Method and system for detecting settling performance of activated sludge based on atlas analysis

Abstract

The invention belongs to the technical field of environmental engineering sewage treatment and intelligent detection, and discloses a method and a system for detecting the settling performance of activated sludge based on atlas analysis, the method comprises the following steps: regularly acquiring sludge settling process images through an image acquisition device, and constructing a colored three-dimensional settling atlas containing dynamic settling information after preprocessing; and based on a convolutional neural network AlexNet architecture, SV30 and SVI30 intelligent detection models taking the three-dimensional sedimentation map as input are respectively established, so that real-time, accurate and automatic prediction of the sludge sedimentation performance is realized. The method overcomes the defects that a traditional manual measurement method is high in hysteresis quality, prone to interference, dependent on manual calibration and the like, has the advantages of being high in anti-interference capacity, low in cost, convenient to integrate and apply and the like, and remarkably improves the intelligent detection level of the sewage treatment process.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of wastewater treatment and intelligent detection technology in environmental engineering, and specifically relates to a method and system for detecting the settling performance of activated sludge based on spectral analysis. Background Technology

[0002] The activated sludge process, as the dominant technology for modern municipal wastewater and industrial wastewater treatment, fundamentally relies on the aggregation state and settling properties of the activated sludge microbial community within the reactor to ensure treatment efficiency and stable effluent quality. Good sludge-water separation is crucial for guaranteeing clear effluent and maintaining stable biomass within the system. In this process, the sludge settling ratio (SV) is... 30 Sludge Volume Index (SVI) 30 As two core empirical parameters, they are widely used in daily operation monitoring and process diagnostics. 30 It directly reflects the settling velocity and compressibility of sludge, while SVI 30 This further eliminates the influence of sludge concentration, and can more scientifically characterize the settling characteristics of sludge, making it an important early warning indicator for predicting abnormal phenomena such as sludge bulking.

[0003] For a long time, SV 30 With SVI 30 The sludge volume percentage (SV) is primarily obtained through classic manual measurement methods. This method requires operators to periodically collect mixed liquor samples from the inlet of the aeration tank or secondary sedimentation tank, inject them into a 1-liter graduated cylinder, and precisely observe the sludge-water interface settling process for 30 minutes under static conditions. The sludge volume percentage after settling is recorded to obtain the SV. 30 Furthermore, a separate sample needs to be taken to determine the suspended solids concentration (MLSS) of the mixture, and the SVI can be calculated accordingly. 30 While this method is direct, it has significant drawbacks: the entire process takes over an hour, suffers from severe feedback lag, and cannot provide real-time data support for the control of key operating parameters such as aeration rate and reflux ratio. Furthermore, the process is highly dependent on the experience and responsibility of the operators; subjective errors in interface interpretation, deviations in settling time control, and fluctuations in laboratory MLSS measurements all cumulatively affect the accuracy of the final results, while also imposing a heavy human burden.

[0004] To overcome the lag issues of manual measurements and achieve continuous process monitoring, online detection instruments based on optical sensing or electrochemical principles have gradually emerged in the market. These devices attempt to indirectly determine sludge concentration through immersion probes or flow measurement tanks. However, these technological approaches face significant challenges in practical applications. Optical sensors are susceptible to interference from water color, turbidity, suspended impurities, oil films, and bubbles, leading to distortion of scattered or transmitted light signals. Electrochemical sensors are sensitive to changes in water ionic strength and pH, and suffer from prominent issues of probe membrane fouling and electrode passivation. This necessitates frequent manual cleaning and professional calibration of these instruments, resulting in high maintenance costs and insufficient stability. Measurement results often become less reliable under complex actual water quality conditions, limiting their widespread application in wastewater treatment scenarios that demand low maintenance and high reliability.

[0005] As the water treatment industry moves towards intelligent and refined operation and maintenance, leveraging machine vision, big data, and artificial intelligence technologies to empower traditional process monitoring has become a research hotspot and an inevitable trend. Although existing research has attempted to monitor sludge characteristics through image analysis technology, most studies remain limited to simple analysis of static appearance morphology (such as floc size and color), or merely attempt to replace the single step of manually reading 30-minute settling volumes. These studies fail to fundamentally achieve a comprehensive and efficient digital analysis of information related to the dynamic process of settling. How to transform the complete settling dynamics into features that can be deeply mined by machine learning models, thereby establishing a new detection method with advantages of real-time performance, accuracy, robustness, and low cost, remains a crucial technology that urgently needs breakthroughs in this field.

[0006] Therefore, existing technologies have significant gaps in meeting the needs of modern wastewater treatment plants for real-time, accurate, and low-maintenance intelligent monitoring of settling performance. Developing an innovative method that can comprehensively capture settling dynamics, has strong anti-interference capabilities, and requires no complex calibration is crucial for improving the intelligent control level and operational stability of wastewater treatment processes. Summary of the Invention

[0007] The purpose of this invention is to address the problems in existing sludge settling performance testing methods, which mainly rely on manual sampling, volume readings, or online monitoring equipment based on optical sensors. These methods suffer from operational lag, susceptibility to water quality interference, and the need for frequent calibration. This invention provides a method and system for testing activated sludge settling performance based on spectral analysis, achieving a sludge settling ratio (SV). 30 Sludge Volume Index (SVI) 30 Fast, accurate, and automated predictions.

[0008] To achieve the above objectives, the present invention employs the following technical solution: The present invention proposes a method for detecting the settling performance of activated sludge based on spectral analysis, comprising the following steps: Obtain sludge settling ratio, sludge volume index, and color three-dimensional settling maps containing sludge settling dynamic information; process the color three-dimensional settling maps to obtain sub-maps. Based on sub-maps, sludge settling ratio, and the AlexNet model, an SV was constructed. 30 Regression prediction model; SVI constructed based on sub-map, sludge volume index and AlexNet model. 30 Regression prediction model; Input the three-dimensional settlement map to be analyzed into SV 30 / SVI 30 The regression prediction model outputs predicted values ​​of sludge settling ratio or sludge volume index, enabling the detection of activated sludge settling performance.

[0009] Preferably, the color three-dimensional settling map containing sludge settling dynamic information specifically comprises: Starting from the initial settling of the sludge, images of the settling process were acquired every 10 seconds for 15 minutes. The sludge images were preprocessed, including background subtraction, image alignment and cropping, to retain the effective settling area. The preprocessed images were arranged in chronological order to obtain a continuous map reflecting the sludge interface settling process. The continuous map was then processed with pseudo-color to obtain a color three-dimensional settling map containing dynamic information about sludge settling.

[0010] Preferably, the process of processing the color three-dimensional settlement map to obtain sub-maps specifically includes: The color three-dimensional settlement map was divided into three 5-minute sub-maps along the time axis: the first, middle, and last.

[0011] Preferably, the SV model is constructed based on sub-maps, sludge settling ratio, and the AlexNet model. 30 The regression prediction model is as follows: Three-dimensional sedimentation maps of sludge under different operating conditions were collected during actual operation, and the sludge settling ratio (SV) corresponding to the three-dimensional sedimentation maps was measured simultaneously. 30 As labeling data, sub-maps were obtained based on the 3D sludge settling map; using the AlexNet convolutional neural network architecture, the sub-maps were used as input, and the sludge settling ratio (SV) was calculated. 30 As output, construct SV 30 The regression prediction model incorporates weighted transfer learning to accelerate convergence during training, employs regularization to suppress overfitting, and optimizes training performance by dynamically adjusting the learning rate.

[0012] Preferably, the SVI is constructed based on the sub-map, sludge volume index, and AlexNet model. 30 The regression prediction model is as follows: Three-dimensional sedimentation maps of sludge under different operating conditions were collected during actual operation, and the sludge volume index (SVI) corresponding to the three-dimensional sedimentation maps was measured simultaneously. 30 As labeling data, sub-maps were obtained based on the 3D sedimentation map of sludge; using the AlexNet convolutional neural network architecture, the sludge volume index SV was calculated as input. 30 As output, an SVI regression prediction model is constructed; during training, weight transfer learning is introduced to accelerate convergence, regularization is used to suppress overfitting, and the training effect is optimized by dynamically adjusting the learning rate.

[0013] Preferably, the AlexNet convolutional neural network architecture is selected with a depth of 34 layers; the training process adopts an adaptive learning rate adjustment strategy based on validation loss, specifically using the ReduceLROnPlateau learning rate scheduler. During the training process, the validation loss is used as the monitoring metric. When the validation loss no longer decreases within 10 consecutive epochs, the learning rate is automatically multiplied by 0.5 to reduce it, achieving the effect of rapid convergence in the early stage of training and fine-tuning in the later stage.

[0014] This invention proposes an activated sludge settling performance testing system based on spectral analysis, comprising: The sub-map acquisition module is used to acquire sludge settling ratio, sludge volume index and color three-dimensional settling map containing sludge settling dynamic information, and to process the color three-dimensional settling map to obtain sub-maps. The model building module is used to construct SV based on the sub-map, sludge settling ratio, and AlexNet model. 30 Regression prediction model; SVI constructed based on sub-map, sludge volume index and AlexNet model. 30 Regression prediction model; The sludge settling performance testing module is used to input the three-dimensional settling map to be analyzed into the SV / SVI regression prediction model and output the predicted value of sludge settling ratio or sludge volume index, thereby realizing the settling performance testing of activated sludge.

[0015] Preferably, the color three-dimensional settling map containing sludge settling dynamic information specifically comprises: Starting from the initial settling of the sludge, images of the settling process were acquired every 10 seconds for 15 minutes. The sludge images were preprocessed, including background subtraction, image alignment and cropping, to retain the effective settling area. The preprocessed images were arranged in chronological order to obtain a continuous map reflecting the sludge interface settling process. The continuous map was then processed with pseudo-color to obtain a color three-dimensional settling map containing dynamic information about sludge settling.

[0016] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of a method for detecting the sedimentation performance of activated sludge based on atlas parsing are implemented.

[0017] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of a method for detecting the sedimentation performance of activated sludge based on atlas parsing are implemented.

[0018] Compared with the prior art, the present invention has the following beneficial effects: A method for detecting the sedimentation performance of activated sludge based on atlas parsing proposed by the present invention first obtains the sludge volume ratio, the sludge volume index, and a color three-dimensional sedimentation atlas containing dynamic information of sludge sedimentation, and processes them to obtain sub-atlases. Compared with the offline operation mode of manual sampling and volume reading, the color three-dimensional sedimentation atlas can capture the dynamic process of sludge sedimentation in real time, without manual intervention in sampling and reading, avoiding the lag caused by manual operation from the detection source; compared with the optical sensor monitoring that is easily interfered by water quality turbidity, impurities, etc., the three-dimensional sedimentation atlas contains richer dynamic characteristics information of sedimentation, and the extraction of sub-atlases further focuses on key features, reducing the interference of complex components of water quality on detection. Subsequently, based on the sub-atlases and the corresponding SV 30

[0017] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of a method for detecting the sedimentation performance of activated sludge based on atlas parsing are implemented.

[0018] Compared with the prior art, the present invention has the following beneficial effects: A method for detecting the sedimentation performance of activated sludge based on atlas parsing proposed by the present invention first obtains the sludge volume ratio, the sludge volume index, and a color three-dimensional sedimentation atlas containing dynamic information of sludge sedimentation, and processes them to obtain sub-atlases. Compared with the offline operation mode of manual sampling and volume reading, the color three-dimensional sedimentation atlas can capture the dynamic process of sludge sedimentation in real time, without manual intervention in sampling and reading, avoiding the lag caused by manual operation from the detection source; compared with the optical sensor monitoring that is easily interfered by water quality turbidity, impurities, etc., the three-dimensional sedimentation atlas contains richer dynamic characteristics information of sedimentation, and the extraction of sub-atlases further focuses on key features, reducing the interference of complex components of water quality on detection. Subsequently, based on the sub-atlases and the corresponding SV 30 、SVI 30 data, combined with the AlexNet model, SV 30 、SVI 30 regression prediction models are respectively constructed. By using the feature learning and fitting capabilities of the models, accurate mapping of sedimentation performance indicators is achieved, without the need for frequent calibration operations like optical sensors due to device drift, etc. Finally, the three-dimensional sedimentation atlas to be analyzed is input into the constructed regression prediction model, and the predicted value can be output. The entire process does not require manual intervention in the core detection link, realizing fast, accurate, and automated prediction of SV 30 and SVI 30 , comprehensively solving the problems of the prior art.

[0019] A system for detecting the sedimentation performance of activated sludge based on atlas parsing proposed by the present invention divides the system into a sub-atlas acquisition module, a model construction module, and a sludge sedimentation performance detection module, obtains the predicted value of the sludge volume ratio or the sludge volume index, and realizes the detection of the sedimentation performance of activated sludge. Using the modular idea makes each module independent of each other, facilitating unified management of each module. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the activated sludge settling performance testing method based on spectral analysis according to the present invention.

[0022] Figure 2 This is a detailed flowchart of the activated sludge settling performance testing method based on spectral analysis according to the present invention.

[0023] Figure 3 This is a schematic diagram illustrating the construction of the activated sludge settling characteristic spectrum of the present invention.

[0024] Figure 4 The SV of the present invention 30 and SVI 30 Mean squared error (MSE) and loss changes during the training process of intelligent detection models.

[0025] Figure 5 The SV of the present invention 30 and SVI 30 The test set is used to evaluate the model's performance under the model's predictions.

[0026] Figure 6 This is a diagram of the activated sludge settling performance testing system based on graph analysis according to the present invention.

[0027] Figure 7 This is a schematic diagram of the structure of an electronic device according to the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0029] The present invention will now be described in further detail with reference to the accompanying drawings: This invention discloses a method for testing the settling performance of activated sludge based on spectral analysis, such as... Figure 1 As shown, the steps are as follows: S1. Obtain the sludge settling ratio, sludge volume index, and color three-dimensional settling map containing sludge settling dynamic information, and process the color three-dimensional settling map to obtain sub-maps. The color three-dimensional sedimentation map containing dynamic information on sludge sedimentation is specifically as follows: Starting from the initial settling of the sludge, images of the settling process were acquired every 10 seconds for 15 minutes. The sludge images were preprocessed, including background subtraction, image alignment and cropping, to retain the effective settling area. The preprocessed images were arranged in chronological order to obtain a continuous map reflecting the sludge interface settling process. The continuous map was then processed with pseudo-color to obtain a color three-dimensional settling map containing dynamic information about sludge settling.

[0030] The process of processing the color three-dimensional settlement map to obtain sub-maps specifically involves: The color three-dimensional settlement map was divided into three 5-minute sub-maps along the time axis: the first, middle, and last.

[0031] S2. Based on sub-maps, sludge settling ratio, and the AlexNet model, construct SV. 30 Regression prediction model; SVI constructed based on sub-map, sludge volume index and AlexNet model. 30 Regression prediction model; The SV model is constructed based on sub-maps, sludge settling ratio, and the AlexNet model. 30 The regression prediction model is as follows: Three-dimensional sedimentation maps of sludge under different operating conditions were collected during actual operation, and the sludge settling ratio (SV) corresponding to the three-dimensional sedimentation maps was measured simultaneously. 30 As labeling data, sub-maps were obtained based on the 3D sludge settling map; using the AlexNet convolutional neural network architecture, the sub-maps were used as input, and the sludge settling ratio (SV) was calculated. 30 As output, an SV regression prediction model is constructed; during training, weight transfer learning is introduced to accelerate convergence, regularization is used to suppress overfitting, and the training effect is optimized by dynamically adjusting the learning rate.

[0032] The SVI is constructed based on sub-maps, sludge volume index, and the AlexNet model. 30 The regression prediction model is as follows: Three-dimensional sedimentation maps of sludge under different operating conditions were collected during actual operation, and the sludge volume index (SVI) corresponding to the three-dimensional sedimentation maps was measured simultaneously. 30 As labeling data, sub-maps were obtained based on the 3D sedimentation map of sludge; using the AlexNet convolutional neural network architecture, the sludge volume index SV was calculated as input. 30 As output, construct SVI 30The regression prediction model incorporates weighted transfer learning to accelerate convergence during training, employs regularization to suppress overfitting, and optimizes training performance by dynamically adjusting the learning rate.

[0033] The AlexNet convolutional neural network architecture uses a depth of 34 layers. The training process employs an adaptive learning rate adjustment strategy based on validation loss, specifically using the ReduceLROnPlateau learning rate scheduler. During training, validation loss is used as the monitoring metric. When the validation loss no longer decreases within 10 consecutive epochs, the learning rate is automatically multiplied by 0.5 to reduce it, achieving rapid convergence in the early stages of training and fine-tuning in the later stages.

[0034] S3. Input the three-dimensional settlement map to be analyzed into SV 30 / SVI 30 The regression prediction model outputs predicted values ​​of sludge settling ratio or sludge volume index, enabling the detection of activated sludge settling performance.

[0035] This invention discloses a method for testing the settling performance of activated sludge based on spectral analysis, the flowchart of which is shown below. Figure 2 As shown, the process of constructing the activated sludge settling characteristic map is as follows: Figure 3 As shown, its SV 30 and SVI 30 The training process and test results of the model are as follows: Figure 4 and Figure 5 As shown, the specific steps are as follows: Specifically, Figure 4 (a) in the text is SV 30 The change in the loss function value during the prediction model training process, (b) is the SV. 30 The changes in MSE value during the prediction model training process, (c)-(d) represent SVI. 30 The changes in the loss function value and MSE value during the training process of the prediction model. Figure 5 (a) in the text is SV 30 Performance results of the prediction model under optimal parameters (R) 2 (b) is SVI 30 Performance results (R2) of the prediction model under optimal parameters.

[0036] Step 1: Method for constructing three-dimensional sedimentation maps of sludge An image acquisition device was installed in the activated sludge reactor. Images of the settling process were acquired every 10 seconds from the start of sludge settling for 15 minutes, resulting in 90 images. The images underwent preprocessing, including background removal, image alignment, and cropping, preserving the effective settling area. The processed images were then arranged chronologically to create a continuous atlas reflecting the sludge interface settling process. To further enhance visual appeal, the composite images were processed with pseudo-color: water color represented the supernatant, sludge color represented the settled sludge, and transition areas were color-gradientd according to the sludge-water ratio, forming a color three-dimensional settling atlas containing dynamic information about sludge settling. Background removal was performed to remove only the columnar biological reaction body of the SBR reactor.

[0037] The image acquisition device is typically positioned at a fixed location 1 meter outside the reactor. Red represents the color of the supernatant water, and blue represents the color of the settled sludge. The transition zone uses a gradient of red and blue colors based on the sludge-to-water ratio, forming a three-dimensional red-blue sedimentation map containing dynamic information about sludge settling. The flowchart illustrating the constructed three-dimensional sludge sedimentation map is provided below. Figure 3 .

[0038] To verify the generalization performance of this invention, this study selected five laboratory-scale sequencing batch reactors (SBRs), numbered R1 to R5, and set different aeration rates to simulate the operation of activated sludge in a real wastewater treatment plant. The activated sludge samples used were collected from the biological reactor of a wastewater treatment plant in Xi'an.

[0039] Preferably, the aeration rates of R1 to R5 are set to 0.5 to 2 L / s, respectively, and are distributed at equal intervals.

[0040] Specifically, the five reactors used in the experiment were all cylindrical devices of uniform specifications, each with an effective volume of 3 L and specific structural dimensions of 1000 mm in height and 60 mm in inner diameter. Each reactor had an inlet on the upper side wall, an aeration head at the bottom, and a drain on the middle side wall.

[0041] Specifically, this invention sets each operating cycle to 6 hours, and divides it into five continuous stages: static influent, anaerobic digestion, aeration, sedimentation, and drainage. The specific duration allocation for each stage is shown in Table 1.

[0042] Specifically, during the influent stage, simulated wastewater is injected into the reactor at a uniform speed through the bottom inlet, controlled by a peristaltic pump, while constant-speed stirring ensures uniform distribution of the mixture. The aeration process is driven by an air compressor, generating bubbles through aeration sand heads, with the air volume precisely regulated by a glass rotor flow meter. During the drainage stage, the supernatant is discharged from the outlet located in the middle of the reactor (500mm from the bottom) at a 50% replacement rate via a controlled drain valve and collected in a waste liquid tank.

[0043] Specifically, the operation of the aforementioned equipment (peristaltic pump, drain valve, air compressor, and agitator) is all programmed and controlled by a time controller to ensure the accuracy and stability of the entire operation.

[0044] Table 1 Reactor operating parameters / min

[0045] Preferably, this experiment uses artificially prepared synthetic wastewater as the reaction substrate. Its main nutrient sources include sodium acetate as a carbon source, ammonium sulfate as a nitrogen source, and potassium dihydrogen phosphate as a phosphorus source, to ensure the necessary nutrients for microbial growth. The concentrations of C, N, and P in the wastewater are strictly controlled within the set ranges; the specific ratios are shown in Table 2.

[0046] Table 2. Influent Input Parameter Range for the Reactor

[0047] Specifically, the experiment measured MLSS and SV. 30 With SVI 30 The range of indicators is shown in Table 3.

[0048] Table 3 Measurement range of MLSS and SV indices

[0049] Step 2: SV based on 3D settlement map 30 Construction of regression prediction model Three-dimensional settling maps of sludge under different operating conditions were collected during actual operation, and the corresponding sludge settling ratios (SV) were measured simultaneously. 30 ) and sludge volume index (SVI) 30 ) as label data. The AlexNet convolutional neural network architecture is used, with 3D settlement maps as input, SV 30 As output, a regression prediction model is constructed. To adapt to the network input size and preserve time-series information, the original 3D settlement map is divided into three 5-minute sub-maps along the time axis, and each sub-map is input into the network to extract features. During training, weighted transfer learning is introduced to accelerate convergence, regularization is used to suppress overfitting, and the training effect is optimized by dynamically adjusting the learning rate.

[0050] The method for determining the sludge settling ratio (SV) is as follows: First, during the aeration stage of the reactor, take 100 ml of the sludge-water mixture from the reactor. Then, pour the mixture into a 100 ml graduated cylinder. Next, allow the mixture in the graduated cylinder to settle for 30 minutes, and then read the volume of the settled sludge. This volume value is the SV. 30 This method can be used to obtain SV.30 The values ​​are used to understand the settling performance of the sludge.

[0051] Sludge Volume Index (SVI) 30 The determination method for MLSS is as follows: First, determine the MLSS concentration. The specific steps are as follows: Preheat quantitative filter paper to a temperature of 103-105℃ in an oven for 2 hours until constant weight, then weigh it and record the weight as m0 (mg); Take 100 mL of the sludge-water mixture sample and filter it using a Buchner funnel. During the operation, be sure to wash the cup wall with pure water to ensure that the sludge is completely transferred to the filter paper; Place the filter paper and the activated sludge solids on the filter paper in an oven and bake for 2 hours; After drying, place the filter paper in a desiccator to cool, and then weigh it in an analytical balance and record it as m1 (mg); Calculate the MLSS (mg / L) according to the formula m0 - m1, and then calculate the MLSS according to the SV. 30 Using Formula 1, SVI is calculated. 30 . (1) The AlexNet convolutional neural network architecture uses a depth of 34 layers. L2 regularization is chosen as the regularization method. The training process employs an adaptive learning rate adjustment strategy based on validation loss, specifically using the ReduceLROnPlateau learning rate scheduler. During training, validation loss is used as the monitoring metric. When the validation loss no longer decreases within 10 consecutive epochs, the learning rate is automatically multiplied by 0.5 to reduce it, thus achieving rapid convergence in the early stages of training and fine-tuning in the later stages.

[0052] Step 2.1 Dataset Preparation and Division: Collect a large number of activated sludge samples from different wastewater treatment plants, different seasons, and different operating conditions. Simultaneously construct their three-dimensional sedimentation maps according to the method described in Step 1, and strictly determine the corresponding SV according to standard methods. 30 The values ​​are used to form the original dataset. Then, the dataset is randomly divided into a training set and a validation set in a 7:3 ratio. The training set is used for learning model parameters, the validation set is used for hyperparameter tuning and monitoring the training process to prevent overfitting, and the test set is used to finally evaluate the model's generalization performance.

[0053] It should be noted that, in order to test the generalization performance of the model, this example uses the new data collected from 5 laboratory-scale reactors in step 1 as a test set to verify the model's performance.

[0054] Step 2.2 Model Input Preprocessing (Sub-map Generation): To adapt to the standard input size of the AlexNet model and retain the time-series information of the settlement process to the maximum extent, the long strip-shaped continuous three-dimensional settlement map obtained in Step 1 is preprocessed as follows: The map representing a total duration of 15 minutes is evenly divided into three segments along the time axis (horizontal direction), corresponding to three sub-maps for 0-5 minutes (first segment), 5-10 minutes (middle segment), and 10-15 minutes (last segment) after the settlement begins. Each sub-map is independently scaled to the input size required by the model. This operation transforms a single settlement process into three temporally correlated feature samples, satisfying the network input requirements while enriching the information in the time dimension.

[0055] Specifically, the standard input size of the AlexNet model is typically 224×224 pixels.

[0056] Step 2.3 Model Architecture and Transfer Learning Initialization: In this example, AlexNet is preferred as the base network architecture. The model input is a sub-map of RGB three channels, and the output layer is set to a single neuron for direct regression prediction of SV. 30 To accelerate convergence and improve model performance on limited data, a transfer learning strategy is adopted: the weights of an AlexNet model pre-trained on a large image dataset (such as ImageNet) are used as the initial parameters of the network, and its last fully connected layer is replaced to adapt to the regression output of this task.

[0057] Step 2.4 Model Training and Hyperparameter Settings: Model training is performed on a GPU-equipped computing platform using the PyTorch deep learning framework. Key training hyperparameter settings are as follows: Specifically, the loss function is the mean squared error (MSE), which minimizes the predicted SV. 30 The difference between the value and the true value.

[0058] Specifically, the optimizer is Adam, with an initial learning rate set to 1e-4.

[0059] Specifically, the learning rate adjustment strategy is as follows: The ReduceLROnPlateau scheduler is used to monitor the validation set loss. If the loss does not decrease for 10 consecutive training epochs, the learning rate is multiplied by 0.5 to decay until it drops below 1e-8.

[0060] Specifically, regularization: To suppress overfitting, weight decay (i.e., L2 regularization) is set in the optimizer with a coefficient of 1e-4; at the same time, dropout is used during training.

[0061] Specifically, the batch size is set to 16 based on the GPU memory.

[0062] Specifically, the training epochs are 1000, employing an early stopping strategy. Training automatically terminates when the validation set loss no longer improves within 50 consecutive epochs, preserving the model parameters that best perform on the validation set. The training process is detailed below. Figure 4 .

[0063] Step 2.5 Model Evaluation: After training, the final model's performance is evaluated using the reserved test set. The main evaluation metric is the coefficient of determination (R²). This is achieved by comparing the model's predicted SV... 30 The values ​​were compared with actual measured values ​​to verify the accuracy and reliability of the model. A schematic diagram of the model's training loss versus validation loss changes can be found in [reference needed]. Figure 3 The scatter plot of the model's predicted and actual values ​​on the test set can be found in [link to relevant documentation]. Figure 5 .

[0064] It should be noted that, in order to test the generalization performance of the model, this example uses the new data collected from 5 laboratory-scale reactors in step 1 as a test set to verify the model's performance.

[0065] Specifically, the final trained SV 30 The predictive performance of the intelligent detection model for the five SBR reactors is R² = 0.893.

[0066] Step 3: SVI based on three-dimensional settlement map 30 Construction of regression prediction model The SV established in step two 30 Based on the prediction model, while maintaining the same neural network structure, graph preprocessing, and training strategies, the model output is changed to the sludge volume index (SVI). 30 Retraining yields SVI 30 Intelligent detection model. This model can directly extract features from three-dimensional settlement maps to achieve SVI (Smart Detection Model). 30 End-to-end prediction.

[0067] SVI 30 The optimal depth for the AlexNet convolutional neural network architecture in regression prediction models is 34 layers. (SVI) 30The specific strategy of the regularization method for the intelligent detection model is as follows: weight decay is adopted as the regularization strategy, which is implemented by setting the WEIGHT_DECAY parameter in the AdamW optimizer. This method can limit the size of the model parameters, effectively reduce the risk of overfitting, and, combined with the adaptive learning rate feature of the Adam optimizer, can improve the generalization ability while maintaining the model performance.

[0068] Step 3.1 Preparation of data labels: for training SVI 30 The dataset for the intelligent detection model has the same input portion (i.e., 3D sedimentation maps) as described in step 2.1. The key difference lies in the preparation of the label data. For each 3D sedimentation map, the corresponding label value must be the sludge volume index (SVI) measured using standard methods. 30 SVI 30 The calculation depends on two measured values: SV as described in step two. 30 The concentration of suspended solids (MLSS) in the liquid mixture is also determined. MLSS determination must strictly follow the standard gravimetric method: a certain volume of the liquid mixture sample is filtered through filter paper pre-dried to constant weight at 103-105℃, and the filter residue is then dried again to constant weight at the same temperature. The MLSS concentration (mg / L) is then calculated. Finally, the SVI for each sample is calculated according to Formula 1. 30 Value (mL / g).

[0069] Step 3.2 Model Architecture Adjustment and Initialization: SVI 30 The prediction model uses SV 30 The same AlexNet network is used as the feature extraction backbone. Due to SVI... 30 Even though it's still a continuous numerical value, this task is still defined as a regression problem. Therefore, it's only necessary to convert the SV... 30 The number of neurons in the model's output layer is adjusted to 1 to directly output SVI. 30 The predicted values. Model initialization is also recommended to use pre-trained weights on large datasets such as ImageNet, and fine-tuning should be performed using transfer learning strategies.

[0070] Step 3.3 Model Training Strategy: SVI 30 Model training strategy and SV 30 The models remain consistent to ensure comparability and consistency of methods.

[0071] Specifically, the data partitioning scheme is exactly the same as that used in step 2.1 for the training set, validation set, and test set partitioning.

[0072] Specifically, the hyperparameter settings are as follows: optimizer (Adam), initial learning rate (1e-4), learning rate scheduling strategy (ReduceLROnPlateau, patience=10, factor=0.5), regularization method (L2 weight decay, coefficient 1e-4), batch size, and early stopping strategy. These key hyperparameter settings are consistent with those described in step S2.4.

[0073] Specifically, the training process involves using a prepared (3D settlement map, SVI) 30 The network is retrained using the data pairs (values). During training, the loss changes on the validation set are monitored, and the best-performing model is saved. See the training process below. Figure 4 .

[0074] Step 3.4 Model Validation and Application: After training, evaluate SVI using the test set. 30 Performance of the intelligent detection model. Predicting SVI through calculation. 30 Value and measured SVI 30 The accuracy and reliability of the model are quantified by metrics such as root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²). Once validated, the model can be deployed in practical applications to achieve end-to-end intelligent detection: simply input the 3D settlement map of the new sample constructed according to step one into the model, and it can directly and quickly output its SVI. 30 Predicted values ​​no longer require cumbersome laboratory chemical analysis. The model's predictive performance is comparable to SV. 30 The model is together Figure 5 The results will be displayed and compared in the middle.

[0075] Specifically, the final trained SVI 30 The predictive performance of the intelligent detection model for the five SBR reactors is R² = 0.863.

[0076] In summary, compared with existing technologies that rely on manual sampling and laboratory analysis, this invention, by constructing a three-dimensional sedimentation map, transforms the dynamic process of sludge sedimentation into image information that can be used for deep learning for the first time, fundamentally overcoming the inherent lag and human error problems of traditional methods. Unlike online detection instruments based on optical or electrochemical principles, which are susceptible to interference from water color, bubbles, and impurities and require frequent calibration, this invention directly measures based on the macroscopic morphology of the sludge mass sedimentation, exhibiting strong anti-interference capabilities. Therefore, this invention has significant advantages in terms of real-time performance, accuracy, and robustness. Furthermore, the core data acquisition of this method only requires a common digital camera, making it inexpensive and easily integrated into existing wastewater treatment systems to achieve sludge settling performance (SFS) measurement. 30 / SVI 30This invention enables in-situ, real-time, and automatic detection of activated sludge, significantly improving the intelligent operation and maintenance level of water plants. Through the specific design and implementation methods described above, this invention can achieve high-precision, real-time online intelligent detection of activated sludge settling performance, providing reliable technical support for the refined control and intelligent operation of wastewater treatment plants.

[0077] The beneficial effects of this invention are as follows: 1) It realizes complete, visualized, and digital information on the settlement process, fundamentally solving the problem of detection lag.

[0078] This invention constructs a time-series three-dimensional settlement map, which condenses up to 30 minutes of continuous settlement dynamic information into a single image, and realizes a complete record of the entire process, including settlement rate, interface changes and compression characteristics. This makes it possible to conduct rapid and dynamic performance evaluation based on a single measurement, overcoming the inherent drawback of traditional manual measurement methods that are time-consuming.

[0079] 2) The test results are accurate, objective, and have strong anti-interference capabilities.

[0080] This invention directly measures the macroscopic settling morphology of the sludge, avoiding signal interference and measurement errors caused by changes in water quality factors such as water color, bubbles, suspended impurities, and charged ions, which are common with instruments based on optical or electrochemical principles. Furthermore, intelligent judgment based on a deep learning model replaces manual visual readings, eliminating subjective errors and resulting in more objective and reliable results.

[0081] 3) Significantly reduced equipment costs and maintenance complexity.

[0082] The core data acquisition device for this method is simply a regular digital camera, eliminating the need for expensive dedicated optical or electrochemical sensors, resulting in extremely low hardware costs. Furthermore, this method is a non-contact measurement; the sensor (camera) does not directly contact the wastewater, avoiding problems such as probe contamination and membrane damage, significantly reducing the intensity and frequency of subsequent maintenance.

[0083] 4) It possesses strong generalization capabilities and a high level of intelligence.

[0084] By applying convolutional neural network models, complex features deeply related to sedimentation performance can be automatically extracted from 3D sedimentation maps without the need for manual feature design or reliance on prior knowledge, demonstrating strong adaptability. Through techniques such as transfer learning, the model can quickly adapt to the operating conditions of different wastewater treatment plants, exhibiting excellent generalization performance and laying the foundation for truly in-situ, real-time, and intelligent detection.

[0085] 5) It has achieved end-to-end intelligent detection, which has improved operation and maintenance efficiency.

[0086] This method integrates image acquisition, map construction, and model prediction, allowing direct input of sedimentation sequence images and one-click output of SV (Sedimentation Sequence). 30 With SVI 30 Predictive values ​​and full-process automation greatly reduce the workload of operators and provide efficient technical tools for the refined and intelligent operation and maintenance of wastewater treatment plants.

[0087] Example 2 This invention proposes an activated sludge settling performance testing system based on spectral analysis, such as... Figure 6 The above includes: The sub-map acquisition module is used to acquire sludge settling ratio, sludge volume index and color three-dimensional settling map containing sludge settling dynamic information, and to process the color three-dimensional settling map to obtain sub-maps. The color three-dimensional sedimentation map containing dynamic information on sludge sedimentation is specifically as follows: Starting from the initial settling of the sludge, images of the settling process were acquired every 10 seconds for 15 minutes. The sludge images were preprocessed, including background subtraction, image alignment and cropping, to retain the effective settling area. The preprocessed images were arranged in chronological order to obtain a continuous map reflecting the sludge interface settling process. The continuous map was then processed with pseudo-color to obtain a color three-dimensional settling map containing dynamic information about sludge settling.

[0088] The process of processing the color three-dimensional settlement map to obtain sub-maps specifically involves: The color three-dimensional settlement map was divided into three 5-minute sub-maps along the time axis: the first, middle, and last.

[0089] The model building module is used to construct SV based on the sub-map, sludge settling ratio, and AlexNet model. 30 Regression prediction model; SVI constructed based on sub-map, sludge volume index and AlexNet model. 30 Regression prediction model; The SV model is constructed based on sub-maps, sludge settling ratio, and the AlexNet model. 30 The regression prediction model specifically involves: collecting three-dimensional sludge settling maps under different operating conditions during actual operation, and simultaneously measuring the sludge settling ratio (SV) corresponding to the three-dimensional sludge settling maps. 30 As labeling data, sub-maps were obtained based on the 3D sludge settling map; using the AlexNet convolutional neural network architecture, the sub-maps were used as input, and the sludge settling ratio (SV) was calculated. 30 As output, an SV regression prediction model is constructed; during training, weight transfer learning is introduced to accelerate convergence, regularization is used to suppress overfitting, and the training effect is optimized by dynamically adjusting the learning rate.

[0090] The SVI is constructed based on sub-maps, sludge volume index, and the AlexNet model. 30 The regression prediction model is as follows: Three-dimensional sedimentation maps of sludge under different operating conditions were collected during actual operation, and the sludge volume index (SVI) corresponding to the three-dimensional sedimentation maps was measured simultaneously. 30 As labeling data, sub-maps were obtained based on the 3D sedimentation map of sludge; using the AlexNet convolutional neural network architecture, the sludge volume index SV was calculated as input. 30 As output, construct SVI 30 The regression prediction model incorporates weighted transfer learning to accelerate convergence during training, employs regularization to suppress overfitting, and optimizes training performance by dynamically adjusting the learning rate.

[0091] The AlexNet convolutional neural network architecture uses a depth of 34 layers. The training process employs an adaptive learning rate adjustment strategy based on validation loss, specifically using the ReduceLROnPlateau learning rate scheduler. During training, validation loss is used as the monitoring metric. When the validation loss no longer decreases within 10 consecutive epochs, the learning rate is automatically multiplied by 0.5 to reduce it, achieving rapid convergence in the early stages of training and fine-tuning in the later stages.

[0092] The sludge settling performance testing module is used to input the three-dimensional settling map to be analyzed into the SV / SVI regression prediction model and output the predicted value of sludge settling ratio or sludge volume index, thereby realizing the settling performance testing of activated sludge.

[0093] Example 3 Please see Figure 7 As shown, the present invention also provides an electronic device 100 for a method of detecting the settling performance of activated sludge based on spectral analysis; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0094] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the activated sludge settling performance detection method based on graph analysis described in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0095] The at least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor. The processor 102 is the control center of the electronic device 100, connecting various parts of the electronic device 100 via various interfaces and lines.

[0096] The memory 101 in the electronic device 100 stores multiple instructions to implement a method for detecting the settling performance of activated sludge based on graph analysis, and the processor 102 can execute the multiple instructions to achieve the following: Obtain sludge settling ratio, sludge volume index, and color three-dimensional settling maps containing sludge settling dynamic information; process the color three-dimensional settling maps to obtain sub-maps. Based on sub-maps, sludge settling ratio, and the AlexNet model, an SV was constructed. 30 Regression prediction model; SVI constructed based on sub-map, sludge volume index and AlexNet model. 30 Regression prediction model; Input the three-dimensional settlement map to be analyzed into SV30 / SVI 30 The regression prediction model outputs predicted values ​​of sludge settling ratio or sludge volume index, enabling the detection of activated sludge settling performance.

[0097] Example 4 If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0098] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0099] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0100] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0101] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting the settling performance of activated sludge based on spectral analysis, characterized in that, Includes the following steps: Obtain sludge settling ratio, sludge volume index, and color three-dimensional settling maps containing sludge settling dynamic information; process the color three-dimensional settling maps to obtain sub-maps. Based on sub-maps, sludge settling ratio, and the AlexNet model, an SV was constructed. 30 Regression prediction model; SVI constructed based on sub-map, sludge volume index and AlexNet model. 30 Regression prediction model; Input the three-dimensional settlement map to be analyzed into SV 30 / SVI 30 The regression prediction model outputs predicted values ​​of sludge settling ratio or sludge volume index, enabling the detection of activated sludge settling performance.

2. The method for detecting the settling performance of activated sludge based on spectral analysis according to claim 1, characterized in that, The color three-dimensional sedimentation map containing dynamic information on sludge sedimentation is specifically as follows: Starting from the initial settling of the sludge, images of the settling process were acquired every 10 seconds for 15 minutes. The sludge images were preprocessed, including background subtraction, image alignment and cropping, to retain the effective settling area. The preprocessed images were arranged in chronological order to obtain a continuous map reflecting the sludge interface settling process. The continuous map was then processed with pseudo-color to obtain a color three-dimensional settling map containing dynamic information about sludge settling.

3. The method for detecting the settling performance of activated sludge based on spectral analysis according to claim 1, characterized in that, The process of processing the color three-dimensional settlement map to obtain sub-maps specifically involves: The color three-dimensional settlement map was divided into three 5-minute sub-maps along the time axis: the first, middle, and last.

4. The method for detecting the settling performance of activated sludge based on spectral analysis according to claim 1, characterized in that, The SV model is constructed based on sub-maps, sludge settling ratio, and the AlexNet model. 30 The regression prediction model is as follows: Three-dimensional sedimentation maps of sludge under different operating conditions were collected during actual operation, and the sludge settling ratio (SV) corresponding to the three-dimensional sedimentation maps was measured simultaneously. 30 As labeling data, sub-maps were obtained based on the 3D sludge settling map; using the AlexNet convolutional neural network architecture, the sub-maps were used as input, and the sludge settling ratio (SV) was calculated. 30 As output, construct SV 30 The regression prediction model incorporates weighted transfer learning to accelerate convergence during training, employs regularization to suppress overfitting, and optimizes training performance by dynamically adjusting the learning rate.

5. The method for detecting the settling performance of activated sludge based on spectral analysis according to claim 1, characterized in that, The SVI is constructed based on sub-maps, sludge volume index, and the AlexNet model. 30 The regression prediction model is as follows: Three-dimensional sedimentation maps of sludge under different operating conditions were collected during actual operation, and the sludge volume index (SVI) corresponding to the three-dimensional sedimentation maps was measured simultaneously. 30 As labeling data, sub-maps were obtained based on the 3D sedimentation map of sludge; the AlexNet convolutional neural network architecture was used, with the sub-maps as input, and the sludge volume index SV was calculated. 30 As output, an SVI regression prediction model is constructed; during training, weight transfer learning is introduced to accelerate convergence, regularization is used to suppress overfitting, and the training effect is optimized by dynamically adjusting the learning rate.

6. The method for detecting the settling performance of activated sludge based on spectral analysis according to claim 4 or 5, characterized in that, The AlexNet convolutional neural network architecture uses a depth of 34 layers. The training process employs an adaptive learning rate adjustment strategy based on validation loss, specifically using the ReduceLROnPlateau learning rate scheduler. During training, validation loss is used as the monitoring metric. When the validation loss no longer decreases within 10 consecutive epochs, the learning rate is automatically multiplied by 0.5 to reduce it, achieving rapid convergence in the early stages of training and fine-tuning in the later stages.

7. A system for testing the settling performance of activated sludge based on spectral analysis, characterized in that, include: The sub-map acquisition module is used to acquire sludge settling ratio, sludge volume index and color three-dimensional settling map containing sludge settling dynamic information, and to process the color three-dimensional settling map to obtain sub-maps. The model building module is used to construct SV based on the sub-map, sludge settling ratio, and AlexNet model. 30 Regression prediction model; SVI constructed based on sub-map, sludge volume index and AlexNet model. 30 Regression prediction model; The sludge settling performance testing module is used to input the three-dimensional settling map to be analyzed into the SV / SVI regression prediction model and output the predicted value of sludge settling ratio or sludge volume index, thereby realizing the settling performance testing of activated sludge.

8. The activated sludge settling performance testing system based on spectral analysis according to claim 7, characterized in that, The color three-dimensional sedimentation map containing dynamic information on sludge sedimentation is specifically as follows: Starting from the initial settling of the sludge, images of the settling process were acquired every 10 seconds for 15 minutes. The sludge images were preprocessed, including background subtraction, image alignment and cropping, to retain the effective settling area. The preprocessed images were arranged in chronological order to obtain a continuous map reflecting the sludge interface settling process. The continuous map was then processed with pseudo-color to obtain a color three-dimensional settling map containing dynamic information about sludge settling.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the activated sludge settling performance testing method based on graph analysis as described in any one of claims 1 to 6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the activated sludge settling performance testing method based on graph analysis as described in any one of claims 1 to 6.

Images