Environment feature based adaptive data denoising method applied to chemical industrial park

Abstract

The application relates to the technical field of environmental monitoring, and discloses an adaptive data noise reduction method based on environmental characteristics applied to a chemical industry park. The method comprises the following steps: S1. detecting the environmental characteristics of a static pressure water level gauge, and starting a cleaning mode of the static pressure water level gauge based on the environmental characteristics; S2. obtaining water level information based on the static pressure water level gauge, and obtaining turbidity information based on a turbidity sensor; and S3. correcting the turbidity information based on historical tide data, wherein the historical tide data comprises historical water level information and historical turbidity information. The application has the advantages of adapting to complex tide modes and environmental changes, significantly improving the accuracy and reliability of turbidity monitoring, and having certain abnormality detection and adaptive capacity.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of environmental monitoring, and in particular to an adaptive data denoising method based on environmental characteristics applied to a chemical industrial park. BACKGROUND

[0002] In a chemical industrial park, real-time monitoring of water quality and gas is a core requirement for ensuring environmental safety and compliance of factory operation. Among them, water quality monitoring as a key link directly affects the management effect of park emission control, pollution control and ecological environment protection. In a coastal chemical industrial park, turbidity as one of the important indicators of water quality, its monitoring has important significance for reflecting the concentration of suspended particles in water and potential pollution risk.

[0003] The change of the concentration of suspended particles in water may be caused by a variety of factors, including the superposition of natural environmental changes and human activities. For example, in a coastal chemical industrial park, if accidents such as storage tank leakage, pipeline rupture or rainwater flushing of emissions occur, the concentration of suspended particles in water may increase significantly. These accidents are usually accompanied by a large influx of silt, organic matter or chemical particles, which can quickly deteriorate water quality. In addition, potential pollutants such as untreated wastewater discharged during chemical production may directly affect turbidity levels. These wastewater may contain a large amount of undecomposed particles, toxic heavy metal ions or oil and fat substances, which not only cause turbidity to rise, but also pose a serious threat to the ecosystem of the water body and downstream water safety.

[0004] If the change of turbidity is not monitored and effectively controlled in a timely manner, it may cause a series of environmental and safety problems. For example, high turbidity water can significantly reduce the survival ability of aquatic plants and animals, and destroy the local ecological balance; suspended particles may adsorb pollutants and become carriers of pollutant diffusion, exacerbating environmental pollution in downstream areas; more seriously, these pollution may endanger the safety of drinking water for residents, and even cause a public health crisis. In addition, turbidity anomalies may also mask the presence of other more dangerous pollutants (such as chemical toxins or toxic gases) in the water body, leading to pollution incidents that cannot be identified and handled in a timely manner.

[0005] The special geographical environment of a coastal chemical industrial park further exacerbates the complexity of monitoring. For example, the water flow caused by tidal phenomena can significantly increase the concentration of suspended particles, causing abnormal readings of turbidity sensors. This abnormal reading may not only cause false alarms, wasting valuable emergency resources, but also may mask more serious pollution problems. In addition, if water level measurement devices are used to detect water level in real time (such as static pressure water level sensors), suspended and dissolved substances in tidal movements may also form dirt on the detection part of the water level measurement device, affecting the measurement accuracy of the device and causing inaccurate water level data. SUMMARY

[0006] In order to adapt to complex tidal patterns and environmental changes, significantly improve the accuracy and reliability of turbidity monitoring, and have certain abnormal detection and adaptive ability, the application provides an adaptive data noise reduction method based on environmental characteristics applied to a chemical industry park.

[0007] In the first aspect, the application provides an adaptive data noise reduction method based on environmental characteristics applied to a chemical industry park, which adopts the following technical scheme:

[0008] An adaptive data noise reduction method based on environmental characteristics applied to a chemical industry park comprises the following steps:

[0009] S1. Detect the environmental characteristics of the static pressure water level gauge, and start the cleaning mode of the static pressure water level gauge based on the environmental characteristics;

[0010] S2. Obtain the current water level fluctuation range information based on the static pressure water level gauge, and obtain the turbidity information based on the turbidity sensor;

[0011] S3. Correct the turbidity information based on historical tidal data, wherein the historical tidal data includes historical water level information and historical turbidity information.

[0012] Optionally, the static pressure water level gauge body comprises an outer shell, a piezoelectric ceramic, and a sensor arranged inside the outer shell, the outer shell is columnar, a communication port is formed in the bottom surface of the outer shell, and the communication port is used for allowing external fluid to enter to contact the sensor inside the outer shell; the piezoelectric ceramic is installed inside the static pressure water level gauge body, the top of the static pressure water level gauge is connected to an external host through an armored steel cable, and the piezoelectric ceramic is used for outputting ultrasonic vibration to the static pressure water level gauge body.

[0013] Optionally, the static pressure water level gauge body further comprises an ultrasonic probe and a convex ring located in the communication port, the ultrasonic probe is located inside the outer shell, and the ultrasonic probe is used for detecting the ultrasonic detection signal of the convex ring.

[0014] Optionally, the cross section of the outer shell is circular and gradually thickens from both ends to the middle.

[0015] Optionally, S1 comprises the following sub-steps:

[0016] S11. Control the piezoelectric ceramic to emit an ultrasonic measurement signal;

[0017] S12. Obtain the ultrasonic detection signal based on the ultrasonic probe, and perform dimension expansion on the ultrasonic detection signal; wherein the waveform of the ultrasonic detection signal is generated from the ultrasonic emission waveform and the physical properties of the interface, and carries relevant effective information related to the measured medium and the interface;

[0018] S13. Perform noise reduction processing on the ultrasonic detection signal;

[0019] S14. Perform non-local mean filtering;

[0020] S15. Input the non-local mean filtered ultrasonic detection signal into a pre-trained neural network model to obtain the thickness of the fouling; wherein the pre-trained neural network model takes the echo time and echo amplitude of the pre-calibrated ultrasonic reflection echo waveform as input and the calibrated thickness of the fouling as output for training, wherein the outer diameter, inner diameter, material average wall thickness of the pipeline are determined, the composition of seawater is determined, and the waveform of the ultrasonic reflection echo includes a pipeline outer wall echo peak, a pipeline outer wall secondary echo, a pipeline inner wall echo, a pipeline inner wall secondary echo, and a pipeline inner wall tertiary echo;

[0021] S16. Determine whether the thickness of the fouling is greater than a preset threshold value, and if so, start the piezoelectric ceramic to emit ultrasonic vibration for cleaning.

[0022] Optionally, the S12 comprises the following steps:

[0023] S121. Set the ultrasonic detection signal as wherein, , , , ;

[0024] S122. Perform dimension expansion on the one-dimensional ultrasonic detection signal , additionally add a random signal with a constant power spectral density as the original noise , and construct a virtual observation signal to assist in analysis, , i = 2, 3, …, M; is a random signal with the same distribution as ; constitute a virtual observation matrix P, , .

[0025] The S13 comprises the following steps:

[0026] S131. Decompose the virtual observation matrix P to obtain a square matrix ;

[0027] S132. Perform two-dimensional wave atom transformation on the square matrix to obtain a coefficient matrix ;

[0028] S133. Perform threshold filtering on the coefficients in the coefficient matrix to obtain a denoised coefficient matrix ;

[0029] S134. performing two-dimensional wavelet inverse transform on the noise-reduced coefficient matrix to obtain a filtered

[0030] S135. performing one-dimensional noise reduction on the noise-reduced result matrix to obtain a one-dimensional noise-reduced result of the signal

[0031] Optionally, the S3 comprises the following steps:

[0032] S31. determining a corresponding water level range in historical tide data according to a current water level fluctuation range;

[0033] S32. extracting historical turbidity information in the corresponding water level range from the historical tide database;

[0034] S33. analyzing a typical influence mode of tidal fluctuation on turbidity in combination with the historical tide data;

[0035] S34. selecting a correction coefficient based on the typical influence mode, and calculating a corrected turbidity based on the correction coefficient and the real-time turbidity.

[0036] In a second aspect, the present application provides an adaptive data denoising method based on environmental characteristics applied to a chemical industrial park, which adopts the following technical solution:

[0037] An adaptive data denoising device based on environmental characteristics applied to a chemical industrial park, comprising:

[0038] A cleaning module for detecting the environmental characteristics of the static pressure water level gauge and starting the cleaning mode of the static pressure water level gauge based on the environmental characteristics;

[0039] A data acquisition module for acquiring water level information based on the static pressure water level gauge and acquiring turbidity information based on the turbidity sensor;

[0040] A turbidity correction module for correcting the turbidity information based on historical tide data, wherein the historical tide data includes historical water level information and historical turbidity information.

[0041] In a third aspect, the present application provides a computer device, which adopts the following technical solution:

[0042] A computer device comprising a processor, wherein the processor runs a program of the adaptive data denoising method based on environmental characteristics applied to a chemical industrial park according to any one of the above.

[0043] In a fourth aspect, the present application provides a storage medium, which adopts the following technical solution:

[0044] ​​​​A storage medium storing a program for an adaptive data noise reduction method based on environmental characteristics, applicable to chemical industrial parks, as described in any one of the above-mentioned methods.

[0045] In summary, this application includes at least one of the following beneficial technical effects:

[0046] 1. By cleaning and compensating for environmental characteristics (such as scale and temperature) of the hydrostatic level gauge, measurement errors are effectively reduced, ensuring high accuracy of water level data. Simultaneously, by incorporating historical tidal data and real-time water level fluctuation analysis, and by calculating correction coefficients to dynamically adjust real-time turbidity values, abnormal turbidity deviations caused by tidal fluctuations are significantly reduced.

[0047] 2. The method incorporates the average water level and its rate of change, enabling adaptive adjustments for different typical patterns such as high tide, low tide, and non-tidal factors, thereby enhancing the system's adaptability to complex tidal environments. Attached Figure Description

[0048] Figure 1 This is a flowchart of an adaptive data noise reduction method based on environmental characteristics applied to a chemical industrial park, according to a certain embodiment of this application.

[0049] Figure 2 This is a longitudinal cross-sectional schematic diagram of the main body of the hydrostatic level gauge in one embodiment of this application.

[0050] Figure 3 This is a cross-sectional schematic diagram of the main body of the hydrostatic level gauge in one embodiment of this application.

[0051] Figure 4 This is a module connection diagram of an adaptive data noise reduction system based on environmental characteristics applied to a chemical industrial park, according to one embodiment of this application.

[0052] Figure label:

[0053] 11. Outer shell; 12. Raised ring; 13. Sensor; 102. Piezoelectric ceramic; 103. Connecting port; 5. Armored steel cable. Detailed Implementation

[0054] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0055] In the description of the present specification, the description referring to the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "example", "specific example" or "some examples" means that the particular feature, structure, material or characteristic being described in connection with the embodiment or example is included in at least one embodiment or example of the application. Descriptions of the above terms in the present specification do not necessarily refer to the same embodiment or example. Moreover, the particular features, structures, materials, or characteristics being described can be combined in any suitable manner in one or more embodiments or examples.

[0056] The embodiment of the present application discloses an adaptive data denoising method based on environmental characteristics applied to a chemical industry park, referring to Figure 1 , comprising S1-S3.

[0057] S1. Detect the environmental characteristics of the static pressure water level gauge, and start the cleaning mode of the static pressure water level gauge based on the environmental characteristics.

[0058] Due to the influence of tides, the water level will change frequently. In the early stage, the range of water level change is accurately obtained, so that it can be matched with historical tide data in the later stage. However, it is found in actual operation that the static pressure water level placed in seawater is easily affected by dirt or scale. Scale deposition can cover the pressure surface of the sensor, increasing the resistance of pressure transmission. And the blockage of the pressure surface may cause the pressure to be unable to be completely transmitted to the sensor chip, so that the measured value is low. In addition, scale itself has weight, which will increase the pressure on the surface of the sensor. If the scale deposition is heavy, it will cause the measured pressure value of the sensor to be high, resulting in the water level height being overestimated. The deposition of scale may cause the change of the density of the liquid nearby, especially the scale adhesion affects the fluidity, which may cause the local liquid density to be uneven, and then affect the accuracy of calculation. Scale deposition may also cause physical damage to the sensor, such as corrosion of the pressure diaphragm or blockage of the exhaust hole, so that the measurement error increases with time accumulation, or affects the zero drift of the sensor (change of the initial reference point of measurement), resulting in unstable long-term measurement data.

[0059] Therefore, the embodiment of the present application provides a static pressure water level gauge capable of self-cleaning, referring to Figure 2 and Figure 3The static pressure water level gauge body includes an outer shell 11 and a sensor 13 arranged inside the outer shell 11. Specifically, the outer shell 11 is a cylinder with an axis, and in different embodiments, the cross section of the outer shell 11 can have different shapes. As an example, the cross section of the outer shell 11 is circular. When the outer shell 11 is placed underwater, the tidal current will repeatedly impact the outer shell 11. The impact of the water flow has two aspects: on the one hand, the impurities contained in the tidal water will continuously deposit on the contact surface between the outer shell 11 and the water under the flow of the water flow; on the other hand, the impact of the water flow will accelerate the impurities on the surface of the outer shell 11 to separate, and the two reach a dynamic balance.

[0060] The top surface of the outer shell 11 is connected to the external host computer through the armored steel cable 5. The bottom surface of the outer shell 11 is provided with a communication port 103 for the external fluid to enter to contact the sensor 13 inside the outer shell 11. In different embodiments, the communication port 103 can have different shapes, such as circular, square or other shapes. In order to make the shape of the communication port 103 not affect the water flow into the inside of the outer shell 11, or make the water flow entering the outer shell 11 at different rotation angles not to be different, therefore, preferably, the communication port 103 is selected to be circular, and the channel inside the communication port 103 is cylindrical. The pressure input port of the sensor 13 is arranged on the side surface of the channel. In different embodiments, the sensor 13 can adopt different models, such as isolated diffusion silicon sensitive element or ceramic capacitor pressure sensitive sensor, to convert static pressure into electrical signal. Here, the input position capable of obtaining water flow pressure information is referred to as pressure input port.

[0061] The piezoelectric ceramic 102 is installed inside the static pressure water level gauge body. In different embodiments, the piezoelectric ceramic 102 can be started to vibrate according to the preset time interval through the intelligent control module, or started to vibrate according to the degree of dirt accumulation, to vibrate at a specific frequency and amplitude, so that the dirt on the surface of the sensor 13 falls off, keeps the surface of the sensor 13 clean, and ensures that the measurement accuracy is not affected by the dirt accumulation. At present, in some existing technical routes, the measurement of the degree of dirt accumulation can adopt laser detection or ultrasonic detection method or other methods. As an example, in the present scheme, a convex ring 12 surrounding the outer shell 11 is designed inside the communication port, which can not only reduce the influence of water flow fluctuation on the internal area of the communication port, but also provide echo for ultrasonic waves.

[0062] For piezoelectric ceramic 102, when an electric field is applied in the polarization direction of the crystal, the lattice in the crystal will be deformed by the internal stress under the action of the electric field force. Under the action of the alternating electric field, the internal stress and deformation of the crystal will change periodically, thereby generating mechanical vibration. Specifically, if pressure is applied to the piezoelectric ceramic 102, it will generate a potential difference, which is called the positive piezoelectric effect. Conversely, if voltage is applied, mechanical stress will be generated, which is called the inverse piezoelectric effect. If the pressure is a high-frequency vibration, a high-frequency current will be generated. When a high-frequency electric signal is applied to the piezoelectric ceramic 102, a high-frequency acoustic signal, i.e. mechanical vibration, is generated, which is called ultrasonic vibration.

[0063] Since the energizing position of the piezoelectric ceramic 102 is determined, the characteristic direction of the piezoelectric ceramic 102 is determined, and here the characteristic direction of the piezoelectric ceramic 102 is the vibration direction of the piezoelectric ceramic 102 at the corresponding energizing position. In an embodiment, the vibration direction of the piezoelectric ceramic 102 is designed to form a certain angle with the axis of the outer shell 11, so that the ultrasonic waves generated by the piezoelectric ceramic 102 can be used to clean the side surface of the outer shell 11. In another embodiment, the cross section of the outer shell 11 is circular and gradually thickens from both ends to the middle. In this way, the vibration direction of the piezoelectric ceramic 102 forms a certain angle with the outer surface of the outer shell 11, which can also play a cleaning role.

[0064] Specifically, on the scale of a natural day, the composition and content of the wastewater discharged by the chemical plant are relatively fixed every day, and the composition of seawater is also relatively fixed. However, due to the change in the speed of the water flow, such as the influence of tides, etc., the thickness of the dirt and the like formed is not uniform. If the piezoelectric ceramic is started to clean the static pressure water gauge at a fixed time every day, on the one hand, there will be a problem of over-cleaning, because the formation of the dirt layer is actually relatively slow. On the other hand, relatively frequent start of ultrasonic vibration, such as cleaning once a day, may adversely affect the internal components, such as loosening of the solder joints, etc., reducing the service life.

[0065] It should be noted that the causes and compositions of water scale or dirt are usually complex, and the formation characteristics of the dirt are a nonlinear function of multiple time-dependent parameters. If only theoretical numerical analysis is performed, it is difficult to quantify and the accuracy is difficult to estimate. Therefore, an embodiment is given. Since the piezoelectric ceramic itself emits ultrasonic signals, the appropriate frequency can be adjusted to detect the ultrasonic detection signal, thereby measuring the thickness of the dirt. This is because the ultrasonic detection signal carries a large amount of information related to the characteristics of the reflection surface. However, due to the presence of a large amount of noise and interference in the ultrasonic detection signal, specific steps are required to reduce the noise.

[0066] Specifically, S1 includes the following steps S11-S15.

[0067] S11. Control the piezoelectric ceramic to emit an ultrasonic wave measurement signal.

[0068] In this step, an ultrasonic wave signal is generated by the piezoelectric ceramic vibration, and the echo signal containing the sensor surface dirt information is obtained by using the physical characteristics of ultrasonic wave propagation, reflection, etc. in the elastic medium.

[0069] The piezoelectric ceramic is a core device that realizes the conversion of mechanical vibration and electrical signal by using the piezoelectric effect. When an electric field is applied in the polarization direction of the crystal, the crystal lattice in the piezoelectric ceramic will deform under the action of the electric field force, and high-frequency mechanical vibration will be generated with the periodic change of the electric field. This mechanical vibration further forms an ultrasonic wave signal in the surrounding medium. Ultrasonic wave is a mechanical vibration wave with a frequency higher than 20 kHz, and its frequency range is wide, usually a high frequency band (such as 10 MHz) is selected in underwater detection to obtain higher resolution and sensitivity. The propagation speed and reflection characteristics of the ultrasonic wave signal are closely related to the physical properties (such as density, acoustic impedance) of the propagation medium, therefore, the reflection echo contains rich information of the medium.

[0070] The specific process of controlling the piezoelectric ceramic to emit an ultrasonic wave signal includes the following stages. First, an alternating voltage of a certain frequency is generated by the control module to excite the piezoelectric ceramic, so that it vibrates and emits an ultrasonic wave signal. In this embodiment, in order to optimize the penetration ability and resolution of the signal, a working frequency of 10 MHz and a pulse width of 100 ns are selected. The emission of high-frequency ultrasonic waves can effectively improve the detection ability of thin layer dirt, while avoiding excessive signal attenuation. Subsequently, the ultrasonic wave propagates through the elastic medium (such as water) and is reflected when it encounters an interface between media (such as the dirt layer and the sensor surface). The time of arrival of the echo signal (i.e. the time of flight) is proportional to the distance of the interface, while the amplitude of the reflected signal reflects the difference in acoustic impedance between the media. According to these characteristics, the thickness of the dirt can be further calculated.

[0071] The ultrasonic sensor (i.e. ultrasonic probe) is a key component for realizing the emission and reception of ultrasonic signals. Its working principle is based on the inverse piezoelectric effect of the piezoelectric ceramic, which is driven by a high-frequency alternating current signal to produce mechanical vibration, converting electrical energy into mechanical wave energy and transmitting it to the medium. When the ultrasonic wave encounters an interface and is reflected, the mechanical vibration received by the ultrasonic probe is converted into an electrical signal again for system analysis. In this step, the design of the probe needs to ensure that its frequency matches the resonant frequency of the piezoelectric ceramic, in order to maximize the signal energy and reduce loss.

[0072] To further improve the spatial and temporal resolution of the ultrasonic signal, the transmitted pulse signal is in the form of a narrow pulse. This design helps to reduce the overlap of the echo signal, enabling the system to more accurately distinguish multiple interface reflection echoes. In addition, the width and frequency of the pulse signal also need to be considered in combination with the attenuation characteristics in the water flow environment and the detection depth requirement to ensure that the signal can effectively cover the target area.

[0073] S12. Obtain an ultrasonic wave detection signal based on the ultrasonic probe and perform dimension expansion on the ultrasonic wave detection signal; wherein the waveform of the ultrasonic wave detection signal is generated from an ultrasonic wave transmission waveform and physical characteristics of an interface, and carries a large amount of effective information related to the measured medium and the interface.

[0074] When controlling the piezoelectric ceramic to emit an ultrasonic wave measurement signal, the system needs to generate a specific alternating voltage to excite the piezoelectric ceramic to emit an ultrasonic wave signal, and after it receives the reflected echo, convert the mechanical vibration information into an electrical signal for processing.

[0075] In a specific implementation, the working frequency of the piezoelectric ceramic needs to be consistent with its resonance frequency to ensure that the generated ultrasonic wave energy is maximized and the measurement sensitivity is improved. As an example, when 10MHz is selected as the working frequency, the control circuit generates an excitation voltage with a frequency of 10MHz to drive the piezoelectric ceramic. Under the drive of this alternating voltage, the piezoelectric ceramic is converted into an ultrasonic pulse in an efficient manner.

[0076] To obtain sufficient time resolution and detection accuracy, the duration of the transmitted pulse is set to 100ns. This ensures that the pulse width can match the requirements of the measurement environment, improving the penetration of the signal and the sensitivity of the echo reception.

[0077] Specifically, S12 includes the following steps S121 and S122.

[0078] S121. Let the ultrasonic wave detection signal be wherein, , , , .

[0079] is the result of the ultrasonic wave signal transmitted by the piezoelectric ceramic after being reflected at the medium interface, received by the ultrasonic probe. is the effective ultrasonic echo information, reflecting the physical characteristics of the medium interface (such as the dirt layer, the sensor surface), including the thickness between the interfaces, the acoustic impedance, etc. k, α and β are parameters. is a noise signal, which can come from:

[0080] Medium scattering: When the ultrasonic wave encounters a non-smooth surface, it will generate a random scattering signal.

[0081] Environmental interference: e.g. mechanical vibration or electromagnetic noise.

[0082] System noise: signal acquisition and transmission noise from the electronic device itself.

[0083] Mathematically, the signal can be modeled as: . is the effective component of the signal, which can usually be decomposed into the superposition of multiple echoes: . Where P is the number of echoes, representing the multiple reflections of ultrasonic waves at different medium interfaces or the same interface. is the i-th echo signal, whose amplitude and phase reflect the characteristics of the medium interface. is the time delay of the echo, i.e. the time difference from the transmission to the return of the ultrasonic wave, reflecting the position or thickness of the medium interface. Since the ultrasonic wave comes from the same sound source, has obvious similarity.

[0084] is the noise component, which can be white noise or other types of random signals, whose characteristics are usually described by statistical distribution (such as Gaussian distribution).

[0085] S122. One-dimensional ultrasonic detection signal is dimensionally expanded, and additional random signals with the same distribution as the original noise with constant power spectral density are added, and virtual observation signals are constructed to assist analysis, , i=2, 3, …, M; are random signals with the same distribution as ; constitute a virtual observation matrix P, , M is the number of expanded observation signals, .

[0086] In one-dimensional signal , the noise may have partial overlap with the frequency domain characteristics of the effective signal g(t), making it difficult to separate by direct filtering. In this case, by adding random noise and constructing a matrix form, the spatial distribution characteristics of the two-dimensional signal can be utilized to enhance the ability to distinguish between signal and noise.

[0087] The constructed virtual observation signal is: , i=2, 3, …, M. is the original ultrasonic detection signal, is the same as The random signal with the same distribution, M is the number of extended observation signals. By this method, the generated virtual observation signal can retain the original signal characteristics, while diluting the correlation of the noise in the two-dimensional plane.

[0088] The M virtual observation signals are organized into a matrix P, P is an MxN matrix, N is the number of sampling points. The noise is randomly distributed in the matrix, while the effective signal g(t) has high similarity between the matrix lines. By constructing the matrix P, the original one-dimensional signal is converted into a two-dimensional representation, which provides a basis for subsequent extraction of effective signals using wave atom transform and other algorithms.

[0089] The waveform of the ultrasonic detection signal is generated by the transmission waveform of the ultrasonic wave and the physical properties of the medium interface, and contains information related to the medium and the medium interface, which needs to be preserved as much as possible in the process of reducing noise. The noise , which usually comes from the scattering of the medium to the ultrasonic detection signal, or from the interference of the external environment.

[0090] In this step, for the dimension M of the dimension expansion, the operation complexity and the noise reduction effect need to be selected. Here is used to select, and p should take a small integer, so that M is greater than the period length of the wave signal, as an example, here P=2.

[0091] S13. Perform noise reduction processing on the ultrasonic detection signal.

[0092] The virtual observation matrix P is a two-dimensional matrix constructed by the one-dimensional signal after dimension expansion. Each row represents a virtual observation signal , which is superimposed by the original signal and random noise . The key characteristics of the matrix are as follows:

[0093] 1. The original signal g(t) remains consistent in each row, forming a highly correlated regular structure between the matrix lines. This regularity is manifested as a highly regular texture in two-dimensional space, similar to a vertical oscillating texture.

[0094] 2. The added random noise is an independent signal with the same distribution as the original noise, with low correlation. The randomness of the noise is manifested as disordered distribution in the two-dimensional matrix, making it difficult to form a sustained structure or significant texture.

[0095] 3. There may be significant noise peaks in the one-dimensional signal, which are diluted by row and column distribution in the two-dimensional matrix, reducing their interference with the overall signal processing.

[0096] By regarding the elements of matrix P as the gray values of pixels, the signal can be represented in the form of a two-dimensional image. In this representation: the regularity of the effective signal makes the image exhibit an oscillatory texture in the vertical direction; the random noise appears as isolated points or regions randomly distributed in the image, with less prominence. The advantage of this two-dimensional representation is that: the characteristics of the signal are more intuitive and clear, facilitating processing using spatial filtering techniques (such as wavelet atom transform); the noise appears more scattered in the two-dimensional space, making it easier to separate.

[0097] Wavelet atom transform is a sparse representation method suitable for multi-dimensional signals, which can decompose the signal into basis functions with local characteristics and direction selectivity. By applying wavelet atom transform on the two-dimensional matrix, the separation effect of noise and signal is significantly improved:

[0098] Wavelet atom transform can decompose the two-dimensional matrix into a sparse coefficient matrix, and the main information of the signal is concentrated in a few large amplitude coefficients, while the noise corresponds to a more uniform distribution of coefficients with smaller amplitude. By threshold filtering (such as hard threshold or soft threshold filtering), large amplitude signal coefficients can be retained while low amplitude noise coefficients are removed.

[0099] The high correlation of the signal in the two-dimensional matrix makes it exhibit a concentrated energy distribution in the transformed coefficient matrix, facilitating separation from noise. Wavelet atom transform can also handle local changes in the signal, avoiding information loss in traditional Fourier transform.

[0100] Performing wavelet atom inverse transform on the filtered coefficient matrix, we get the denoised two-dimensional matrix. By matrix row and column averaging operation, the signal is further restored to one-dimensional form .

[0101] In ultrasonic detection, matrix P contains the superposition results of multiple virtual observation signals. Two-dimensional wavelet atom transform extracts the regular texture in the matrix (corresponding to the effective signal) and removes the randomly distributed isolated points (corresponding to the noise), significantly improving the quality of the signal.

[0102] Specifically, in an embodiment, S13 includes steps S131-S135.

[0103] S131. Decompose the virtual observation matrix P to get a square matrix .

[0104] In this step, several square sub-matrices are cut out from the virtual observation matrix P, laying the foundation for subsequent two-dimensional wavelet atom transform and denoising processing. This operation decomposes the complete matrix P into overlapping small matrices through framing technology, ensuring that the signal characteristics are fully preserved during the denoising process and avoiding data loss.

[0105] The reason for decomposing the virtual observation matrix P is that the original matrix P is usually large, and direct transformation and processing will lead to high computational complexity and large memory occupation. By decomposing the matrix into smaller square sub-matrices , the computational burden can be reduced, and the signal characteristics can be analyzed more effectively in a local range.

[0106] Specifically, frame division and overlap processing are required here.

[0107] Frame division: divide the columns of matrix P into several groups, and each group of columns forms a square sub-matrix .

[0108] Overlap: set a certain overlap degree η (usually 50%) between the columns of adjacent square matrices to ensure the continuity of information between adjacent frames and avoid the loss of boundary information caused by segmentation.

[0109] The formula for the square matrix is , where represents the k-th column of matrix P. M is the square side length of the matrix, , and k represents the starting column position of the matrix.

[0110] The number of square matrices c can be calculated by the following formula: , where N is the number of columns of matrix P.

[0111] For example, set the square matrix side length M and the overlap degree η, for example, M = 64 and η = 50%. Calculate the starting column k of each square matrix: . According to the size of k and M, extract the corresponding columns from matrix P to generate the square matrix . Repeat the above steps for each i until all columns of matrix P are processed.

[0112] S132. Perform two-dimensional wavelet atom transformation on the square matrix to obtain the coefficient matrix .

[0113] Two-dimensional wavelet atom transformation is a sparse representation technique that is particularly suitable for processing local and directional features contained in two-dimensional matrix data. It can decompose the matrix signal into a sparse coefficient matrix, where: the energy of the effective signal g(t) is concentrated in a small number of high-amplitude transformation coefficients; the noise h(t) is distributed in the entire domain of the transformation space, showing low-amplitude coefficients.

[0114] Through decomposition, the matrix can be represented as: . Where is the signal in the wavelet atom basis function the projection coefficients on the wavelet atom basis, denote the basis functions with localization and direction selectivity. The two-dimensional wavelet atom transform can capture the regular texture in the signal (e.g., oscillatory features in the vertical direction), while the noise appears as disordered high-frequency components.

[0115] Specifically, the square matrix is projected onto the wavelet atom basis to obtain the sparse coefficient matrix : . Where is the transformed sparse coefficient matrix, containing the distribution information of the effective signal and noise.

[0116] First, the fast Fourier transform (FFT) is applied to the matrix to convert the signal into the frequency domain representation: .

[0117] Then, the directionality and multi-scale analysis of the wavelet atom basis function are applied to the to decompose the signal into sparse coefficients: , where is the wavelet atom basis function, and j and m represent the scale and direction indices, respectively.

[0118] The non-zero coefficients in the are concentrated in a small number of directions and scales, representing the main information of the signal, while the coefficients of the noise are more dispersed and have lower amplitudes.

[0119] S133. Threshold filtering is performed on the coefficients in to obtain the denoised coefficient matrix .

[0120] By setting a threshold λ, coefficients with amplitudes below the threshold (mainly noise coefficients) can be removed, and coefficients with larger amplitudes (mainly signal coefficients) can be retained.

[0121] The threshold filtering formula is , where denotes the elements of the coefficient matrix ; λ is the threshold value used to distinguish between signal and noise.

[0122] For the value of λ, the threshold value can be dynamically adjusted according to the statistical properties of the coefficient matrix, for example, based on a Gaussian distribution noise model, using the mean or variance to determine λ:

[0123] , where σ is the standard deviation of the noise and N is the number of elements in the matrix.

[0124] Here, hard threshold filtering is used for filtering. Coefficients below the threshold are directly set to zero:

[0125] .

[0126] In this step, the effective signal is concentrated on a small number of high amplitude coefficients, and the key information of the signal can be preserved by filtering to ensure the integrity of the denoised signal.

[0127] S134. Perform two-dimensional wavelet atom inverse transform on the denoised coefficient matrix to obtain the filtered .

[0128] Perform two-dimensional wavelet atom inverse transform on the filtered to obtain the filtered . Wherein, is the sparse coefficient matrix after threshold filtering; represents the wavelet atom inverse transform operation; is the reconstructed denoised matrix. In the reconstruction process, the linear combination of wavelet atom basis functions is used to generate the signal: . Wherein, is the non-zero coefficient in ; is the wavelet atom basis function. Through this process, the denoised matrix signal retains the key characteristics of the effective signal while eliminating random noise.

[0129] In the reconstruction process, according to the non-zero coefficients in , the corresponding wavelet atom basis functions are selected for weighted combination. The directional and scale characteristics of each basis function ensure that the local features of the signal are restored. Since the threshold filtering has eliminated the low amplitude noise coefficients, the noise component is significantly suppressed during reconstruction, and the signal-to-noise ratio of the final generated matrix signal is greatly improved. The result of reconstruction is a two-dimensional matrix consistent in size with the original matrix , but the noise is eliminated and the signal is enhanced.

[0130] S135. Perform one-dimensional denoising on the denoised result matrix to obtain the one-dimensional denoised result .

[0131] In this step, the denoised square matrix processed in the previous step is recombined into a complete denoised matrix . This step solves the problem of data redundancy and boundary discontinuity introduced by the frame and overlap processing through weighted averaging, thereby generating a continuous and complete matrix.

[0132] The values of the plurality of square matrices in the overlapping area can be effectively fused by weighted average, the signal redundancy problem is solved, and the boundary transition is smoothed. The weighted formula is:

[0133]

[0134] wherein, is the pixel value of position (k, l) in the reconstruction matrix; S(k, l) is a set of square matrices covering position (k, l); is the square matrix is the weight of position (k, l).

[0135] In this embodiment, the weight Gaussian weighting is adopted. The weight is calculated using a Gaussian distribution function, the weight of the center of the overlapping area is larger, and the weight gradually decreases at the edge, and the formula is:

[0136] wherein σ controls the width of the Gaussian distribution.

[0137] S14. Non-local mean filtering is performed.

[0138] In the foregoing steps, the original one-dimensional signal f(t) is converted into a two-dimensional matrix P through dimension expansion, and a denoised complete matrix P' is generated through denoising processing. In order to be applied to actual scenes, the matrix signal after denoising needs to be restored to the same one-dimensional form f'(t) as the original signal.

[0139] Signal restoration is realized by weighted superposition of row data of the matrix P'. The specific operation includes the following steps:

[0140] Each row represents a signal copy generated in the virtual observation signal expansion process. By weighted accumulation of these rows, the main features of the original signal can be restored. The formula is as follows:

[0141] wherein M is the number of rows of the matrix; is the weighting coefficient of the i-th row. It is assumed that the weight of each row is the same, that is, The one-dimensional signal f'(t) after restoration is the final form of the denoised signal.

[0142] S15. A pre-trained neural network model is input to obtain the thickness of the dirt; wherein the pre-trained neural network model takes the echo time and echo amplitude of the pre-calibrated ultrasonic reflection echo waveform as input, and the calibrated thickness of the dirt as output to obtain the pre-trained neural network model, wherein the outer diameter, the inner diameter, the average wall thickness of the material of the pipeline are determined, the composition of seawater is determined, and the waveform of the ultrasonic reflection echo includes the pipeline outer wall echo peak, the pipeline outer wall secondary echo, the pipeline inner wall echo, the pipeline inner wall secondary echo, and the pipeline inner wall tertiary echo.

[0143] The training of neural network model is a quite mature technology and will not be described in detail here. It should be noted that, in the present embodiment, in order to provide abundant sample data, the sample set is provided in a simulated manner. Specifically, the sample set acquisition steps are as follows:

[0144] S151. Simulate a one-dimensional signal f'(t) in a real scene as an input signal; wherein, , A is the signal amplitude, corresponding to the characteristics of different thicknesses; f is the signal frequency, simulating the characteristics of the physical environment; is the phase, corresponding to the echo time; n(t) is the noise component, conforming to Gaussian distribution.

[0145] S152. According to the setting of signal amplitude and phase, generate the corresponding target output; wherein, the thickness , k1 and k2 are coefficients. The environmental parameters .

[0146] S153. Simulate the real signal by using a random number generator and a mathematical model, and introduce a certain range of random noise to simulate environmental interference.

[0147] S16. Determine whether the dirt thickness is greater than a preset threshold, and if so, start the piezoelectric ceramic to emit ultrasonic vibration for cleaning.

[0148] When the dirt thickness predicted by the neural network exceeds the set threshold, the ultrasonic cleaning operation of the static pressure water level gauge is automatically triggered.

[0149] S2. Obtain the current water level fluctuation range information based on the static pressure water level gauge, and obtain the turbidity information based on the turbidity sensor.

[0150] The static pressure information of the water body is collected in real time by the static pressure water level gauge, and the water level value is calculated according to the formula. The turbidity data of the water body is obtained by the turbidity sensor, reflecting the change of the suspended particulate matter concentration. The collected data of the static pressure water level gauge and the turbidity sensor are recorded synchronously, and the two types of data are synchronized by time stamp.

[0151] S3. Correct the turbidity information based on historical tide data, wherein the historical tide data includes historical water level information and historical turbidity information.

[0152] Specifically, in an embodiment, S3 includes the following steps S31-S34.

[0153] S31. According to the current water level fluctuation range, determine the corresponding water level range in the historical tide data.

[0154] The current water level fluctuation range is , wherein, The minimum water level for real-time monitoring, The maximum water level for real-time monitoring. This range can be calculated by a short time window of real-time water level information: , where, is the current water level, and w is the time window length (e.g., 10 minutes).

[0155] The water level fluctuation range of historical tidal data: , the matching condition is , that is, find all historical tidal records that overlap with the current water level fluctuation range. For example, if the current water level is 2.45m-2.50m, then the matching range is the historical water level interval [2.4m, 2.6m].

[0156] S32. Extract historical turbidity information corresponding to the water level range from the historical tidal database.

[0157] The historical tidal database stores the correlation information between water level and turbidity, usually including: sampling time of tidal data, historical water level value at corresponding time, turbidity value at corresponding time.

[0158] The data structure can be represented as: .

[0159] S33. Combine historical tidal data to analyze the typical influence mode of tidal fluctuation on turbidity.

[0160] Calculate the current average water level according to the water level fluctuation range, and determine the current typical influence mode according to the average water level and the change speed of the average water level.

[0161] According to the average water level and its change speed, combined with historical tidal data, determine the current typical influence mode:

[0162] In the rising tide influence mode, turbidity usually rises due to more suspended solids being stirred up by the water flow. In the falling tide influence mode, turbidity usually decreases due to the settling of particles. In the non-tidal influence mode, the change of turbidity is related to non-tidal factors, such as industrial emissions, rainfall, etc. Different typical influence modes correspond to different correction coefficients K, which are calibrated from historical data and stored in the database.

[0163] S34. Select correction coefficient based on typical influence mode, and calculate corrected turbidity based on correction coefficient and real-time turbidity.

[0164] According to the typical influence mode (rising tide, falling tide, non-tidal) determined by S33, find the corresponding correction coefficient database. Combine the current average water level and change speed to select the most matching correction coefficient K.

[0165] Use the formula to calculate the corrected turbidity.

[0166] Referring to Figure 4 The embodiment of the present application also discloses an adaptive data noise reduction device based on environmental characteristics applied to a chemical industry park, comprising:

[0167] a cleaning module, configured to detect environmental characteristics of the static pressure water level gauge, and start a cleaning mode of the static pressure water level gauge based on the environmental characteristics;

[0168] a data acquisition module, configured to acquire water level information based on the static pressure water level gauge, and acquire turbidity information based on the turbidity sensor;

[0169] a turbidity correction module, configured to correct the turbidity information based on historical tide data, wherein the historical tide data comprises historical water level information and historical turbidity information.

[0170] The embodiment of the present application also discloses a computer device comprising a processor, wherein the processor runs a program of the adaptive data noise reduction method based on environmental characteristics applied to a chemical industry park according to any one of the above.

[0171] The embodiment of the present application also discloses a storage medium, which stores the program of the adaptive data noise reduction method based on environmental characteristics applied to a chemical industry park according to any one of the above.

[0172] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. An environmental feature-based adaptive data denoising method applied to a chemical industrial park, characterized in that, The method comprises the following steps: S1. Detecting the environmental characteristics of the static pressure water level gauge, and starting the cleaning mode of the static pressure water level gauge based on the environmental characteristics; wherein the static pressure water level gauge body comprises an outer shell, a piezoelectric ceramic and a sensor arranged inside the outer shell, the outer shell is columnar, the bottom surface of the outer shell is provided with a communication port for the external fluid to enter to contact with the sensor inside the outer shell; the piezoelectric ceramic is installed inside the static pressure water level gauge body, the top of the static pressure water level gauge is connected to the external host through an armored steel cable, and the piezoelectric ceramic is used to output ultrasonic vibration to the static pressure water level gauge body; the static pressure water level gauge body further comprises an ultrasonic probe and a convex ring in the communication port, the ultrasonic probe is located in the outer shell, and the ultrasonic probe is used to detect the ultrasonic detection signal of the convex ring; the cross section of the outer shell is circular and gradually thickens from both ends to the middle; S2. Obtaining current water level fluctuation range information based on the static pressure water level gauge, and obtaining turbidity information based on the turbidity sensor; S3. Correcting the turbidity information based on historical tide data, wherein the historical tide data comprises historical water level information and historical turbidity information; The S1 comprises the following sub-steps: S11. Controlling the piezoelectric ceramic to emit an ultrasonic wave measurement signal; S12. Obtaining an ultrasonic detection signal based on the ultrasonic probe, and performing dimension expansion on the ultrasonic detection signal; wherein the waveform of the ultrasonic detection signal is generated from the ultrasonic wave transmission waveform and the physical characteristics of the interface, and carries relevant effective information related to the measured medium and the interface; S13. Noise reduction processing is performed on the ultrasonic detection signal; S14. Non-local mean filtering is performed; S15. The ultrasonic detection signal after non-local mean filtering is input into a pre-trained neural network model to obtain the thickness of the dirt; wherein the pre-trained neural network model takes the echo time and echo amplitude of the pre-calibrated ultrasonic reflection echo waveform as input, and the calibrated dirt thickness as output for training, wherein the outer diameter, inner diameter and average wall thickness of the pipe are determined, the composition of seawater is determined, and the waveform of the ultrasonic reflection echo comprises a pipe outer wall echo peak, a pipe outer wall secondary echo, a pipe inner wall echo, a pipe inner wall secondary echo and a pipe inner wall tertiary echo; S16. Determine whether the dirt thickness is greater than a preset threshold, and if so, start the piezoelectric ceramic to emit ultrasonic vibration for cleaning; The S12 comprises the following steps: S121. Let the ultrasonic probe signal be wherein, , , , ; is the result of the ultrasonic signal emitted by the piezoelectric ceramic being reflected after the medium interface is received by the ultrasonic probe; is the effective ultrasonic echo information, which is used to reflect the physical characteristics of the medium interface; is the noise signal; is the i-th echo signal, the amplitude and phase reflect the characteristics of the medium interface; is the time delay of the echo; M is the number of extended observation signals; N is the number of sampling points; S122. One-dimensional ultrasonic detection signal Dimensional expansion, additional and original noise The same distribution of power spectral density is constant random signal, and the virtual observation signal is constructed To assist in analysis, , i=2, 3, …, M; For the same distribution of random signals; The same distribution of power spectral density is constant random signal, and the virtual observation signal is constructed The virtual observation matrix P is composed of , ； The S13 comprises the following steps: S131. decompose the virtual observation matrix P to obtain a square matrix ; S132. The square matrix Performing a two-dimensional wavelet transform to obtain a coefficient matrix ; S133. threshold filtering coefficients in the coefficient matrix to obtain a denoised coefficient matrix ; S134. The coefficient matrix after noise reduction performing two-dimensional wavelet inverse transform to obtain the filtered ; S135. The result matrix after noise reduction Cumulative average in the longitudinal direction to obtain a one-dimensional noise reduction result of the signal .

2. The method for adaptive data denoising based on environmental characteristics applied to a chemical industrial park according to claim 1, characterized in that, The sample set of the pre-trained neural network model is obtained as follows: S151. Simulate a one-dimensional signal f'(t) in a real scenario as an input signal; wherein, , A is the signal amplitude, corresponding to features of different thicknesses; f is the signal frequency, simulating the characteristics of the physical environment; φ is the phase, corresponding to the echo time; n(t) is the noise component, conforming to the Gaussian distribution; S152. According to the setting of the signal amplitude and phase, the corresponding target output is generated; wherein, the thickness k1 and k2 are coefficients; S153. Simulate the real signal by using a random number generator and a mathematical model, and introduce random noise in a preset range to simulate environmental interference.

3. The method for adaptive data denoising based on environmental characteristics applied to a chemical industrial park according to claim 1, characterized in that, The S3 comprises the following steps: S31. According to the current water level fluctuation range, determine the corresponding water level range in the historical tide data; S32. Extract the historical turbidity information in the corresponding water level range from the historical tide database; S33. Analyze the typical influence mode of tidal fluctuation on turbidity in combination with the historical tide data; S34. Select a correction coefficient based on the typical influence mode, and calculate the corrected turbidity based on the correction coefficient and the real-time turbidity.

4. An environmental feature-based adaptive data denoising device applied to a chemical industrial park, characterized in that, The application is applied to the environment feature-based adaptive data noise reduction method for a chemical industry park as claimed in any one of claims 1-3, comprising: a cleaning module for detecting the environment feature where the static pressure water level gauge is located and starting the cleaning mode of the static pressure water level gauge based on the environment feature; a data acquisition module for acquiring water level information based on the static pressure water level gauge and acquiring turbidity information based on the turbidity sensor; a turbidity correction module for correcting the turbidity information based on historical tide data, wherein the historical tide data comprises historical water level information and historical turbidity information.

5. A computer apparatus, characterized by The application comprises a processor, wherein the processor runs the program of the environment feature-based adaptive data noise reduction method for a chemical industry park as claimed in any one of claims 1-3.

6. A storage medium, characterized by The application stores the program of the environment feature-based adaptive data noise reduction method for a chemical industry park as claimed in any one of claims 1-3.

Images