement data according to claim 8, characterized in that: The attention head number adjustment function is specifically used to adjust the multi-head attention mechanism parameters of the gravity anomaly detection model. The attention adjustment value is calculated based on three data: measurement area complexity, data sampling density, and environmental noise level. When the attention adjustment value is in the range of 0 to 0.25, 4 attention heads are used and the attention range of each attention head is 256 time steps. When the attention adjustment value is in the range of 0.25 to 0.5, 6 attention heads are used and the attention range of each attention head is 128 time steps. When the attention adjustment value is in the range of 0.5 to 0.75, 8 attention heads are used and the attention range of each attention head is 64 time steps. When the attention adjustment value is in the range of 0.75 to 1, 12 attention heads are used and the attention range of each attention head is 32 time steps to adjust the multi-head attention mechanism parameters of the model.

10. The method for on-site quality assessment of gravity measurement data according to claim 9, characterized in that: The attention weight threshold of each attention head is dynamically set according to the variance of the gravity data sequence. When the variance of the gravity data sequence is less than 0.01mGal, the attention weight threshold is set to 0.8; when the variance of the gravity data sequence is between 0.01 and 0.05mGal, the attention weight threshold is set to 0.6; when the variance of the gravity data sequence is greater than 0.05mGal, the attention weight threshold is set to 0.

4. The attention head uses a sliding window attention method to process the gravity data time series and screens important features according to the attention weight threshold.

Images