Micro-nano gas floatation solid-liquid separation control method based on scum density monitoring

Abstract

The application discloses a micro-nano air floatation solid-liquid separation control method based on scum density monitoring, and relates to the technical field of automatic water treatment control. The method comprises the following steps: obtaining scum layer terahertz scanning data, analyzing free water and bubble distribution to calculate a density index; calibrating the index by using physical viscoelastic parameters measured by an ultrasonic standing wave field; inputting the calibrated index and a feedforward signal into a joint prediction optimization model to solve optimal operation variables; and outputting fluid and residue discharge parameters, which are used for adaptive adjustment of the underlying fluid and phased array acoustic nondestructive residue discharge. The application is used to solve the problems of the traditional optical monitoring, such as the existence of a surface sensing blind area, the lack of physical confidence in pure data state evaluation, and the easy breaking of flocs and secondary backmixing caused by mechanical residue discharge.

Description

Technical Field

[0001] This invention relates to the field of water treatment automatic control technology, and more specifically, to a micro-nano air flotation solid-liquid separation control method based on scum density monitoring. Background Technology

[0002] Micro-nano air flotation technology is a core process for achieving efficient solid-liquid separation in modern water treatment engineering. Its operation is based on utilizing microbubbles to adhere suspended flocs, causing them to float and form a complex scum layer consisting of a solid-liquid-gas three-phase mixture. The efficient and stable operation of the air flotation process highly depends on the accurate monitoring of the scum layer density and the adaptive control of the scum discharge system. If the scum has a loose internal structure and is rich in free water, it will not only lead to a decrease in separation purity but also significantly increase the operational load of subsequent sludge dewatering processes. Therefore, establishing a high-fidelity scum state monitoring and closed-loop control system is crucial.

[0003] Currently, the mainstream control schemes for solid-liquid separation in air flotation mostly employ machine vision combined with empirical models. A common approach is to deploy optical cameras or infrared sensors above the liquid surface to capture the color, texture, and morphological features of the scum surface, and then use data-driven image recognition algorithms such as deep learning to indirectly assess the thickness and maturity of the scum, thereby controlling the traditional mechanical sludge scraper to perform contact-type scum discharge operations.

[0004] However, existing technologies have limitations. First, conventional optical sensors only have surface sensing capabilities and cannot penetrate the surface of scum to obtain the true spatial distribution of free water and air bubbles inside, resulting in a serious blind spot in the assessment of the internal structural compactness. Second, deep learning models based purely on data fitting lack the absolute constraint of microscopic physical rheological mechanisms, and are prone to statistical bias and "black box" prediction failure when faced with complex water quality shocks or fluctuations in operating conditions. Finally, the physical shear force applied by traditional mechanical scum scraping devices during scum discharge can easily cause secondary breakage of fragile flocs and backmixing and leakage of pollutants, making it difficult to meet the high standard of scum discharge requirements for zero-contact and non-destructive handling at the solid-liquid-gas multiphase interface. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a micro-nano air flotation solid-liquid separation control method based on scum density monitoring. This method uses terahertz penetrating scanning to visualize the internal structure of the scum and combines it with ultrasonic standing wave field for physical rheological calibration. Then, a joint predictive optimization model is used to solve the optimal operating variables to synergistically regulate the bottom fluid and phased array acoustic scum discharge. This addresses the technical problems of traditional optical monitoring having a surface perception blind zone, pure data state assessment lacking physical confidence, and conventional air flotation control systems having insufficient adaptive accuracy and low scum discharge purity.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A micro-nano air flotation solid-liquid separation control method based on scum density monitoring includes the following steps: acquiring terahertz penetration scan echo data of the scum layer and analyzing the distribution profile characteristics of free water and bubbles; calculating the density index characterizing the compactness of the three-dimensional structure inside the scum layer based on the profile characteristics; acquiring physical viscoelastic parameters measured based on ultrasonic standing wave fields and calibrating the density index using the viscoelastic parameters; inputting the calibrated density index and the feedforward signal of the influent into a joint prediction optimization model to solve for the optimal combination of operating variables that satisfies multi-objective constraints; the combination of variables includes at least fluid control parameters and scum discharge execution parameters; and outputting the fluid control parameters and scum discharge execution parameters for adaptive adjustment of the bottom fluid and scum discharge operation of the phased array acoustic emission device, respectively.

[0007] In a preferred embodiment, acquiring terahertz transmission scan echo data of the scum layer and analyzing the distribution profile features of free water and bubbles includes: acquiring terahertz echo signals reflected and transmitted by the scum layer; analyzing the dielectric constant differences of the scum layer at different depth profiles based on the time-domain delay and amplitude attenuation of the echo signals; and reconstructing the three-dimensional distribution profile features of free water and bubbles within the scum layer based on the dielectric constant differences.

[0008] In a preferred embodiment, the calculation of the density index, which characterizes the compactness of the three-dimensional structure inside the scum layer, and the calibration using the viscoelastic parameter, includes: processing the profile features using a pre-trained deep fusion network to output an initial density index; using the viscoelastic parameter as a physical anchor point to calibrate the weights of the deep fusion network, and outputting the calibrated density index.

[0009] In a preferred embodiment, before inputting the calibrated density index and the feedforward signal of the influent into the joint prediction and optimization model, the method further includes: obtaining the low-field nuclear magnetic resonance scanning spectrum of the mixed fluid in the separation zone of the flotation tank; quantifying the microscopic physical binding force state between micro- and nano-bubbles and pollutant flocs based on the displacement characteristics of the relaxation peaks in the scanning spectrum; and inputting the binding force state as a precondition into the joint prediction and optimization model.

[0010] In a preferred embodiment, the joint prediction optimization model includes a machine learning prediction model based on feature importance attribution and a model predictive control optimizer. The process of obtaining the optimal combination of operands that satisfies multi-objective constraints includes: using the machine learning prediction model, combined with the binding force state, to predict in real time the marginal contribution weights of multidimensional adjustment variables to maintaining the density index; and the model predictive control optimizer using the marginal contribution weights as a search guide to perform multi-objective rolling optimization under preset operating cost and effluent water quality constraints to obtain the optimal combination of operands.

[0011] In a preferred embodiment, the fluid control parameters include the reagent dosage concentration and the multiphase flow dissolved air pump speed; the output fluid control parameters for adaptive adjustment of the bottom fluid include: outputting the reagent dosage concentration to the flocculant metering pump to adjust the stroke frequency; and outputting the multiphase flow dissolved air pump speed to adaptively adjust the dissolved air-water reflux ratio.

[0012] In a preferred embodiment, outputting slag discharge execution parameters for slag discharge operation of the phased array acoustic transmitter includes: determining the target slag discharge thrust and sound field focusing path based on the slag discharge execution parameters; generating control signals for dynamically adjusting the emission phase and amplitude of each transducer element in the phased array acoustic transmitter to excite directional acoustic traveling waves at the interface.

[0013] In a preferred embodiment, the process of outputting slag discharge execution parameters further includes: determining in real time whether the calculated density index is greater than or equal to a preset slag discharge critical threshold; if the density index is less than the slag discharge critical threshold, generating a sleep command and sending it to the phased array acoustic transmitter; if the density index is greater than or equal to the slag discharge critical threshold, generating a slag discharge command and sending it to the phased array acoustic transmitter to generate directional acoustic traveling waves.

[0014] In a preferred embodiment, the process of multi-objective rolling optimization under preset operating cost and effluent water quality constraints further includes: acquiring the environmental noise level and effluent suspended solids concentration; when the environmental noise level exceeds a preset safety threshold, reducing the tracking weight of the density index and narrowing the constraint boundary of the effluent suspended solids concentration, and dynamically updating the cost function of the model predictive control optimizer to ensure that the effluent water quality meets the preset standards.

[0015] This invention provides a micro / nano air flotation solid-liquid separation control system based on scum density monitoring, comprising: a feature analysis module for acquiring terahertz transmission scan echo data of the scum layer and analyzing the distribution profile features of free water and bubbles; a density calculation module for calculating a density index characterizing the compactness of the three-dimensional structure inside the scum layer based on the profile features; a density calibration module for acquiring physical viscoelastic parameters measured based on ultrasonic standing wave fields and calibrating the density index using the viscoelastic parameters; a parameter optimization module for inputting the calibrated density index and the feedforward signal of the influent into a joint prediction optimization model to solve for the optimal combination of operating variables that satisfies multi-objective constraints; the variable combination includes at least fluid control parameters and scum discharge execution parameters; and a solid-liquid separation module for outputting the fluid control parameters and scum discharge execution parameters for adaptive adjustment of the bottom fluid and scum discharge operation of the phased array acoustic emission device, respectively.

[0016] The technical effects and advantages of this invention's micro-nano air flotation solid-liquid separation control method based on scum density monitoring are as follows: This invention introduces terahertz penetration scanning technology to deeply analyze the spatial distribution characteristics of free water and air bubbles within the scum layer, breaking through the blind spot of traditional monitoring technologies that can only focus on the surface of the scum. This achieves high-fidelity quantification of the compactness of the three-dimensional structure within the scum. Simultaneously, it utilizes ultrasonic standing wave fields to obtain real physical viscoelastic parameters for mechanistic-level calibration of the density index, effectively eliminating the bias of simple state assessment and significantly improving the physical confidence of the parameters. Furthermore, the calibrated density index and the feedforward signal of the influent are jointly input into a joint prediction and optimization model to solve for the optimal operating variables under multi-objective constraints in real time, and to link the adaptive control of the bottom multiphase fluid with the non-destructive scum removal operation of the top phased array acoustic device. This invention constructs a closed-loop control system of "internal perspective and physical calibration perception—multi-objective rolling optimization prediction—coordinated execution of fluid and acoustic scum removal," improving the adaptive control accuracy, system shock resistance robustness, and purity of scum removal operations in the air flotation solid-liquid separation process. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the micro-nano air flotation solid-liquid separation control method based on scum density monitoring provided in an embodiment of the present invention; Figure 2 This is a diagram showing the terahertz wave frequency domain amplitude attenuation curves at different depths of the scum layer provided in an embodiment of the present invention. Figure 3 The deformation relaxation response curve of the slag sample based on microacoustic standing waves is provided in the embodiment of the present invention. Figure 4 This is a comparison of the low-field nuclear magnetic resonance relaxation time spectrum of the air-flotation mixed fluid provided in an embodiment of the present invention; Figure 5 This is a block diagram of a micro-nano air flotation solid-liquid separation control system based on scum density monitoring, provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1, Figure 1 This invention presents a micro / nano air flotation solid-liquid separation control method based on scum density monitoring, comprising the following steps: S1, acquire terahertz penetration scan echo data of the scum layer, and analyze the profile characteristics of free water and bubble distribution.

[0020] In this embodiment, acquiring terahertz penetration scan echo data of the scum layer and analyzing the distribution profile characteristics of free water and bubbles includes: S101, Acquire the terahertz echo signals reflected and transmitted by the scum layer; as a preferred embodiment, a terahertz transceiver antenna array is suspended and installed via a seismic support at a position approximately 30 cm to 50 cm directly above the liquid surface in the flotation tank separation zone. The operating frequency band of this array is preferably configured between 0.1 THz and 3.0 THz. Within this specific frequency band, terahertz waves possess extremely unique physical propagation and attenuation characteristics: they have extremely high penetration for non-polar substances (such as air inside micro / nanobubbles and non-polar organic pollutants), and the Rayleigh scattering effect is weak; however, for polar molecules (especially liquid free water molecules), the torsional and vibrational modes of their intermolecular hydrogen bond networks fall precisely within the terahertz band, thus exhibiting strong dipole moment resonance absorption characteristics. This significant physical difference based on phase dielectric properties allows the terahertz pulse to accurately capture the interface reflection echo and transmission energy attenuation signals caused by the overlapping of gas, liquid, and solid phases when penetrating vertically downwards through the complex scum interface.

[0021] S102, based on the time-domain delay and amplitude attenuation of the echo signal, the difference in dielectric constant of the scum layer at different depth profiles is analyzed; further, after capturing a continuous terahertz waveform sequence containing multipath effects, signal decoupling and time-domain windowing operations are performed. Specifically, utilizing the time-domain delay characteristics generated by the reflection of the echo signal at interfaces of different dielectric impedances (such as gas-scum interfaces, scum-water interfaces), combined with the propagation speed loss of light in different phase states, the corresponding spatial depth coordinates of the scum are calculated. Simultaneously, a Fourier transform is performed on the acquired signal to extract the amplitude attenuation characteristics of specific absorption peaks in the frequency domain. Due to the spatial heterogeneity of bubble embedding density and free water retention at different depths of the scum layer, the terahertz wave exhibits a stepped absorption loss along the propagation path. By inverting this amplitude attenuation, the water content and bubble porosity at the corresponding depth can be quantified, thereby calculating the difference in dielectric constant of the scum layer sliced ​​layer by layer in the vertical direction. Figure 2 The figure shows a simulation diagram of the frequency domain amplitude attenuation curve of the terahertz transmitted wave in the scum layer under a typical working condition in this embodiment. The horizontal axis of the figure represents the frequency (0.1~3.0 THz), and the vertical axis represents the normalized amplitude. The curves clearly show that as the depth of the probe slice advances from the surface layer (z=0) of the scum to the bottom layer (z=15cm), the absorption peak area of ​​the terahertz wave at the hydrogen bond resonance frequency of water molecules (around 1.5 THz) increases significantly and nonlinearly due to the higher concentration of free water in the bottom layer. This simulation result intuitively demonstrates that by extracting the characteristic attenuation of a specific frequency band, the vertical distribution of polar water and non-polar bubbles inside the scum can be inverted with high confidence.

[0022] The absorption coefficient and complex permittivity of the terahertz echo signal are calculated using the following formulas: Absorption coefficient calculation formula: (1) Refractive index calculation formula: (2) Formulas for solving the real and imaginary parts of the complex permittivity: (3) (4) in, The angular frequency of the incident terahertz wave; Calculate the penetration thickness of the profile layer for the target; The amplitude of the frequency domain sample signal after penetrating the slag layer slice; The amplitude of the frequency domain reference signal measured in free space; The absorption coefficient of a substance is highly linearly sensitive to the free water content. This is the time-domain delay of the echo signal relative to the reference signal, used to map depth coordinates; The speed of light in a vacuum; This represents the equivalent refractive index at the corresponding depth. The real part of the dielectric constant represents the dispersion capability of the material at this depth to terahertz waves and its polarization energy storage characteristics. The imaginary part of the dielectric constant represents the electromagnetic wave energy loss caused by dipole polarization relaxation inside the medium.

[0023] S103, based on the aforementioned dielectric constant difference, reconstruct the three-dimensional distribution profile of free water and bubbles within the scum layer. In this embodiment, the core of the reconstruction lies in using a three-dimensional total variational regularized (3D-TV) tomography algorithm to perform inverse mapping on the slice data. The input data obtained by the algorithm is a series of two-dimensional dielectric constant slice matrices at different depth profiles obtained in step S102. During the algorithm initialization phase, the voxel space resolution parameter of the three-dimensional mesh is set (preferably set to...). The two-dimensional slice matrix is ​​then mapped to the initial three-dimensional tensor space. To suppress signal noise caused by multiple scattering in the multiphase flow medium, an iterative reconstruction logic with 3D-TV regularization constraints is used to update the dielectric constant tensor. The update formula is as follows: (5) in, For the first The three-dimensional dielectric constant tensor of the next iteration; is the relaxation factor of the algorithm; This is the spatial forward projection mapping matrix. Its inverse projection matrix; Two-dimensional slice data input from actual observations; The parameter is the total variational regularization penalty parameter; It is a three-dimensional spatial gradient operator.

[0024] After the algorithm outputs a stable three-dimensional dielectric constant tensor, a soft-threshold phase mapper based on the sigmoid function is set up to transform it into a specific physical phase distribution, and the volume fraction probabilities of free water and bubbles are calculated: (6) in, Representation in three-dimensional coordinate space The volume percentage of the probability that free water is present at a given location; To reconstruct the real part of the corresponding dielectric constant output by the algorithm; The preset critical dielectric constant separation threshold for the gas-liquid-solid three-phase interface; This is the smoothing gain coefficient of the mapping curve.

[0025] By introducing terahertz high-frequency penetration scanning and a total variational regularized tomographic reconstruction mechanism, the physical blind spot of traditional optical or infrared monitoring technologies, which can only perceive the surface, is overcome. This enables non-destructive imaging measurement of the deadly liquid water cavities and hidden bubble embedding state inside scum. It solves the control lag problem caused by the invisibility of the internal rheological state, providing solid and high-dimensional physical state data support for the subsequent accurate calculation of the density index.

[0026] S2, based on the profile features, calculate the density index, which characterizes the compactness of the three-dimensional structure inside the scum layer; obtain the physical viscoelastic parameters measured based on the ultrasonic standing wave field, and use the viscoelastic parameters to calibrate the density index.

[0027] In this embodiment, the calculation of the density index, which characterizes the compactness of the three-dimensional structure inside the scum layer, and the calibration using the viscoelastic parameters, includes: S201, the profile features are processed using a pre-trained deep fusion network to output an initial density index; as a preferred embodiment, a deep fusion network based on a three-dimensional convolutional neural network (3D-CNN) is constructed. The input to this network is the three-dimensional distribution profile feature tensor of free water and bubbles reconstructed in step S103 (e.g., with a fixed size). The network hierarchy consists of alternating stacks of multi-level 3D convolutional layers and 3D max-pooling layers, aiming to extract high-dimensional topological features of bubble mosaic density and free water sac encapsulation from a spatial dimension. After flattening, the feature maps are connected to a fully connected layer, and a scalar value ranging from 0 to 100 is output through a Sigmoid activation function. This scalar is the initial compactness index, used to preliminarily characterize the macroscopic compactness of the scum layer's micro-network from a pure data manifold perspective.

[0028] S202, using the viscoelastic parameters as physical anchors, the weights of the deep fusion network are calibrated, and the calibrated density index is output. Further, scum samples are diverted from the bypass of the flotation tank and measured online using a non-contact microrheological detection unit (such as an ultrasonic rheological analyzer) based on an ultrasonic standing wave field. Specifically, a piezoelectric phased array of transducers with operating frequencies from 2.0 MHz to 5.0 MHz is symmetrically arranged inside the detection unit, and a stable spatial ultrasonic standing wave field is excited through interference superposition. The extraction volume of the bypass microfluidic sampling valve is controlled to be approximately 10... Up to 50 A small sample of scum droplets is suspended in the center of the detection area without contact using acoustic radiation pressure at the nodes of a standing wave field. Subsequently, the system instantaneously increases the excitation voltage of the transducer to generate high-frequency converging sound pressure, applying non-destructive dynamic micro-compression to the suspended scum sample, causing it to undergo ellipsoidal deformation. At the instant of sound pressure unloading, a high-frequency optical sensing module continuously captures the deformation recovery process of the sample under the influence of the elasticity and surface tension of the internal cross-linked flocs, calculating the "relaxation time" and the true Young's modulus (i.e., the physical viscoelastic parameter) characterizing the sample's physical rheological properties. Figure 3 The figure shows the simulation curve of the deformation recovery time-series response of scum droplets obtained based on ultrasonic suspension microrheological monitoring. The horizontal axis represents time (ms), and the vertical axis represents the major and minor axis deformation differences of the droplets under optical field of view. The curve trajectory shows that during the high-frequency focusing sound pressure phase (0~50 ms), the deformation difference rapidly climbs to its peak; after the sound pressure is instantaneously removed (after 50 ms), the droplets exhibit a typical viscoelastic exponential decay recovery trend under the elastic force within the cross-linked flocculants. By curve fitting the slope of this exponential decay segment, the deformation relaxation time constant, unaffected by optical noise, can be accurately extracted, thus providing an absolute physical benchmark for deep networks.

[0029] Finally, the viscoelastic parameter is introduced as a physical anchor into the dynamic loss function of the deep fusion network, and the weight matrix of the fully connected layer of the 3D-CNN is fine-tuned online through the backpropagation mechanism, thereby outputting the density index calibrated by the physical rheological mechanism.

[0030] The formula for extracting the physical viscoelasticity (Young's modulus and deformation relaxation time) is as follows: (7) (8) in, The relaxation time for deformation recovery of the scum sample; This refers to the absolute deformation difference (i.e., the difference between the major and minor axes) of the sample captured by the optical sensing module when it is under maximum pressure. After removing the sound pressure Transient deformation difference at any given moment; The equivalent Young's modulus is used to characterize the physical viscoelastic parameters of the sample. The system acoustic coupling constant of the microacoustic testing chamber; The peak value of the transient acoustic radiation pressure exerted on the droplet by the ultrasonic standing wave field; Let be the initial equivalent spherical radius of the scum sample droplet in an unpressurized suspended state.

[0031] The dynamic loss function formula is as follows: (9) in, To include weight parameters after introducing physical anchor points The dynamic calibration loss function; This represents the number of sample batches within the sliding data window. and These are the predicted index and historical benchmark labels output by the network, respectively. The dynamic weighting coefficient for the physical anchor point penalty term; The initial density index is the output of the network's real-time forward propagation at the current moment; This is a mechanism mapping function used to map the absolute viscoelastic parameters obtained through acoustic deformation measurements. Nonlinear mapping is applied to the same dimensional scale as the density index.

[0032] This implementation method achieves non-destructive online quantification of the macroscopic phase mechanical properties of trace amounts of scum through a physical anchor point calibration mechanism in acoustic tweezers microrheological measurements. This method avoids the prediction failure caused by "black box drift" in purely data-driven AI models under extreme water quality shocks or complex nonlinear multiphase flow conditions, thereby improving the adaptive robustness and physical confidence of the air flotation separation control system.

[0033] S3. Input the calibrated density index and the feedforward signal of the influent into the joint prediction and optimization model, and solve for the optimal combination of operating variables that satisfies the multi-objective constraints; the combination of variables includes at least fluid control parameters and slag discharge execution parameters.

[0034] Furthermore, before inputting the calibrated density index and the feedforward signal of the influent into the joint prediction and optimization model, the method further includes: obtaining the low-field nuclear magnetic resonance scanning spectrum of the mixed fluid in the separation zone of the flotation tank; quantifying the microscopic physical binding force state between micro-nano bubbles and pollutant flocs based on the displacement characteristics of the relaxation peaks in the scanning spectrum; and inputting the binding force state as a precondition into the joint prediction and optimization model.

[0035] In this embodiment, as a preferred implementation, a mixed fluid sample is bypassed in the separation zone of the flotation tank and subjected to online non-destructive testing using low-field nuclear magnetic resonance (NMR) scanning technology. Specifically, hydrogen protons in the fluid sample are radio-frequency excited by applying a specific sequence of continuous radio frequency pulses (such as a CPMG spin echo sequence), and then the spin echo signals generated by spin relaxation in an applied static magnetic field are continuously acquired. At the quantum spin level, the transverse relaxation time of hydrogen protons in water molecules (… The relationship between micro- and nano-bubbles and pollutant flocs is highly dependent on their microscopic physical confinement state. When micro- and nano-bubbles strongly adhere and associate with pollutant flocs through hydrophobic interactions, the originally free water is squeezed out, and "bound water" tightly bound by physicochemical forces forms at the interface. This confinement state greatly restricts the degree of freedom of hydrogen proton movement, causing it to... The relaxation peak in the spectrum exhibits a significant left shift (i.e., the relaxation time becomes shorter). This can be observed by analyzing a broad range of... By inverting and fitting the relaxation time spectrum, and extracting the area of ​​the short relaxation peak on the left (representing tightly bound water) and its leftward shift on the time axis, the microscopic bonding strength of bubbles and flocs can be quantified with high precision without completely destroying the fluid physics structure. For example... Figure 4 As shown, the low-field NMR spectra of the air-flotation mixed fluid sample at different control stages are presented. A comparison of transverse relaxation time spectra. The horizontal axis represents relaxation time (0.01~10000 ms), and the vertical axis represents the amplitude of the NMR spin signal. The solid line represents the spectrum under the conventional unoptimized state, and the dashed line represents the spectrum optimized by the joint model of this invention. The comparison shows that after optimization, the area of ​​the short relaxation peak (1~10 ms interval) on the left, representing "bound water," is significantly increased, and the peak position is noticeably shifted to the left, while the amplitude of the long relaxation peak (>100 ms) on the right, representing "free water," is significantly reduced. This data visualization confirms from the quantum spin level that this control method can effectively drive free water to transform into bound water, promoting the dense microscopic embedding of bubbles and flocs.

[0036] The low-field nuclear magnetic resonance The relaxation time characteristics and the derived formula for extracting the microscopic binding force state are as follows: (10) (11) in, This is the overall magnetization vector attenuation signal of the low-field nuclear magnetic spin echo; For the first The signal amplitude of each relaxor component (corresponding to the number of hydrogen protons); This is the corresponding transverse relaxation time constant; This is a background noise signal; It is the quantified microscopic physical bonding force state index; and This constitutes a short relaxation time integration interval, representing the water that is tightly bound by the interface; and This constitutes a long relaxation time integration interval, representing free water in the environment; The coordinates of the standard transverse relaxation peak position of free water in pure water; The transverse relaxation time coordinates of the main peak of bound water in the actual measured spectrum.

[0037] In this embodiment, the joint prediction optimization model includes a machine learning prediction model based on feature importance attribution and a model prediction control optimizer; the solution to obtain the optimal combination of operational variables that satisfies the multi-objective constraints includes: S301, utilizing a machine learning prediction model and combining the aforementioned binding force state, the marginal contribution weights of multidimensional adjustment variables to maintaining the density index are predicted in real time; further, a machine learning prediction model based on gradient boosting trees (such as the LightGBM architecture) is constructed. Unlike traditional statistical mappings that rely solely on macroscopic indicators such as influent turbidity, this model incorporates the aforementioned microscopic binding force state measured based on quantum relaxation. The influent feedforward signal, along with other inputs, is forced to serve as a high-dimensional feature input. In practice, the training process of the LightGBM prediction model includes: firstly, constructing a supervised learning dataset, and extracting the influent feedforward signal and binding force state from the historical operating conditions of the flotation tank. The model uses multidimensional operational variables as feature matrices and the ground truth index, calibrated in step S2 at the corresponding time point, as the ground truth. During training, the model employs a histogram-based decision tree algorithm for splitting and growth, selecting mean squared error (MSE) as the optimization objective. Gradient descent minimizes the residual between the predicted density and the ground truth label. The core structural parameters of the model are set as follows: a learning rate to control the error fitting step size of a single tree, and a maximum tree depth and minimum number of leaf node samples to limit model complexity and prevent overfitting. As a preferred example of specific parameters in this embodiment, the learning rate of the LightGBM prediction model is set to 0.01 to 0.05 (preferably 0.02 in this embodiment), the maximum tree depth is set to 5 to 8 layers (preferably 6 layers in this embodiment), the minimum number of leaf node samples is set to 20 to 50 (preferably 30 in this embodiment), and the total number of base learner iterations (i.e., the number of decision trees generated) is set to 200 to 500 (preferably 300 in this embodiment). Using the Shapley Additive Interpretation (SHAP) algorithm, the marginal contribution of each multidimensional adjustment variable to driving the scum layer to reach the density index is calculated in real time under the current specific water quality and binding force state, and then a physically oriented SHAP weight vector is dynamically generated.

[0038] The SHAP weight vector The basic attribution solution and diagonalization mapping formulas are as follows: (12) (13) in, For the first The basic SHAP feature attribution values ​​corresponding to each moderating variable; It is the set of all input features, including the state of binding force. The total number of features; To exclude the first The feature subset following each feature; The expected output exponent of a machine learning prediction model given a subset of feature inputs; This represents the total dimension of the variables ultimately input to the controller. The bottom of the formula uses Softmax nonlinear normalization on the absolute-valued base SHAP values ​​to transform the marginal contributions into diagonal elements with values ​​between (0, 1) and a sum of 1, thus constructing a smooth and dimensionally uniform SHAP weight diagonal matrix. .

[0039] S302, the model predictive control optimizer uses the marginal contribution weights as a search guide to perform multi-objective rolling optimization under preset operating cost and effluent quality constraints, solving for the optimal combination of operating variables. As a preferred implementation, the model predictive control (MPC) algorithm within the finite prediction time domain is used for online solving. The MPC solver uses the generated SHAP weight diagonal matrix... As a search guidance factor in the control increment matrix, the optimizer prioritizes gradient descent search along the variable with the largest marginal contribution during the rolling prediction time domain. The solution process is simultaneously constrained by multiple physical hard boundaries, such as preset energy consumption (e.g., multiphase pump power consumption) and drug consumption.

[0040] The set of constraint inequalities for the multiple physical hard boundaries is as follows: (14) (15) (16) (17) in, and These represent the absolute state of the operational variables and the control increment within the prediction time domain, respectively. and This represents the physical saturation limit of the underlying actuators. This is a transient power consumption estimation model for the multiphase pump in the system. This is the preset peak energy consumption boundary; This refers to the calculated consumption of flocculant. This is the preset maximum dosage limit for the drug.

[0041] After solving the quadratic programming problem, the optimal combination matrix of operational variables for the first control cycle is output. : (18) in, To optimize the dosage of the reagent, To optimize the speed of the multiphase dissolved air pump, the two together constitute the fluid control parameters, which can be directly used for the adaptive adjustment of the multiphase fluid at the bottom of the flotation tank. The system extracts the combined control parameters, including the target acoustic field thrust and the phase of the focusing array elements, for the operation of the top-level phased array acoustic device. This optimization process is repeated as the time window slides forward.

[0042] Furthermore, the multi-objective rolling optimization process under preset operating cost and effluent quality constraints also includes: acquiring the environmental noise level and effluent suspended solids concentration; when the environmental noise level exceeds a preset safety threshold, reducing the tracking weight of the density index and narrowing the constraint boundary of the effluent suspended solids concentration, and dynamically updating the cost function of the model predictive control optimizer to ensure that the effluent quality meets the preset standards. In this embodiment, the environmental noise caused by strong winds, heavy rain, or external mechanical vibration is captured in real time by the anemometer and tank vibration sensor at the top of the flotation tank. When external physical oscillations cause liquid surface fluctuations or the interference signal-to-noise ratio exceeds the safety threshold, the scum interface will become extremely unstable. At this time, the MPC optimizer automatically triggers adaptive defense logic: actively reducing the error tracking penalty weight of the "density index" in the cost function to avoid nonlinear oscillations caused by over-adjustment of the system; at the same time, forcibly shrinking the constraint boundary of the effluent suspended solids (SS) concentration (for example, adaptively reducing the maximum allowable SS concentration by 50%).

[0043] The model predictive control optimizer includes a multi-objective cost function with adaptive weights for environmental disturbances, and the dynamic update formula for this function is as follows: (19) in, It is a multi-objective cost function; To predict the length of the time domain; To control the length of the time domain; This is the adaptive attenuation coefficient for environmental disturbances under normal operating conditions. When environmental interference exceeds the standard To weaken the target tracking intensity; For state error weights; For the model in The predicted value of the temporal density index at any given time. The desired density index; To control the incremental suppression weighting coefficient; The above-mentioned SHAP marginal contribution weight diagonal matrix is ​​dynamically generated by the machine learning model; For the dynamic increment of multidimensional control variables; As an operating cost penalty factor; This is a function representing the combined operating cost of drug consumption and energy consumption. A very high penalty coefficient for violations of soft constraints; The predicted concentration of suspended solids in the effluent; The water quality constraint safety boundary is dynamically adjusted and shrinks in response to environmental disturbance intensity.

[0044] Through the above steps, the physical causality at the microscopic quantum state level replaces the purely statistical black-box prediction in traditional control logic, accurately revealing the underlying mechanism of multiphase flow separation. Combined with an MPC cost function incorporating an adaptive update mechanism for environmental factors and SHAP marginally guided weights, the control strategy of this invention not only achieves global optimization of separation accuracy and operating energy consumption under stable conditions, but also endows the control system with unbreakable optimization robustness and a bottom-line guarantee of effluent water quality when facing severe weather shocks and drastic water quality fluctuations.

[0045] S4 outputs fluid control parameters and slag discharge execution parameters, which are used for adaptive adjustment of the bottom fluid and slag discharge operation of the phased array acoustic transmitter, respectively.

[0046] In this embodiment, the fluid control parameters include the reagent dosage concentration and the multiphase flow dissolved gas pump speed; the output fluid control parameters, used for adaptive adjustment of the underlying fluid, include: The output agent dosage concentration is fed to the flocculant metering pump to adjust the stroke frequency; the output multiphase flow dissolved air pump speed is used to adaptively adjust the dissolved air-water reflux ratio. In a preferred embodiment, when the bottom fluid execution device receives optimized control commands, the flocculant metering pump adjusts the instantaneous dosage concentration of the polymeric agent by changing the mechanical stroke frequency. After the agent molecules enter the raw water, they neutralize the Zeta potential on the surface of suspended colloids, compressing and disrupting the colloidal double layer of pollutants. Simultaneously, the multiphase flow dissolved air pump adjusts its speed via a frequency converter, changing the system pressure and reflux ratio of the dissolved air-water. This response determines the particle size distribution and volume fraction of micro-nano bubbles at the release end. The two work synergistically at the fluid dynamics level, enabling a specific concentration of microbubble clusters and the destabilized floc skeleton to reach a certain "eddy current capture frequency" in the separation zone eddy current, achieving "bubble embedding" between bubbles and pollutants, providing conditions for the subsequent three-dimensional dense structure of the scum layer.

[0047] Furthermore, the output slag removal execution parameters, used for the slag removal operation of the phased array acoustic transmitter, include: Based on the slag discharge execution parameters, the target slag discharge thrust and sound field focusing path are determined; control signals are generated to dynamically adjust the emission phase and amplitude of each transducer element in the phased array acoustic emission device, so as to excite directional acoustic traveling waves at the interface.

[0048] In this embodiment, the slag discharge execution parameters output by the joint prediction optimization model are first analyzed to extract the desired slag discharge rate command. The system combines the calibrated density index and viscoelastic parameters, substitutes them into the fluid dynamics resistance model, and calculates the target slag discharge thrust required to overcome the cross-linking viscosity within the slag layer and the surface tension at the gas-liquid interface. Simultaneously, based on the reconstructed three-dimensional morphology of the slag layer and the fixed relative spatial coordinates of the flotation tank's slag discharge trough, a horizontally sweeping acoustic field focusing path with a progressive spatial movement gradient is planned.

[0049] The formula for calculating the target slag discharge thrust and the spatial coordinate system equation of the sound field focusing path are as follows: (20) (twenty one) in, The target macroscopic slag discharge thrust that the system needs to output; The contact area between the scum layer and the bottom clear water; The equivalent dynamic viscosity coefficient after calibration using acoustic tweezers physical anchor points (its value is similar to the aforementioned Young's modulus). (positive correlation) The desired slag discharge propulsion rate; This represents the average thickness of the scum layer. The length of the three-phase perimeter where the scum layer contacts the pool wall. This refers to the infinitesimal element of the line segment along the surrounding path; The effective surface tension coefficient of the gas-liquid interface; For dynamic contact angle; The estimated total mass of the scum layer; for The absolute physical coordinate vector of the focal point of the sound field in three-dimensional space at any given moment; , , These are the transient coordinate components of the focal point in the spatial rectangular coordinate system along the horizontal slag discharge direction, the horizontal transverse direction, and the vertical depth. This is the time integration variable used to calculate the sweeping advance distance; The longitudinal coordinate of the slag discharge starting point; The coordinates of the lateral center of the flotation tank; This refers to the effective lateral width for sludge discharge in the dissolved air flotation tank. The angular frequency of the sound wave focus as it sweeps in a zigzag pattern laterally; The depth coordinates are the gas-slag-water interface.

[0050] With the above-mentioned target driving force clearly defined With focus path Subsequently, in the air zone above the liquid surface in the flotation tank separation zone (preferably at a distance from the liquid surface) to The location is equipped with an inclined two-dimensional ultrasonic transducer array. This array uses... The piezoelectric ceramic matrix arrangement is set to a working frequency band of 20kHz to 40kHz, and the tilt angle is set to be perpendicular to the horizontal liquid surface. to In order to accurately track the planned sound field focusing path. And output the target slag discharge thrust. By using an FPGA digital delay generator, the phase delay and amplitude of ultrasonic waves emitted by each transducer element in the array are independently controlled. This causes the originally divergent high-frequency sound waves to undergo spatial coherent superposition at the air-scum-water interface along a set path, synthesizing a directional acoustic traveling wave with macroscopic momentum. Due to the huge acoustic impedance difference between the high-density scum layer (a solid-gas-liquid multiphase coupler) before scum discharge and the water at the bottom, when the directional acoustic traveling wave passes through this interface, the strong sound wave reflection and momentum gradient transfer are directly converted into horizontally forward acoustic radiation pressure. Using this radiation pressure, the scum layer is smoothly, continuously, and non-contactly pushed into the scum discharge tank.

[0051] In addition, the process of outputting slag discharge execution parameters also includes: The system continuously monitors whether the calculated density index is greater than or equal to a preset critical threshold for slag discharge. If the density index is less than the critical threshold, a dormancy command is generated and sent to the phased array acoustic transmitter. If the density index is greater than or equal to the critical threshold, a slag discharge command is generated and sent to the phased array acoustic transmitter to generate directional acoustic traveling waves. As a preferred implementation, this threshold judgment logic provides the slag layer with sufficient time for natural gravity dehydration. Only when the free water bladders inside the slag layer are fully squeezed out, the bubble network shrinks, and the physical density reaches a mature state, will the acoustic phased array be activated and trigger slag discharge. This effectively prevents the premature discharge of "immature slag" containing a large amount of free water and loose flocs, greatly improving the dryness of the mud cake after solid-liquid separation.

[0052] Furthermore, after determining that the density meets the standard and formally triggering the slag discharge command, in order to accurately synthesize the required directional acoustic traveling wave at the complex interface, the calculation formula for the transmission phase delay of each transducer element of the acoustic phased array is as follows: (twenty two) Furthermore, under this phase delay control, the final acoustic radiation force excited and acting on the surface of the scum is calculated as follows: (twenty three) in, For the first in the array The transmit phase delay angle of each transducer element; This refers to the ultrasonic wave emission frequency; The speed at which sound waves propagate in the air medium; For the first The physical coordinates of each transducer element on the two-dimensional array plane; To synthesize the spatial focusing elevation and azimuth angles of the directional traveling wave; The macroscopic acoustic radiation force acting on the entire slag layer along the slag discharge direction; The incident sound intensity projected onto the interface by the transducer array; The equivalent acoustic impedance of a dense scum layer; The acoustic impedance of the bottom clear water; the square term in the formula The acoustic reflectivity and momentum gradient transfer intensity caused by phase differences were characterized. The angle of incidence of the sound wave at the interface; The effective integral area of ​​the scum surface covered by the sound wave.

[0053] Through the above steps, this invention utilizes an acoustic phased array to achieve non-contact "virtual sludge scraping." This closed-loop control eliminates the problems of floc breakage, trace pollutant shedding, and secondary back-mixing leakage caused by physical shearing forces during contact scraping in traditional mechanical sludge scrapers. While ensuring absolute physical isolation between sludge and clean water, it improves the purity of the discharged sludge and the level of adaptive operation.

[0054] Example 2, Figure 5 A micro / nano air flotation solid-liquid separation control system based on scum density monitoring is presented, including: The feature analysis module is used to acquire terahertz penetration scan echo data of the scum layer and analyze the profile features of free water and bubble distribution. The density calculation module is used to calculate the density index, which characterizes the compactness of the three-dimensional structure inside the scum layer, based on the profile features. The density calibration module is used to obtain physical viscoelastic parameters based on ultrasonic standing wave fields and to calibrate the density index using the viscoelastic parameters. The parameter optimization module is used to input the calibrated density index and the feedwater feedforward signal into the joint prediction optimization model to solve for the optimal combination of operating variables that satisfies the multi-objective constraints; the combination of variables includes at least fluid control parameters and slag discharge execution parameters. The solid-liquid separation module is used to output fluid control parameters and slag discharge execution parameters, which are used for adaptive adjustment of the bottom fluid and slag discharge operation of the phased array acoustic transmitter, respectively.

[0055] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0056] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0057] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0058] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0059] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0060] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A micro / nano air flotation solid-liquid separation control method based on scum density monitoring, characterized in that, include: Acquire terahertz penetration scan echo data of the scum layer and analyze the profile characteristics of free water and bubble distribution; Based on the aforementioned profile features, the density index, which characterizes the compactness of the three-dimensional structure inside the scum layer, is calculated. Obtain the physical viscoelastic parameters based on the ultrasonic standing wave field, and use the viscoelastic parameters to calibrate the density index; The calibrated density index and the feedforward signal are input into a joint prediction and optimization model to solve for the optimal combination of operating variables that satisfies the multi-objective constraints; the combination of variables includes at least fluid control parameters and slag discharge execution parameters. Output fluid control parameters and slag discharge execution parameters for adaptive adjustment of the bottom fluid and slag discharge operation of the phased array acoustic transmitter, respectively.

2. The method according to claim 1, characterized in that, The acquisition of terahertz penetration scan echo data of the scum layer and the analysis of the free water and bubble distribution profile characteristics include: Acquire terahertz echo signals reflected and transmitted by the scum layer; Based on the time-domain delay and amplitude attenuation of the echo signal, the difference in dielectric constant of the scum layer at different depth profiles is analyzed. Based on the difference in dielectric constant, the three-dimensional distribution profile of free water and bubbles inside the scum layer is reconstructed.

3. The method according to claim 1, characterized in that, The calculation of the density index, which characterizes the compactness of the three-dimensional structure within the scum layer, and the calibration using the viscoelastic parameters, includes: The profile features are processed using a pre-trained deep fusion network to output an initial density index; The viscoelastic parameters are used as physical anchors to calibrate the weights of the deep fusion network, and the calibrated density index is output.

4. The method according to claim 1, characterized in that, Before inputting the calibrated density index and the influent feedforward signal into the joint prediction and optimization model, the following steps are also included: Obtain the low-field nuclear magnetic resonance scanning spectrum of the mixed fluid in the separation zone of the flotation tank; Based on the displacement characteristics of the relaxation peaks in the scanned spectrum, the microscopic physical bonding state between micro- and nano-bubbles and pollutant flocs is quantified. The combined force state is used as a precondition input into the joint prediction optimization model.

5. The method according to claim 4, characterized in that, The joint prediction optimization model includes a machine learning prediction model based on feature importance attribution and a model prediction control optimizer; The solution yields the optimal combination of operational variables that satisfies the multi-objective constraints, including: Using a machine learning prediction model, combined with the binding force state, the marginal contribution weight of multidimensional adjustment variables to maintaining the density index is predicted in real time. The model predictive control optimizer uses the marginal contribution weights as a search guide to perform multi-objective rolling optimization under preset operating cost and effluent water quality constraints, and solves for the optimal combination of operating variables.

6. The method according to claim 1, characterized in that, Fluid control parameters include reagent concentration and multiphase flow dissolved air pump speed; The output fluid control parameters, used for adaptive adjustment of the underlying fluid, include: The dosage concentration of the agent is output to the flocculant metering pump to adjust the stroke frequency; Output the speed of the multiphase flow dissolved air pump to adaptively adjust the dissolved air-water reflux ratio.

7. The method according to claim 1, characterized in that, Output slag removal execution parameters for use in the slag removal operation of the phased array acoustic transmitter, including: Based on the slag discharge execution parameters, determine the target slag discharge thrust and the sound field focusing path; Control signals are generated to dynamically adjust the emission phase and amplitude of each transducer element in the phased array acoustic transmitter, so as to excite directional acoustic traveling waves at the interface.

8. The method according to claim 7, characterized in that, The process of outputting slag discharge execution parameters also includes: Real-time determination of whether the calculated density index is greater than or equal to the preset critical threshold for slag discharge; If the density index is less than the critical threshold for slag discharge, a dormancy command is generated and sent to the phased array acoustic transmitter. If the density index is greater than or equal to the critical threshold for slag discharge, a slag discharge command is generated and sent to the phased array acoustic transmitter to generate directional acoustic traveling waves.

9. The method according to claim 5, characterized in that, The process of multi-objective rolling optimization under preset operating cost and effluent quality constraints also includes: Acquire environmental noise levels and effluent suspended solids concentration; When the level of environmental noise exceeds the preset safety threshold, the tracking weight of the density index is reduced and the constraint boundary of the effluent suspended solids concentration is narrowed. The cost function of the model predictive control optimizer is dynamically updated to ensure that the effluent quality meets the preset standards.

10. A system using the micro / nano air flotation solid-liquid separation control method based on scum density monitoring as described in any one of claims 1 to 9, characterized in that, include: The feature analysis module is used to acquire terahertz penetration scan echo data of the scum layer and analyze the profile features of free water and bubble distribution. The density calculation module is used to calculate the density index, which characterizes the compactness of the three-dimensional structure inside the scum layer, based on the profile features. The density calibration module is used to obtain physical viscoelastic parameters based on ultrasonic standing wave fields and to calibrate the density index using the viscoelastic parameters. The parameter optimization module is used to input the calibrated density index and the feedwater feedforward signal into the joint prediction optimization model to solve for the optimal combination of operating variables that satisfies the multi-objective constraints; the combination of variables includes at least fluid control parameters and slag discharge execution parameters. The solid-liquid separation module is used to output fluid control parameters and slag discharge execution parameters, which are used for adaptive adjustment of the bottom fluid and slag discharge operation of the phased array acoustic transmitter, respectively.

Images