Radar liquid level meter system fusing adaptive algorithm and multi-modal data

Abstract

The invention belongs to the technical field of liquid level measurement and industrial process control, and particularly relates to a radar liquid level meter system fusing an adaptive algorithm and multi-modal data. Comprising a data acquisition unit, a signal acquisition and preprocessing unit, a signal denoising and optimizing unit, a data fusion and anomaly detection unit, a self-adaptive calibration and drift compensation unit and a display and alarm terminal. The system has the advantages that the radar liquid level meter, the auxiliary sensor, signal processing, data fusion, anomaly detection, self-adaptive calibration, drift compensation and other technical means are organically combined, and the system can achieve high-precision and real-time monitoring of the liquid level under the complex working condition.

Description

Technical Field

[0001] The present invention belongs to the technical field of liquid level measurement and industrial process control, and particularly relates to a radar liquid level gauge system integrating an adaptive algorithm and multi-modal data. Background Art

[0002] In the context of the continuous improvement of industrial automation and intelligence today, the accurate measurement and control of liquid level are crucial for ensuring production efficiency and equipment safety. As a non-contact liquid level measurement instrument, the radar liquid level gauge is widely used in industries such as petroleum, chemical, and metallurgy due to its high precision and reliability. However, in complex working conditions, such as when there is enamel on the inner wall of the storage tank, irregular structures, or violent fluctuations in the liquid surface, microwave signals are easily affected by multiple reflections, interference, and noise, resulting in a decrease in measurement accuracy. In addition, during the long-term operation of the radar liquid level gauge, changes in environmental conditions and equipment aging may cause measurement drift, affecting the stability and safety of the production process.

[0003] Traditional radar liquid level gauges usually rely on fixed algorithms and parameter settings, making it difficult to adapt to complex and changing working conditions in a timely manner, and unable to effectively eliminate noise interference or perform adaptive calibration. Summary of the Invention

[0004] The object of the present invention is to provide a radar liquid level gauge system integrating an adaptive algorithm and multi-modal data, which integrates an adaptive algorithm and multi-modal data fusion technology to improve the accuracy and reliability of liquid level measurement. The design goal is to achieve noise elimination and signal optimization for the signals received by the radar liquid level gauge under complex working conditions, combine other sensor data for data fusion and anomaly detection, and ensure high-precision measurement during long-term operation of the device through adaptive calibration and drift compensation. Through this technical integration, the present invention is expected to improve the real-time performance and accuracy of liquid level monitoring, reduce production risks caused by measurement errors, and enhance the efficiency and safety of industrial process control.

[0005] The technical solution of the present invention is as follows: A radar liquid level gauge system integrating an adaptive algorithm and multi-modal data, including a data acquisition unit, a signal acquisition and preprocessing unit, a signal denoising and optimization unit, a data fusion and anomaly detection unit, an adaptive calibration and drift compensation unit, and a display and alarm terminal;

[0006] The data acquisition unit includes a radar liquid level gauge and auxiliary sensors. The radar liquid level gauge is used to detect the liquid level information in the container, transmit microwave signals, and receive the signals reflected from the liquid surface. The auxiliary sensors include a temperature sensor and a pressure sensor, which are used to measure environmental parameters;

[0007] The signal acquisition and preprocessing unit is used to collect the original data of the radar level gauge and auxiliary sensors in real time, and perform synchronous sampling, filtering, and normalization processing on the signals;

[0008] The signal denoising and optimization unit is used to perform denoising and optimization processing on the preprocessed radar level gauge signals, and improve the signal-to-noise ratio and measurement accuracy by eliminating complex noises and interferences;

[0009] The data fusion and anomaly detection unit is used to fuse and analyze the radar level gauge signals and the data of auxiliary sensors that have undergone denoising and optimization processing, and use anomaly detection algorithms to identify abnormal situations in the data, and provide early warnings and fault diagnoses;

[0010] In the case where no anomaly is detected, the adaptive calibration and drift compensation unit dynamically adjusts the calibration coefficient k according to the residual r(t) to compensate for measurement drift, reduce measurement errors, and improve the accuracy of liquid level estimation;

[0011] The display and alarm terminal is used to display the liquid level information, environmental parameters, and system status in real time, provide a friendly human-machine interface, and send an alarm signal when an abnormal situation is detected.

[0012] The synchronous sampling in the signal acquisition and preprocessing unit is to determine a common sampling frequency f s , for sensors with a sampling frequency lower than f s , the interpolation method is used to increase the sampling rate; for sensors with a sampling frequency higher than f s , the downsampling method is used to reduce the sampling rate;

[0013] The filtering process uses a low-pass filter to eliminate high-frequency noise and retain the low-frequency components of the signal, a high-pass filter is used to eliminate low-frequency drift and retain the high-frequency components of the signal, and a band-pass filter is used to extract the signal components in a specific frequency band.

[0014] The denoising and optimization processing includes wavelet transform denoising and signal smoothing.

[0015] The wavelet transform denoising is to perform wavelet decomposition, threshold processing, and signal reconstruction on the preprocessed signal to eliminate noise components. The specific process is as follows:

[0016] Wavelet decomposition: Perform multi-scale wavelet decomposition on the normalized radar signal X norm (t) to decompose the signal into detail coefficients d j,k and approximation coefficients c j,k :

[0017]

[0018] Among them: Xnorm (t) The normalized radar signal, φ J,k (t) The scaling function, representing the low-frequency component of the signal, ψ j,k (t) The wavelet function, representing the high-frequency detail components of the signal at different scales, c j,k The approximation coefficient, representing the low-frequency part of the signal at the Jth layer, d j,k The detail coefficient, representing the high-frequency details of the signal at the jth layer, (j = 1, 2, …, J).

[0019] Threshold processing: For the detail coefficient d j,k Perform threshold processing to eliminate the influence of noise. The soft threshold method is adopted, specifically:

[0020]

[0021] Where: The detail coefficient after threshold processing;

[0022] sgn(d j,k ) The sign function, taking 1 when d j,k > 0, taking -1 when d j,k < 0, and taking 0 when d j,k = 0;

[0023] max(|d j,k |- λ j , 0) The threshold of the jth layer, usually calculated according to the statistical characteristics of the detail coefficient;

[0024] Signal reconstruction: Utilize the processed detail coefficient and the approximation coefficient c j,k to reconstruct the denoised signal through the inverse wavelet transform

[0025]

[0026] Where: The signal after wavelet denoising.

[0027] The signal smoothing means performing smoothing processing on the denoised signal to eliminate the residual high-frequency noise and further improve the signal quality. The moving average filter is adopted:

[0028]

[0029] Where: The smoothed signal, The length of the smoothing window, usually an odd number, determines the degree of smoothing, Δt is the sampling time interval, and n is the sampling point index within the window.

[0030] Data fusion includes the following steps:

[0031] Weighted average fusion: The measurement signals of different sensors are weighted and averaged according to certain weights to obtain the fused liquid level estimation value. The specific process is as follows:

[0032] (1) Signal acquisition: Obtain the signal of the radar liquid level gauge after denoising and optimization Auxiliary sensor signal

[0033] (2) Weight determination: Determine the weight coefficients w radar and w radar , satisfying w radar +w radar =1;

[0034] (3) Weighted average calculation: The fused liquid level estimation value is:

[0035]

[0036] Where: S fusion (t) The fused liquid level estimation value, w radar , w aux The weight coefficients of the radar liquid level gauge and the auxiliary sensor, The sensor signal after denoising and optimization.

[0037] The results of multi-sensor data are input into the Kalman filter for optimization, including the following steps:

[0038] Step 1: Establish a state space model

[0039] Define the system state and observation model:

[0040] (1) State variable: x(t), representing the true liquid level value;

[0041] (2) State equation:

[0042] x(t) = x(t - 1) + v(t) (6)

[0043] Where: v(t) is the process noise, assumed to be Gaussian white noise with zero mean and variance Q;

[0044] (3) Observation equation:

[0045] z(t) = H·x(t) + w(t) (7)

[0046] Where: z(t) Observation value, i.e., the weighted fused liquid level estimation value S fusion (t)

[0047] H: Observation matrix, with a value of 1, w(t) is measurement noise, assumed to be Gaussian white noise with zero mean and variance R;

[0048] Step 2: Kalman filter recursion formula

[0049] (1) Initialization

[0050] Initial state estimate:

[0051]

[0052] Initial error covariance:

[0053]

[0054] (2) Prediction stage

[0055] State prediction:

[0056]

[0057] Error covariance prediction:

[0058] P - (t) = P(t - 1) + Q (11)

[0059] (3) Update stage

[0060] Kalman gain calculation:

[0061]

[0062] State update:

[0063]

[0064] Error covariance update:

[0065] P(t) = [I - K(t)·H]·P - (t) (14)

[0066] Where: The predicted value of the state at time t, The estimated value of the state at time t, P - (t) The predicted error covariance, P(t) The updated error covariance, K(t) The Kalman gain, Q The process noise covariance, R The measurement noise covariance, I The identity matrix;

[0067] Output the estimated liquid level value:

[0068]

[0069] Where: L KF(t) is the estimated liquid level output by the Kalman filter.

[0070] The anomaly detection is used to ensure the reliability of the system and the accuracy of the measurement results. By identifying anomalies in the data, including sensor failures, environmental interferences, or system malfunctions, it includes the following steps:

[0071] (1) Residual calculation: Calculate the residual r(t) between the observed value and the predicted value:

[0072]

[0073] where: r(t): the residual vector, reflecting the difference between the measured value and the predicted value;

[0074] (2) Threshold setting: According to historical data and system requirements, set the threshold ξ of the residual. When the absolute value of the residual exceeds the threshold, it is determined that there is an anomaly:

[0075] If |r(t)| > ξ, it is determined as an anomaly;

[0076] (3) Anomaly handling: When an anomaly is detected, trigger the corresponding warning mechanism, record the anomaly information, or enable the backup sensor and redundant system to ensure the stable operation of the system.

[0077] The adaptive calibration and drift compensation unit includes,

[0078] Adjustment of the calibration coefficient k:

[0079] k t = k t-1 + α·r(t) (17)

[0080] where: k t The updated calibration coefficient at the current moment, k t-1 The calibration coefficient at the previous moment, α calibration step size, controlling the adjustment speed, α > 0, r(t) residual,

[0081] Drift compensation:

[0082] The adjustment of the calibration coefficient k realizes the compensation for measurement drift;

[0083] Calibrated liquid level estimate:

[0084] Use the updated calibration coefficient k t Calculate the calibrated liquid level estimate L KF (t)

[0085] L cal (t) = k t ·L KF (t) (18)

[0086] Wherein: L cal (t) is the calibrated liquid level estimation value, L KF (t) is the liquid level estimation value output by the Kalman filter.

[0087] The display and alarm terminal includes the following functional modules:

[0088] Data storage and management module: used to store the generated liquid level data, environmental parameters and historical records, and provide functions of data query, statistics and analysis;

[0089] Real-time display module: visually display the current liquid level information, temperature and pressure parameters through an intuitive interface;

[0090] Historical trend analysis module: provides functions of historical data playback and trend analysis of liquid level and environmental parameters, and supports multi-dimensional display and comparison of data;

[0091] Alarm and early warning module: according to the set thresholds and abnormal detection results, when the liquid level exceeds the safe range or a system abnormality is detected, it issues an audible and visual alarm and information prompt in a timely manner;

[0092] Human-computer interaction and parameter setting module: supports operators to set and adjust system parameters, including alarm thresholds, display modes, and data acquisition frequencies;

[0093] Network communication and remote monitoring module: supports communication with the upper computer, control center or mobile devices to achieve remote monitoring and data sharing.

[0094] The beneficial effects of the present invention are as follows: by organically combining various technical means such as radar liquid level gauges, auxiliary sensors, signal processing, data fusion, anomaly detection, adaptive calibration and drift compensation, the system can achieve high-precision and real-time monitoring of the liquid level under complex working conditions. Description of the Drawings

[0095] Figure 1 It is a flowchart of an improved radar liquid level gauge system with an adaptive algorithm and multi-modal data fusion technology provided by the present invention. Detailed Embodiments

[0096] The following details the specific embodiments of the present invention for the convenience of those skilled in the art to understand and implement. It should be noted that the following description is only an exemplary implementation of the present invention and is not used to limit the protection scope of the present invention.

[0097] A radar level gauge system integrating an adaptive algorithm and multi-modal data, comprising: a data acquisition unit based on a radar level gauge and auxiliary sensors, a signal acquisition and preprocessing unit, a signal denoising and optimization unit, a data fusion and anomaly detection unit, an adaptive calibration and drift compensation unit, and a display and alarm terminal.

[0098] The radar level gauge is used to detect the liquid level information in the container, transmit microwave signals and receive the signals reflected from the liquid surface. However, under complex working conditions, such as when there is enamel or irregular structures on the inner wall of the storage tank, the microwave signals are prone to multiple reflections, interference and noise, and the received signals need to be further processed to extract effective liquid level information.

[0099] The auxiliary sensors include a temperature sensor, a pressure sensor, etc., which are used to measure environmental parameters, provide an additional information source for multi-modal data fusion, and improve the accuracy and reliability of liquid level measurement.

[0100] The signal acquisition and preprocessing unit is used to collect the raw data of the radar level gauge and auxiliary sensors in real time, perform synchronous sampling, filtering and normalization processing on the signals. Through these preprocessing steps, the impacts caused by sampling frequency, time delay and signal amplitude differences are eliminated, laying a solid foundation for subsequent signal denoising, data fusion and anomaly detection.

[0101] The synchronous sampling is to determine a common sampling frequency f s , for sensors with a sampling frequency lower than f s , the interpolation method is used to increase the sampling rate; for sensors with a sampling frequency higher than f s , the downsampling method is used to reduce the sampling rate.

[0102] The filtering process is that the original signal may contain high-frequency noise, low-frequency drift and other interferences, and it needs to be processed by filters to extract the effective signal components and improve the signal quality. The low-pass filter is used to eliminate high-frequency noise and retain the low-frequency components of the signal, the high-pass filter is used to eliminate low-frequency drift and retain the high-frequency components of the signal, and the band-pass filter is used to extract the signal components in a specific frequency band. The normalization process is that the signal amplitude ranges and units of different sensors may be different, and the signals need to be normalized to make them comparable and facilitate subsequent algorithm processing.

[0103] The signal denoising and optimization unit is used to perform advanced denoising and optimization processing on the preprocessed radar level gauge signals. By eliminating complex noise and interference, the signal-to-noise ratio and measurement accuracy of the signals are improved, providing a high-quality signal basis for subsequent data fusion and anomaly detection. The denoising and optimization processing includes steps such as wavelet transform denoising and signal smoothing, which can effectively suppress non-linear and non-stationary noise and retain the useful features of the signals.

[0104] The wavelet transform denoising is to perform wavelet decomposition, threshold processing, and signal reconstruction on the preprocessed signal to eliminate noise components. The specific process is as follows:

[0105] Wavelet decomposition: Perform multi-scale wavelet decomposition on the normalized radar signal X norm (t) to decompose the signal into detail coefficients d j,k and approximation coefficients c j,k :

[0106]

[0107] Where: X norm (t) is the normalized radar signal, φ J,k (t) is the scaling function, representing the low-frequency component of the signal, ψ j,k (t) is the wavelet function, representing the high-frequency detail components of the signal at different scales, c j,k is the approximation coefficient, representing the low-frequency part of the signal at the Jth layer, d j,k is the detail coefficient, representing the high-frequency details of the signal at the jth layer, (j = 1, 2,..., J).

[0108] Threshold processing: Perform threshold processing on the detail coefficients d j,k to eliminate the influence of noise. The soft threshold method is adopted, and the processing formula is:

[0109]

[0110] Where: is the detail coefficient after threshold processing.

[0111] sgn(d j,k ) is the sign function, which takes 1 when d j,k > 0, -1 when d j,k < 0, and 0 when d j,k = 0.

[0112] max(|d j,k |-λ j , 0) is the threshold at the jth layer, usually calculated according to the statistical characteristics of the detail coefficients.

[0113] Signal reconstruction: Use the processed detail coefficients and approximation coefficients c j,k to reconstruct the denoised signal through inverse wavelet transform

[0114]

[0115] Where: is the signal after wavelet denoising.

[0116] The signal smoothing is to perform smoothing processing on the denoised signal to eliminate the residual high-frequency noise and further improve the signal quality. A moving average filter is adopted, and its formula is:

[0117]

[0118] Where: the smoothed signal, the length of the smoothing window, usually an odd number, determines the degree of smoothing, Δt is the sampling time interval, and n is the sampling point index within the window.

[0119] Through the above denoising and optimization processing, a high-quality signal is obtained providing a reliable data basis for subsequent data fusion and anomaly detection.

[0120] The data fusion and anomaly detection unit is used to fuse and analyze the radar level gauge signal and the data of the auxiliary sensor that have undergone denoising and optimization processing. By integrating multi-source information, the accuracy and reliability of level measurement are improved. At the same time, an anomaly detection algorithm is used to identify abnormal situations in the data and provide early warnings and fault diagnoses. The data fusion and anomaly detection include steps such as weighted average fusion, Kalman filter fusion, and anomaly detection based on statistical characteristics, which can effectively improve the measurement accuracy and robustness of the system.

[0121] Data fusion includes the following steps:

[0122] Weighted average fusion: The measurement signals of different sensors are weighted and averaged according to certain weights to obtain the fused level estimate value

[0123] The specific process is as follows:

[0124] (1) Signal acquisition: Obtain the radar level gauge signal that has undergone denoising and optimization processing the auxiliary sensor signal

[0125] (2) Weight determination: Determine the weight coefficients w radar and w radar , satisfying w radar +w radar = 1.

[0126] (3) Weighted average calculation: The formula for calculating the fused level estimate value is:

[0127]

[0128] Where: S fusion(t) The estimated liquid level value after fusion, w radar , w aux The weight coefficients of the radar liquid level gauge and the auxiliary sensor The sensor signal after denoising and optimization

[0129] The Kalman filter fusion is to use the Kalman filter to perform optimal estimation on multi-sensor data, reduce the influence of noise, and improve the measurement accuracy. Based on the above weighted average fusion result of multi-sensor data, it is input into the Kalman filter for further optimized estimation

[0130] Step 1: Establish a state space model

[0131] Define the system state and observation model

[0132] (1) State variable: x(t), representing the true liquid level value

[0133] (2) State equation

[0134] x(t) = x(t - 1) + v(t) (6)

[0135] Where

[0136] v(t) is the process noise, assumed to be Gaussian white noise with zero mean and variance Q

[0137] (3) Observation equation

[0138] z(t) = H · x(t) + w(t) (7)

[0139] Where

[0140] z(t) Observation value, that is, the estimated liquid level value S after weighted fusion fusion (t)

[0141] H: Observation matrix, with a value of 1. w(t) Measurement noise, assumed to be Gaussian white noise with zero mean and variance R

[0142] Step 2: Kalman filter recursion formula

[0143] (1) Initialization

[0144] Initial state estimate

[0145]

[0146] Initial error covariance

[0147]

[0148] (2) Prediction stage

[0149] State prediction

[0150]

[0151] Error covariance prediction:

[0152] P - (t) = P(t - 1) + Q (11)

[0153] (3) Update phase

[0154] Kalman gain calculation:

[0155]

[0156] State update:

[0157]

[0158] Error covariance update:

[0159] P(t) = [I - K(t)·H]·P - (t) (14)

[0160] Where: The predicted value of the state at time t, The estimated value of the state at time t, P - (t) The predicted error covariance, the updated error covariance of P(t), K(t) the Kalman gain, Q the process noise covariance, R the measurement noise covariance, I the identity matrix.

[0161] Output the liquid level estimated value:

[0162]

[0163] Where: L KF (t) is the liquid level estimated value output by the Kalman filter.

[0164] The anomaly detection is used to ensure the reliability of the system and the accuracy of the measurement results. By identifying anomalies in the data, such as sensor failures, environmental interferences, or system malfunctions, the system can give early warnings in a timely manner and take corrective measures to prevent incorrect data from affecting subsequent processing and decision-making. It includes the following steps:

[0165] (1) Residual calculation: Calculate the residual r(t) between the observed value and the predicted value:

[0166]

[0167] Where: r(t): The residual vector, reflecting the difference between the measured value and the predicted value.

[0168] (2) Threshold setting: Set the threshold ξ of the residual according to historical data and system requirements. When the absolute value of the residual exceeds the threshold, it is determined that there is an anomaly:

[0169] If |r(t)| > ξ, it is determined as an anomaly.

[0170] (3) Anomaly handling: When an anomaly is detected, trigger the corresponding warning mechanism, record the anomaly information, or enable backup sensors and redundant systems to ensure the stable operation of the system.

[0171] In the case where the adaptive calibration and drift compensation unit does not detect an anomaly, the system dynamically adjusts the calibration coefficient k according to the residual r(t) to compensate for measurement drift, reduce measurement error, and improve the accuracy of liquid level estimation.

[0172] Adjustment of the calibration coefficient k:

[0173] k t = k t-1 + α·r(t) (17)

[0174] Where: k t The updated calibration coefficient at the current moment, k t-1 The calibration coefficient at the previous moment, α is the calibration step (learning rate) that controls the adjustment speed, α > 0, r(t) is the residual,

[0175] Drift compensation:

[0176] The adjustment of the calibration coefficient k realizes the compensation for measurement drift. By continuously updating k, the system can adapt to the drift caused by environmental changes, sensor aging, etc., and maintain the accuracy of the measurement results.

[0177] Calibrated liquid level estimate:

[0178] Use the updated calibration coefficient k t Calculate the calibrated liquid level estimate L KF (t)

[0179] L cal (t) = k t ·L KF (t) (18)

[0180] Where: L cal (t) is the calibrated liquid level estimate, L KF (t) is the liquid level estimate output by the Kalman filter.

[0181] The display and alarm terminal is used to display the liquid level information, environmental parameters and system status in real time, provide a friendly human-machine interaction interface, and send out alarm signals when abnormal situations are detected, ensuring that operators can timely understand the system operation status and take necessary measures.

[0182] The display and alarm terminal includes the following functional modules:

[0183] Data storage and management module: It is used to securely store the liquid level data, environmental parameters and historical records generated by the system, provide functions of data query, statistics and analysis, and lay a foundation for subsequent decision support and trend prediction.

[0184] Real-time display module: It displays the current liquid level information, temperature, pressure and other parameters in real time through an intuitive interface, and supports various forms of data display, including numbers, charts and curves, facilitating operators to quickly obtain key information.

[0185] Historical trend analysis module: It provides functions of historical data playback and trend analysis of liquid level and environmental parameters, supports multi-dimensional display and comparison of data, helps operators understand the operation rules of the system, and conducts preventive maintenance and optimization.

[0186] Alarm and early warning module: According to the set thresholds and abnormal detection results, when the liquid level exceeds the safe range or system abnormalities are detected, it timely issues audible and visual alarms and information prompts to remind operators to take corresponding measures.

[0187] Human-machine interaction and parameter setting module: It supports operators to set and adjust system parameters, including alarm thresholds, display modes, data acquisition frequencies, etc., improving the flexibility and controllability of the system.

[0188] Network communication and remote monitoring module: It supports communication with the upper computer, control center or mobile devices, realizes remote monitoring and data sharing, and facilitates managers to obtain system information anytime and anywhere.

[0189] Embodiment:

[0190] A radar liquid level gauge system integrating adaptive algorithms and multi-modal data, which aims to achieve high-precision and real-time monitoring of the liquid level in industrial containers through advanced signal processing technologies and intelligent data analysis methods.

[0191] As Figure 1 shown, the system mainly includes the following six units: radar liquid level gauge and auxiliary sensors, signal acquisition and preprocessing unit, signal denoising and optimization unit, data fusion and anomaly detection unit, adaptive calibration and drift compensation unit, and display and alarm terminal.

[0192] The radar level gauge is installed on the top of the container and is used to transmit microwave signals and receive the signals reflected from the liquid surface. This device can measure the liquid level height without contact. However, under complex working conditions, such as when there is enamel, irregular structure on the inner wall of the storage tank or the liquid surface fluctuates violently, the microwave signal is easily affected by multiple reflections, interference and noise, and the received signal needs to be further processed to extract effective liquid level information.

[0193] The auxiliary sensors include temperature sensors, pressure sensors, etc., which are installed at appropriate positions of the container and are used to monitor environmental parameters in real time. The additional information provided by these sensors provides an important data source for multimodal data fusion and improves the accuracy and reliability of liquid level measurement.

[0194] The main function of the signal acquisition and preprocessing unit is to obtain the raw data of the radar level gauge and the auxiliary sensors in real time, and perform synchronous sampling, filtering and normalization processing on these signals. Through these preprocessing steps, the effects caused by sampling frequency, time delay and signal amplitude difference can be effectively eliminated, laying a solid foundation for subsequent signal denoising, data fusion and anomaly detection. In terms of synchronous sampling, the system will determine a unified common sampling frequency f s ; for sensors with a sampling frequency lower than f s , an interpolation method is used to increase their sampling rate, while for sensors with a sampling frequency higher than f s , a downsampling method is used to reduce their sampling rate. In terms of filtering processing, for the high-frequency noise, low-frequency drift and other interferences that may exist in the original signal, filters are used to extract effective signal components and improve the signal quality. Among them, a low-pass filter is used to eliminate high-frequency noise and retain the low-frequency part of the signal; a high-pass filter is used to eliminate low-frequency drift and retain the high-frequency part of the signal; a band-pass filter is used to extract signal components in a specific frequency band. In terms of normalization processing, since the signal amplitude ranges and units of different sensors may be different, the signals need to be normalized to make them comparable and facilitate the unified processing of subsequent algorithms.

[0195] The main function of the signal denoising and optimization unit is to perform advanced denoising and optimization processing on the preprocessed radar level gauge signal. By effectively eliminating the complex noise and interference in the signal, the signal-to-noise ratio and measurement accuracy are improved, providing a high-quality signal basis for subsequent data fusion and anomaly detection. This process includes steps such as wavelet transform denoising and signal smoothing, which can effectively suppress non-linear and non-stationary noise while retaining the useful features of the signal.

[0196] The wavelet transform denoising is to perform wavelet decomposition, threshold processing and signal reconstruction on the preprocessed signal to eliminate the noise components. The specific process is as follows:

[0197] Wavelet decomposition: For the normalized radar signal X norm (t), perform multi-scale wavelet decomposition to decompose the signal into detail coefficients d j,k and approximation coefficients c j,k :

[0198]

[0199] Wherein:

[0200] X norm (t): The normalized radar signal. φ J,k (t): The scaling function, representing the low-frequency component of the signal. ψ j,k (t): The wavelet function, representing the high-frequency detail components of the signal at different scales. c j,k : The approximation coefficient, representing the low-frequency part of the signal at the Jth layer. d j,k : The detail coefficient, representing the high-frequency details of the signal at the jth layer, (j = 1, 2,..., J).

[0201] Threshold processing: Perform threshold processing on the detail coefficients d j,k to eliminate the influence of noise. The soft threshold method is adopted, and the processing formula is:

[0202]

[0203] Wherein:

[0204] The detail coefficient after threshold processing.

[0205] sgn(d j,k ): The sign function, taking 1 when d j,k > 0, taking -1 when d j,k < 0, and taking 0 when d j,k = 0.

[0206] max(|d j,k |-λ j , 0): The threshold at the jth layer, usually calculated according to the statistical characteristics of the detail coefficients.

[0207] Signal reconstruction: Utilize the processed detail coefficients and approximation coefficients c j,k to reconstruct the denoised signal through inverse wavelet transform

[0208]

[0209] Wherein: The signal after wavelet denoising.

[0210] The signal smoothing is to the denoised signal Perform smoothing to eliminate residual high-frequency noise and further improve the signal quality. A moving average filter is used, and its formula is:

[0211]

[0212] Where: The smoothed signal. The length of the smoothing window, usually an odd number, determines the degree of smoothing. Δt: Sampling time interval. n: Sampling point index within the window.

[0213] Through the above denoising and optimization processes, a high-quality signal is obtained Providing a reliable data basis for subsequent data fusion and anomaly detection.

[0214] The data fusion and anomaly detection unit is used to fuse and analyze the radar level gauge signal and the data of auxiliary sensors that have undergone denoising and optimization processing. By integrating multi-source information, the accuracy and reliability of level measurement are improved. At the same time, using anomaly detection algorithms, abnormal situations in the data are identified to provide early warnings and fault diagnoses. The data fusion and anomaly detection include steps such as weighted average fusion, Kalman filter fusion, and anomaly detection based on statistical characteristics, which can effectively improve the measurement accuracy and robustness of the system.

[0215] Data fusion includes the following steps:

[0216] Weighted average fusion: Perform weighted averaging on the measurement signals of different sensors according to certain weights to obtain the fused level estimation value

[0217] The specific process is as follows:

[0218] (1) Signal acquisition: Obtain the radar level gauge signal that has undergone denoising and optimization processing Auxiliary sensor signal

[0219] (2) Weight determination: Determine the weight coefficients w radar and w radar , satisfying w radar +w radar = 1.

[0220] (3) Weighted average calculation: The formula for calculating the fused level estimation value is:

[0221]

[0222] Where: S fusion (t): The fused level estimation value. w radar 、w aux: Weight coefficients of the radar level gauge and the auxiliary sensor. Sensor signals that have been denoised and optimized.

[0223] The Kalman filter fusion is to use the Kalman filter to perform optimal estimation on multi-sensor data, reduce the influence of noise, and improve the measurement accuracy. Based on the above weighted average fusion results of multi-sensor data, input them into the Kalman filter for further optimal estimation.

[0224] Step 1: Establish a state space model

[0225] Define the system state and observation model:

[0226] (1) State variable: x(t), representing the true liquid level value

[0227] (2) State equation:

[0228] x(t) = x(t - 1) + v(t) (24)

[0229] Where: v(t) is the process noise, assumed to be Gaussian white noise with zero mean and variance Q.

[0230] (3) Observation equation:

[0231] z(t) = H · x(t) + w(t) (25)

[0232] Where: z(t) is the observed value, that is, the liquid level estimated value S after weighted fusion fusion (t)

[0233] H: Observation matrix, with a value of 1.

[0234] w(t) is the measurement noise, assumed to be Gaussian white noise with zero mean and variance R.

[0235] Step 2: Kalman filter recursion formula

[0236] (1) Initialization

[0237] Initial state estimate:

[0238]

[0239] Initial error covariance:

[0240]

[0241] (2) Prediction stage

[0242] State prediction:

[0243]

[0244] Error covariance prediction:

[0245] P - (t) = P(t - 1) + Q (29)

[0246] (3) Update phase

[0247] Kalman gain calculation:

[0248]

[0249] State update:

[0250]

[0251] Error covariance update:

[0252] P(t) = [I - K(t)·H]·P - (t) (32)

[0253] Where: The predicted value of the state at time t; The estimated value of the state at time t; P - (t): Prediction error covariance; P(t): Updated error covariance; K(t): Kalman gain; Q: Process noise covariance; R: Measurement noise covariance; I: Identity matrix.

[0254] Output liquid level estimated value:

[0255]

[0256] Where: L KF (t) is the liquid level estimated value output by the Kalman filter.

[0257] The anomaly detection is used to ensure the reliability of the system and the accuracy of the measurement results. By identifying anomalies in the data, such as sensor failures, environmental disturbances, or system failures, the system can give early warnings in a timely manner and take corrective measures to prevent incorrect data from affecting subsequent processing and decision-making. It includes the following steps:

[0258] (1) Residual calculation: Calculate the residual r(t) between the observed value and the predicted value:

[0259]

[0260] Where: r(t): Residual vector, reflecting the difference between the measured value and the predicted value.

[0261] (2) Threshold setting: Set the threshold ξ of the residual according to historical data and system requirements. When the absolute value of the residual exceeds the threshold, it is determined that there is an anomaly:

[0262] If |r(t)| > ξ, it is determined as an anomaly.

[0263] (3) Anomaly handling: When an anomaly is detected, trigger the corresponding warning mechanism, record the anomaly information, or enable backup sensors and redundant systems to ensure the stable operation of the system.

[0264] In the case where the adaptive calibration and drift compensation unit does not detect an anomaly, the system dynamically adjusts the calibration coefficient k according to the residual r(t) to compensate for measurement drift, reduce measurement error, and improve the accuracy of liquid level estimation.

[0265] Adjustment of the calibration coefficient k:

[0266] k t = k t-1 + α·r(t) (35)

[0267] Where: k t : The updated calibration coefficient at the current moment; k t-1 : The calibration coefficient at the previous moment; α: Calibration step size (learning rate), which controls the adjustment speed, α > 0; r(t): Residual

[0268] Drift compensation:

[0269] The adjustment of the calibration coefficient k realizes the compensation for measurement drift. By continuously updating k, the system can adapt to the drift caused by environmental changes, sensor aging, etc., and maintain the accuracy of measurement results.

[0270] Calibrated liquid level estimation value:

[0271] Use the updated calibration coefficient k t to calculate the calibrated liquid level estimation value L KF (t)

[0272] L cal (t) = k t ·L KF (t) (36)

[0273] Where: L cal (t): Calibrated liquid level estimation value; L KF (t): Liquid level estimation value output by the Kalman filter.

[0274] The display and alarm terminal is used to display liquid level information, environmental parameters, and system status in real time, provide a friendly human-machine interface, and send an alarm signal when an abnormal situation is detected to ensure that operators can timely understand the system operation status and take necessary measures.

[0275] The display and alarm terminal includes the following functional modules:

[0276] Data storage and management module: It is used to securely store the liquid level data, environmental parameters and historical records generated by the system, provide functions for data query, statistics and analysis, and lay a foundation for subsequent decision support and trend prediction.

[0277] Real-time display module: It visually displays current liquid level information, temperature, pressure and other parameters through an intuitive interface, supports various forms of data display, including numbers, charts and curves, and facilitates operators to quickly obtain key information.

[0278] Historical trend analysis module: It provides functions for historical data playback and trend analysis of liquid level and environmental parameters, supports multi-dimensional display and comparison of data, helps operators understand the operation rules of the system, and conducts preventive maintenance and optimization.

[0279] Alarm and early warning module: According to the set thresholds and abnormal detection results, when the liquid level exceeds the safe range or system abnormalities are detected, it promptly issues audible and visual alarms and information prompts to remind operators to take corresponding measures.

[0280] Human-computer interaction and parameter setting module: It supports operators to set and adjust system parameters, including alarm thresholds, display modes, data acquisition frequencies, etc., and improves the flexibility and controllability of the system.

[0281] Network communication and remote monitoring module: It supports communication with the upper computer, control center or mobile devices, realizes remote monitoring and data sharing, and facilitates managers to obtain system information anytime and anywhere.

[0282] The present invention provides an improved radar liquid level gauge system with an adaptive algorithm and multi-modal data fusion technology. By organically combining various technical means such as radar liquid level gauges, auxiliary sensors, signal processing, data fusion, anomaly detection, adaptive calibration and drift compensation, the system can achieve high-precision and real-time monitoring of the liquid level under complex working conditions. Under complex working conditions, such as when there is enamel, irregular structures or violent fluctuations in the liquid level on the inner wall of the storage tank, the system can still accurately measure the liquid level, significantly improving the accuracy and reliability of monitoring.

[0283] In practical applications, when the liquid level changes abnormally, which may affect the stability of the production process and the safety of equipment, this system can promptly issue early warnings, provide accurate liquid level information for operators, and support effective production management and operation decisions. This not only avoids production accidents and equipment failures caused by liquid level measurement errors, but also greatly reduces maintenance costs and downtime, and improves production efficiency.

[0284] The integrated design of the display and alarm terminal of the present invention includes functions of data storage, early warning issuance, and intuitive display, ensuring the high efficiency of liquid level monitoring and management, enabling operators to easily master the liquid level status, and realizing data-driven decision support. In short, through technological innovation, the present invention effectively addresses the limitations of traditional radar liquid level gauges and provides an efficient and reliable solution for industrial process control.

Claims

1. A radar level gauge system integrating adaptive algorithm and multimodal data, characterized in that: It includes data acquisition unit, signal acquisition and preprocessing unit, signal denoising and optimization unit, data fusion and anomaly detection unit, adaptive calibration and drift compensation unit, display and alarm terminal; The data acquisition unit includes a radar level meter and an auxiliary sensor. The radar level meter is used to detect the liquid level information in the container, transmit microwave signals and receive signals reflected from the liquid surface. The auxiliary sensor includes a temperature sensor and a pressure sensor for measuring environmental parameters. The signal acquisition and preprocessing unit is used to collect the raw data of the radar level meter and the auxiliary sensor in real time, and perform synchronous sampling, filtering and normalization processing on the signals; The signal denoising and optimization unit is used to denoise and optimize the pre-processed radar level meter signal, thereby improving the signal-to-noise ratio and measurement accuracy of the signal by eliminating complex noise and interference; The data fusion and anomaly detection unit is used to fuse and analyze the radar level meter signal and the data of the auxiliary sensor after denoising and optimization processing, use the anomaly detection algorithm to identify the abnormal situation in the data, and provide early warning and fault diagnosis; When no abnormality is detected by the adaptive calibration and drift compensation unit, the system dynamically adjusts the calibration coefficient k according to the residual r(t) to compensate for the measurement drift, reduce the measurement error, and improve the accuracy of the liquid level estimation; The display and alarm terminal is used to display liquid level information, environmental parameters and system status in real time, provide a friendly human-computer interaction interface, and send out an alarm signal when an abnormal situation is detected.

2. A radar level gauge system integrating adaptive algorithm and multi-modal data as claimed in claim 1, characterized in that: The synchronous sampling in the signal acquisition and preprocessing unit is to determine a common sampling frequency f s , for sampling frequencies lower than f s For sensors with sampling frequencies higher than f, interpolation is used to increase the sampling rate. s For sensors, the sampling rate is reduced by downsampling method; The filtering process uses a low-pass filter to eliminate high-frequency noise and retain the low-frequency components of the signal. A high-pass filter is used to eliminate low-frequency drift and retain the high-frequency components of the signal. A band-pass filter is used to extract signal components in a specific frequency band.

3. The radar level gauge system integrating adaptive algorithm and multi-modal data according to claim 1, characterized in that: The denoising and optimization processing includes wavelet transform denoising and signal smoothing.

4. A radar level gauge system integrating adaptive algorithm and multi-modal data as claimed in claim 3, characterized in that: The wavelet transform denoising is to perform wavelet decomposition, threshold processing and signal reconstruction on the pre-processed signal to eliminate the noise component. The specific process is as follows: Wavelet decomposition: normalized radar signal X norm (t) Perform multi-scale wavelet decomposition to decompose the signal into detail coefficients d in different frequency bands j,k and the approximation coefficient c j,k : Where: X norm (t) Normalized radar signal, φ J,k (t) Scaling function, representing the low-frequency component of the signal, ψ j,k (t) Wavelet function, which represents the high-frequency detail components of the signal at different scales, c j,k Approximation coefficient, representing the low-frequency part of the signal at layer J, d j,k The detail coefficient represents the high-frequency details of the signal at the jth layer (j=1,2,…,J). Threshold processing: detail coefficient d j,k Threshold processing is performed to eliminate the influence of noise, and the soft threshold method is used, specifically: in: Detail coefficient after threshold processing; sgn(d j,k ) sign function, when d j,k >0, take 1, d j,k <0, take -1, d j,k =0, take 0; max(|d j,k |-λ j ,0) The threshold of the jth layer is usually calculated based on the statistical characteristics of the detail coefficient; Signal reconstruction: using processed detail coefficients and the approximation coefficient c j,k Reconstruct the denoised signal by inverse wavelet transform in: Signal after wavelet denoising.

5. A radar level gauge system integrating adaptive algorithm and multi-modal data as claimed in claim 4, characterized in that: The signal smoothing is to smooth the denoised signal Smoothing is performed to eliminate residual high-frequency noise and further improve signal quality using a moving average filter: in: The smoothed signal, The length of the smoothing window, usually an odd number, determines the degree of smoothing, Δt is the sampling time interval, and n is the sampling point index within the window.

6. The radar level gauge system integrating adaptive algorithm and multi-modal data according to claim 1, characterized in that: Data fusion includes the following steps: Weighted average fusion: The measurement signals of different sensors are weighted averaged according to certain weights to obtain the fused liquid level estimation value. The specific process is as follows: (1) Signal acquisition: Acquire radar level meter signals after denoising and optimization processing Auxiliary sensor signal (2) Weight determination: Determine the weight coefficient w based on the accuracy, reliability and environmental adaptability of the sensor. radar and w radar , satisfying w radar +w radar =1; (3) Weighted average calculation: The estimated value of the liquid level after fusion is: Where: S fusion (t) Estimated value of liquid level after fusion, w radar 、w aux Weight coefficients for radar level gauges and auxiliary sensors, Denoised and optimized sensor signal.

7. A radar level gauge system integrating adaptive algorithm and multi-modal data as claimed in claim 6, characterized in that: The results of multi-sensor data are input into the Kalman filter for optimization, including the following steps: Step 1: Build a state-space model Define the system state and observation model: (1) State variable: x(t), representing the actual liquid level value; (2) Equation of state: x(t)=x(t-1)+v(t) (6) Where: v(t) is the process noise, which is assumed to be Gaussian white noise with zero mean and variance Q; (3) Observation equation: z(t)=H·x(t)+w(t) (7) Among them: z(t) is the observed value, that is, the estimated value of the liquid level after weighted fusion S fusion (t) H: measurement matrix, the value is 1, w(t) measurement noise, assumed to be Gaussian white noise with zero mean and variance R; Step 2: Kalman filter ground push formula (1) Initialization Initial state estimate: Initial error covariance: (2) Prediction stage Status prediction: Error covariance prediction: P - (t)=P(t-1)+Q (11) (3) Update phase Kalman gain calculation: Status Update: Error covariance update: P(t)=[I-K(t)·H]·P - (t) (14) in: The state prediction value at time t, The estimated value of the state at time t, P - (t) prediction error covariance, P(t) updated error covariance, K(t) Kalman gain, Q process noise covariance, R measurement noise covariance, I identity matrix; Output liquid level estimate: Where: L KF (t) is the estimated value of the liquid level output by the Kalman filter.

8. The radar level gauge system integrating adaptive algorithm and multi-modal data according to claim 1, characterized in that: The anomaly detection is used to ensure the reliability of the system and the accuracy of the measurement results by identifying anomalies in the data, including sensor failure, environmental interference or system failure, and includes the following steps: (1) Residual calculation: Calculate the residual r(t) between the observed value and the predicted value: Among them: r(t): residual vector, reflecting the difference between the measured value and the predicted value; (2) Threshold setting: According to historical data and system requirements, the threshold ξ of the residual is set. When the absolute value of the residual exceeds the threshold, it is determined that an abnormality exists: If |r(t)|>ξ, it is considered abnormal; (3) Exception handling: When an abnormality is detected, the corresponding early warning mechanism is triggered, the abnormal information is recorded, or backup sensors and redundant systems are enabled to ensure the stable operation of the system.

9. The radar level gauge system integrating adaptive algorithm and multi-modal data according to claim 1, characterized in that: The adaptive calibration and drift compensation unit comprises: Adjustment of calibration factor k: k t =k t-1 +α·r(t) (17) Where: k t The calibration coefficient updated at the current moment, k t-1 The calibration coefficient of the previous moment, α calibration step, control adjustment speed, α>0, r(t) residual, Drift compensation: Adjustment of the calibration coefficient k achieves compensation for measurement drift; Estimated calibrated level: Use the updated calibration factor k t Calculate the calibrated liquid level estimate L KF (t) L cal (t)=k t ·L KF (t) (18) Where: L cal (t) Estimated value of liquid level after calibration, L KF (t) Liquid level estimate output by the Kalman filter.

10. The radar level gauge system integrating adaptive algorithm and multi-modal data according to claim 1, characterized in that: The display and alarm terminal includes the following functional modules: Data storage and management module: used to store the generated liquid level data, environmental parameters and historical records, and provide data query, statistics and analysis functions; Real-time display module: displays the current liquid level information, temperature, and pressure parameters in real time through an intuitive interface; Historical trend analysis module: provides historical data playback and trend analysis functions for liquid level and environmental parameters, and supports multi-dimensional display and comparison of data; Alarm and early warning module: according to the set threshold and abnormal detection results, when the liquid level exceeds the safe range or the system abnormality is detected, the sound and light alarm and information prompt will be issued in time; Human-computer interaction and parameter setting module: supports operators to set and adjust system parameters, including alarm threshold, display mode, and data collection frequency; Network communication and remote monitoring module: supports communication with host computers, control centers or mobile devices to achieve remote monitoring and data sharing.

Images