A biofilm state visual diagnosis method, system and equipment for spiral rotating disc type MABR autotrophic denitrification reactor

Abstract

The present application relates to sewage treatment and intelligent monitoring technical field, specifically relates to a kind of biological membrane state visual diagnosis method, system and equipment for spiral rotating disc type MABR autotrophic denitrification reactor, to overcome underwater shooting interference, realize the clear visualization of biological membrane state, online monitoring and intelligent diagnosis, so that biological membrane thickness control is changed from passive "water quality feedback" mode into active "state early warning" mode, finally serve to improve the stability of integrated autotrophic denitrification process, processing efficiency and intelligent operation level.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment and intelligent monitoring technology, specifically to a method, system, and equipment for visual diagnosis of biofilm status in a spiral rotary MABR autotrophic denitrification reactor. Background Technology

[0002] Aerated membrane biofilm reactors (MABRs) are a novel wastewater treatment process that organically integrates traditional biofilm methods with gas separation membrane technology. Oxygen-permeable membranes supply oxygen to the bioreactor, while microorganisms readily attach and grow on the membrane surface to form a biofilm, thereby efficiently degrading pollutants in the water. In recent years, MABR technology, while fulfilling its functions of efficient organic matter removal and energy saving, has been more widely applied in the biological denitrification process of wastewater treatment. Its heterogeneous mass transfer characteristics make it easy for microorganisms such as ammonia-oxidizing bacteria and nitrite-oxidizing bacteria to accumulate on the membrane surface. To enable the growth and retention of denitrifying microorganisms such as denitrifying bacteria and anaerobic ammonia-oxidizing bacteria in the MABR reactor, existing technology and device developers often use the coupling method of MABR membranes with various traditional biofilm methods and activated sludge processes to construct various types of composite or combined MABR membrane modules. Based on efficient oxygenation, this allows for the coexistence and symbiosis of various functional bacteria such as heterotrophic bacteria, nitrifying bacteria, denitrifying bacteria, and anaerobic ammonia-oxidizing bacteria, achieving stable and efficient carbon and nitrogen removal performance.

[0003] The biofilm attached to the surface of a MABR membrane has a complex structure and diverse functions, serving as the primary site for the degradation of pollutants such as organic matter and ammonia nitrogen. Within this biofilm, the mass transfer directions of oxygen, reaction metabolites, and the matrix are opposite, constituting anisotropic mass transfer. Therefore, the apparent reaction rate of pollutants is significantly correlated with biofilm thickness, especially in MABR reactors with biological nitrogen removal capabilities. On the one hand, the biofilm thickness on the MABR membrane surface needs to reach at least 450 μm to 600 μm to fully utilize the advantages of biofilm stratification and promote the smooth progress of nitrification and denitrification reactions. On the other hand, excessive biofilm proliferation, especially when its thickness exceeds 1200 μm to 1500 μm, significantly increases the mass transfer resistance between the matrix and metabolites within the biofilm, continuously reducing the biological nitrogen removal performance of the MABR. Furthermore, abnormal biofilm detachment is prone to occur, affecting reactor performance. Therefore, the biofilm thickness on the MABR membrane should not be too thin or too thick. Maintaining the optimal biofilm thickness by controlling parameters such as water circulation velocity, gas flushing intensity and frequency, and pollutant load within the reactor is crucial for stable and efficient nitrogen removal in MABR reactors. Precise control of biofilm thickness in existing MABR reactors remains challenging, and the immature in-situ online monitoring technology for biofilm thickness further complicates the optimization and control of the biofilm.

[0004] In the research and practical operation of MABR technology, biofilm thickness measurement is fundamental for assessing its state and optimizing control strategies. Currently, commonly used thickness measurement methods mainly include offline sampling analysis and online indirect monitoring. Offline sampling analysis involves taking a sample of membrane fibers from the reactor and observing it through frozen sections combined with optical or electron microscopy to obtain relatively accurate thickness data. However, this process is destructive, complex, and significantly time-consuming, making it unsuitable for real-time operation control. Online methods mainly rely on microelectrode technology, inserting dissolved oxygen or pH microelectrodes into the biofilm and calculating the thickness of the active biofilm based on the gradient changes in microenvironmental parameters (such as dissolved oxygen concentration). While this method can achieve a certain degree of in-situ monitoring, the microelectrodes themselves are fragile, costly, and require high maintenance, and are difficult to operate stably for long periods in complex engineering environments. These limitations make the development of a non-destructive, low-cost monitoring technology that can reflect the state of the biofilm in real time an urgent and critical problem to be solved in the optimized operation of MABR processes.

[0005] In recent years, combining image recognition, artificial intelligence, and other technologies with biofilm monitoring has become an important research direction in this field. Existing technologies have seen some attempts to analyze biofilm state using visual methods, particularly through image processing to obtain macroscopic morphological information of biofilms, thereby indirectly assessing their state. For example: Chinese invention patent CN119959217B discloses a method and system for monitoring dissolved oxygen concentration within the biofilm of a membrane aerated biofilm reactor (MABR) based on image processing and machine learning models. The monitoring method includes: acquiring image data of the biofilm in a MABR wastewater treatment device during the wastewater treatment process; extracting features from the image data to obtain image feature parameters; inputting the image feature parameters into a biofilm dissolved oxygen prediction model to obtain the effective dissolved oxygen concentration within the biofilm output by the biofilm dissolved oxygen prediction model; the biofilm dissolved oxygen prediction model is trained based on a machine learning model. This patent provides a method for monitoring dissolved oxygen within a biofilm using image recognition via a machine learning model, overcoming the drawbacks of high cost, easy damage, and difficult maintenance of microelectrodes, achieving low-cost and effective monitoring of dissolved oxygen within the biofilm. The core of this patent lies in acquiring images of the MABR biofilm, extracting image features, and using a pre-trained machine learning model to predict the effective dissolved oxygen concentration inside the biofilm, thereby optimizing the aeration process. This technology focuses on indirectly acquiring the key microenvironmental parameter of dissolved oxygen within the biofilm through visual means.

[0006] Chinese invention patent CN120309083B relates to an online monitoring and control system for the biofilm thickness of suspended packing material in a moving bed biofilm reactor (MBBR). The system mainly consists of three modules: monitoring, early warning, and control of the suspended packing biofilm thickness. The monitoring module includes a biofilm image acquisition unit that acquires real-time images of the biofilm through an online monitoring device; an image processing unit processes the images to obtain the biofilm thickness as monitoring data; and a thickness analysis unit further analyzes this data. The early warning module has two functions: first, it determines whether the biofilm thickness is abnormal, issuing an early warning if so; second, it determines whether the biofilm growth or shedding rate is normal, also issuing an early warning if so. The control module combines the monitoring data from the monitoring module and the early warning signals from the early warning module to precisely control the opening of the automatic control valve of the aeration module in the MBBR tank, maintaining the biofilm thickness within the optimal range and ensuring the efficient and stable operation of the water treatment system. This patent acquires biofilm images of the suspended packing material through an online image device, processes them to obtain biofilm thickness data, and issues early warnings for thickness anomalies. Finally, it adjusts the biofilm thickness by controlling aeration and other means to maintain the optimal operating state of the MBBR system. This technology is specifically designed for the suspended packing material in the MBBR process, with the goal of monitoring and controlling its biofilm thickness.

[0007] In the aforementioned schemes, biofilm monitoring using image-based methods mainly develops in two directions: one is to predict the dissolved oxygen concentration inside the biofilm through image features (e.g., CN119959217B), and the other is to perform online monitoring and control of the thickness of suspended packing material in the MBBR process (e.g., CN120309083B). However, when attempting to apply these technologies to the integrated spiral rotary MABR-Anammox autotrophic denitrification reactor targeted by this invention, a fundamental mismatch problem exists, rendering it unsuitable for direct application.

[0008] First, the mismatch between the application and the process is the core obstacle. CN119959217B aims to predict dissolved oxygen concentration, but its output cannot directly reflect the macroscopic physical state of the biofilm (such as excessive thickness or nodular growth), let alone assess the direct impact of this state on the performance of the specific coupled process of "short-cut nitrification (PN)-anammox". CN120309083B targets the free-flowing suspended packing material in MBBR, and its monitoring method and image acquisition conditions are fundamentally different from the fixed, horizontally rotating helical disc membrane module in this invention in terms of structure and movement, making the technical solution untransferable.

[0009] Secondly, image acquisition methods face practical challenges. For helical rotating disc structures, conventional underwater in-situ imaging suffers from severe interference from water refraction, suspended matter, and air bubbles, resulting in blurred images and difficulties in feature extraction, failing to meet the basic image clarity requirements for diagnostic analysis. Existing technologies do not provide an effective solution for acquiring clear, standardized, and panoramic images of such semi-submersible, rotatable large membrane modules.

[0010] More importantly, even with the integration of existing reactor control patents, the decision-making challenge of "sensing lag" remains unresolved. In the previously filed "A spiral rotary MABR autotrophic denitrification reactor device and operation control method," the start-up decision for "biofilm enhancement reduction operation" relies entirely on indirect monitoring and empirical judgment of effluent water quality parameters (such as a decrease in total nitrogen removal rate and an increase in nitrate nitrogen formation rate). This post-feedback mechanism based on "water quality effect" has inherent defects: 1) Severe lag: The damage to mass transfer and microbial community structure caused by excessive biofilm thickness precedes water quality deterioration; by the time a water quality alarm is triggered, the process performance has already been damaged; 2) Vague attribution: Water quality changes are the combined result of multiple factors such as load, temperature, and microbial community imbalance, making it difficult to accurately attribute to excessive biofilm thickness, let alone pinpoint whether the excessive thickness occurs in the inner or outer ring of the rotary disc; 3) Lack of foresight: It is impossible to directly perceive key macroscopic morphologies such as biofilm thickness, nodular protrusions, and color changes, which are precisely the leading indicators for predicting a decline in biofilm activity and mass transfer performance.

[0011] Therefore, for integrated spiral rotary MABR autotrophic denitrification reactors, biofilm thickness control is essentially a reactive, post-hoc response rather than proactive, preventative management. Due to the lack of direct, online, and visual means of sensing the biofilm state, the control process is characterized by blindness, experiential factors, and lag, severely hindering the stability, processing efficiency, and development towards intelligent and refined operation of this advanced technology. Summary of the Invention

[0012] To address the aforementioned technical problems, this invention provides a method, system, and equipment for visual diagnosis of biofilm status in a spiral rotary MABR autotrophic denitrification reactor. This overcomes underwater imaging interference, enabling clear visualization, online monitoring, and intelligent diagnosis of biofilm status. Consequently, biofilm thickness control is transformed from a passive "water quality feedback" mode to an active "status early warning" mode, ultimately serving to improve the stability, treatment efficiency, and intelligent operation level of the integrated autotrophic denitrification process.

[0013] To solve the above-mentioned technical problems, the present invention provides a visual diagnostic method for biofilm status in a spiral disc type MABR autotrophic denitrification reactor, which is applied to a spiral disc type MABR autotrophic denitrification reactor including a membrane reaction zone and a spiral disc type MABR membrane module. Includes the following steps: S1. Water level control and membrane surface exposure: The water level in the membrane reaction zone is lowered to the biofilm visual diagnosis control liquid level, so that the spiral rotating MABR membrane module is exposed above the water surface at a predetermined ratio, thus exposing its surface biofilm. S2, Image Acquisition: The drive system drives the spiral rotary MABR membrane module to rotate at a preset low speed, while simultaneously acquiring a high-definition image sequence of the biofilm surface exposed to the air through the image acquisition unit. S3. Image processing: Preprocess and segment the acquired high-definition image sequence, and extract features from the segmented sensitive regions to quantitatively characterize the macroscopic morphology of the biofilm. S4. Multi-source data fusion and evaluation: Obtain the real-time process parameters of the reactor operation and fuse them with the features used to quantitatively characterize the macroscopic morphology of the biofilm. Input the data into a pre-trained intelligent evaluation model to output the biofilm health status index and / or the performance risk level of the integrated autotrophic denitrification process. S5. Diagnosis and Early Warning: Based on the output of the intelligent assessment model, a structured biofilm status diagnosis report is generated, and an early warning signal of the corresponding level is triggered according to the preset risk threshold, and corresponding control instructions are executed.

[0014] Further, in step S1: the biofilm visual diagnostic control liquid level is set to make the radial exposure ratio of the spiral disc MABR membrane module 40%~50%. After reaching the biofilm visual diagnostic control liquid level, the spiral disc MABR membrane module is kept stopped rotating for a preset time to allow the water surface ripples to subside.

[0015] Furthermore, in step S2: The rotary encoder on the drive system monitors the rotation angle of the helical rotary MABR membrane module in real time. The rotation synchronization trigger module, based on the rotary encoder signal, triggers the image acquisition unit to synchronously capture images of the two spiral surfaces of the spiral rotary MABR membrane assembly every time the spiral rotary MABR membrane assembly rotates by a preset equal division angle, so as to obtain a high-definition image sequence.

[0016] Furthermore, in step S2: the image acquisition adopts a segmented acquisition strategy: the axial length of the spiral rotary MABR membrane module is divided into multiple segments, and the image acquisition unit sequentially acquires images in each of the segments.

[0017] Furthermore, within each segment, as the spiral rotary MABR membrane module rotates one revolution, the linear module of the image acquisition unit drives at least two industrial cameras arranged on both sides of the spiral rotary MABR membrane module to move synchronously by one pitch along the axial direction of the spiral rotary MABR membrane module.

[0018] Furthermore, in step S4: The real-time process parameters include at least one or more of the following: nitrite nitrogen accumulation rate, influent ammonia nitrogen load, total nitrogen removal load, dissolved oxygen concentration in the membrane reaction zone, pH value, and redox potential. The features used to quantitatively characterize the macroscopic morphology of biological membranes include at least one or more of the following: color features, texture features, morphological features, and membrane filament outline visibility. The pre-trained intelligent evaluation model is a machine learning or deep learning evaluation model that can understand the mapping relationship between the features used to quantitatively characterize the macroscopic morphology of biofilms and real-time process parameters.

[0019] This application also provides a visual diagnostic system for the biofilm status of a spiral disc MABR autotrophic denitrification reactor. The system is integrated into the spiral disc MABR autotrophic denitrification reactor, which includes a spiral disc MABR membrane module and water quality sensors within its membrane reaction zone. The biofilm condition visual diagnostic system includes: A liquid level regulating device is used to control the liquid level in the membrane reaction zone; The image acquisition unit includes at least two acquisition sub-units disposed on both sides of the spiral rotary MABR membrane assembly along the axial direction and a rotation synchronization trigger module. The rotation synchronization trigger module is used to receive the angle signal from the rotary encoder on the drive system and trigger the acquisition sub-units to acquire images when the spiral rotary MABR membrane assembly rotates to a preset angle. The data processing and early warning unit, including an image processing server, is configured to receive and process image data acquired by the image acquisition unit and real-time process parameter data from the reactor, fuse the two data and input them into a pre-trained intelligent evaluation model to generate a diagnostic report and an early warning signal. The PLC controller is connected to the liquid level regulating device, image acquisition unit, data processing and early warning unit, and reactor actuators to receive early warning signals and execute corresponding control commands.

[0020] Furthermore, the acquisition subunit includes an industrial camera, a light source that provides illumination for the industrial camera, and a linear module that drives the industrial camera to move axially along the spiral rotary MABR membrane assembly. The two sets of acquisition subunits are symmetrically arranged on both sides of the spiral rotary MABR membrane assembly via independent camera brackets.

[0021] Furthermore, the rotation synchronization triggering module generates a trigger signal based on the angle signal of the rotary encoder, and controls the acquisition subunit to acquire images every time the helical rotary MABR membrane assembly rotates by a fixed equal division angle. The fixed equal division angle is 10 degrees or an integer multiple thereof.

[0022] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described visual diagnostic method for biofilm status in a spiral rotary MABR autotrophic denitrification reactor.

[0023] The beneficial effects of using the present invention are as follows: This invention aims to solve the core challenges of delayed biofilm state sensing, difficult diagnosis, and passive control in integrated spiral rotating MABR-Anammox reactors. The innovation of its technical solution lies not in the improvement of a single step, but in the construction of a complete technical system encompassing specific physical operations, intelligent image analysis, and closed-loop process control. The key points and protections of this invention should revolve around the methodological flow, system structure, core algorithms, and functional modules of this system.

[0024] (1) Key protection points in terms of method and process: a standardized visual diagnostic process that is deeply coupled with the spiral rotary MABR-Anammox reactor. It originated from a preprocessing method for image acquisition specifically designed for high-definition imaging: "lowering water level - partial exposure." This method involves actively adjusting the water level within the reactor to expose the helical rotary MABR membrane module above the water surface at a predetermined ratio, thus eliminating water interference with imaging. Building upon this, the core is a panoramic image acquisition method based on "rotation - synchronization - indexing - segmentation": The helical rotary MABR membrane module is driven to rotate at a low, uniform speed. Rotary encoder signals are used to synchronously trigger two industrial cameras positioned on either side of the helical rotary MABR membrane module along its axis at preset equal intervals (e.g., every 10 degrees), achieving simultaneous, blind-spot-free imaging of both the forward and reverse helical surfaces. Combined with axial segmentation (front, middle, and rear) sampling and an interpolation scanning strategy where "the industrial camera moves one pitch axially, and the helical rotary MABR membrane module rotates one revolution," a high-definition image sequence covering the entire surface of the helical rotary MABR membrane module is systematically acquired. Furthermore, through an intelligent assessment method that integrates "visual-process" multi-source data, multi-dimensional quantitative features of the biofilm's macroscopic morphology (such as nodular protrusions and color distribution) are extracted from images. These features are then fused with key real-time process parameters reflecting the performance of the short-range nitrification-anaerobic ammonia oxidation process (especially nitrite accumulation rate), and input into a pre-trained intelligent assessment model. The final output is a dual judgment of the biofilm health status index and the process performance risk level. Based on this diagnostic result, the system executes an "adaptive-closed-loop" control method, automatically generating and executing differentiated control commands (such as triggering enhanced reduction procedures), thereby achieving a complete closed loop from state perception and intelligent diagnosis to precise regulation.

[0025] (2) Key protection points of system devices and hardware architecture: dedicated hardware systems for implementing the above methods and their collaborative connection relationships. This system uses a helical rotary MABR autotrophic denitrification reactor as its platform and innovatively integrates a dual industrial camera synchronous image acquisition unit. This unit includes acquisition sub-units symmetrically arranged on both sides of the rotary table. Each sub-unit comprises two industrial cameras, matching lenses, a parallel light source, and a linear module that drives their axial movement. To ensure the spatiotemporal accuracy of image acquisition, the system is equipped with a high-precision rotary synchronous trigger module. A rotary encoder monitors the rotation angle of the helical rotary MABR membrane module in real time, and the rotary synchronous trigger module generates synchronous pulses based on the encoder's angle signal, simultaneously triggering both industrial cameras to capture images. The intelligent decision-making and execution of the entire system rely on a hierarchical control architecture: the image processing server acts as the intelligent hub, simultaneously receiving raw image data from the industrial cameras and real-time process parameters from sensors inside the reactor (via a PLC controller); after the server runs an intelligent evaluation model to complete multi-source data fusion analysis and evaluation, it sends diagnostic results and early warning signal commands to the PLC controller; the PLC controller then coordinates and controls all actuators, including drive motors, linear modules, and dewatering valves, to achieve intelligent closed-loop operation from image acquisition to process control.

[0026] (3) Key protection points of software and algorithm modules: core algorithm model that supports intelligent diagnostic function. First, the automatic segmentation algorithm for sensitive areas of biofilms can intelligently identify, locate, and segment key regions containing abnormal morphologies (such as nodules and abnormal color spots) from massive amounts of acquired images, focusing the target for subsequent detailed analysis. Second, the multi-dimensional feature extraction algorithm for macroscopic morphology of biofilms calculates a series of quantitative features in parallel for the segmented regions, including color statistical features, texture features, morphological features (such as the area and distribution density of protrusions), and the visibility of membrane filament contours. This transforms visual information into features that quantitatively characterize the macroscopic morphology of the biofilm, resulting in a structured data vector representing the health status of the biofilm. Finally, and most importantly, there is a dedicated intelligent assessment model for the collaborative evaluation of biofilm status and process risks. This model, after training, can learn and establish a complex mapping relationship between the appearance and morphological features of the biofilm and integrated process operation performance parameters, thereby achieving the fusion of image and process data to output comprehensive status diagnosis and risk warning.

[0027] (4) Application scenarios and key protection points of functions: specific industry problems solved and advanced intelligent functions implemented. This solution is an effective means to address the challenge of visual monitoring of the inner biofilm in spiral rotary MABR modules. By combining rotational exposure with simultaneous imaging, it completely overcomes the industry bottleneck of traditional fixed MABR modules where the inner biofilm layer is obstructed by structural occlusion. Based on this, the entire method and system ultimately achieves advanced functionality of "process adaptive control based on real-time biofilm status early warning." This revolutionarily transforms biofilm management from a "passive response" and "post-event remediation" model that relies on lagging water quality indicators and experience-based judgments to a "proactive prevention" and "precision management" model based on real-time, intuitive morphological characteristics for intelligent diagnosis and early warning. This significantly improves the operational stability, treatment efficiency, and intelligence level of the integrated autotrophic denitrification process.

[0028] This invention is a highly integrated, hardware-software combined innovative solution. Its protection strategy is multi-dimensional, with the "method process" at its core, using "system devices" to protect the hardware implementation, and covering key software algorithms and intelligent decision-making logic as much as possible through "functional modules" and "media storing programs." Its core innovation lies in creatively providing a complete technical path of "clear imaging—intelligent diagnosis—precise control" for a specific spiral rotary MABR autotrophic denitrification reactor, achieving a fundamental breakthrough in addressing the bottleneck problem of this advanced process.

[0029] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0030] Figure 1 This is a flowchart of the visual diagnostic method for the biofilm state of the spiral rotary MABR autotrophic denitrification reactor of the present invention. Figure 2 This is a schematic diagram of the visual diagnostic system for the biofilm state of the spiral rotary MABR reactor of the present invention. Figure 3 This is a schematic diagram showing the connection between the spiral disc MABR reactor biofilm state visual diagnostic system and the spiral disc MABR autotrophic denitrification reactor of the present invention.

[0031] In the attached diagram: 2—Membrane reaction zone; 3—Composite sludge zone; 4—Inlet water distributor; 5—Spiral rotary MABR membrane module; 6—Double-layer hollow rotating shaft; 7—Stainless steel chain; 8—Drive motor reducer; 9—Rotary encoder; 10—Variable frequency controller; 11—Inlet pipe; 12—Inlet valve; 13—Drainage pipe; 14—Drainage valve; 15—Outlet pipe; 16a—First linear module; 16b—Second linear module; 17a—First camera bracket; 17b—Second camera bracket; 18a—First light source; 18b—Second light source; 19a—First industrial lens; 19b—Second industrial lens; 20a—First industrial camera; 20b—Second industrial camera; 21—Water quality detection sensor; 22—Secondary water quality detection meter in the membrane reaction zone; 23—Ultrasonic level gauge; 24—Image processing server; 25—Human-machine interface; 26—Rotary synchronous trigger module; 27—PLC controller. Detailed Implementation

[0032] Referring to the accompanying drawings, specific embodiments of the present invention will be described in detail.

[0033] Reference Figures 1 to 3 This invention provides a visual diagnostic method for the biofilm status of a spiral rotating disc MABR autotrophic denitrification reactor, which is applied to a spiral rotating disc MABR autotrophic denitrification reactor including a membrane reaction zone 2 and a spiral rotating disc MABR membrane module 5.

[0034] The spiral rotary MABR autotrophic denitrification reactor includes a membrane reaction zone 2 and a composite sludge zone 3 below it. The spiral rotary MABR membrane module 5 is horizontally positioned in the middle of the membrane reaction zone 2 via a double-layer hollow rotating shaft 6. Driven by a drive motor reducer 8 and a stainless steel chain 7, the spiral rotary MABR membrane module 5 rotates around the horizontal axis. A rotary encoder 9 is installed on the drive motor reducer 8 to monitor the rotation angle of the spiral rotary MABR membrane module 5. The influent to the spiral rotary MABR autotrophic denitrification reactor is connected to an influent pipe 11 and an influent distributor 4. An influent valve 12 is installed on the influent pipe 11. A downcomer pipe 13 and an effluent pipe 15 are also provided on the membrane reaction zone 2. The effluent is discharged through the effluent pipe 15. A downcomer valve 14 is installed on the downcomer pipe 13 to control the water level in the membrane reaction zone 2. The membrane reaction zone 2 of the spiral rotary MABR autotrophic denitrification reactor is designed with at least two working liquid levels: when the water level in the membrane reaction zone 2 is the normal operating level, it is used for conventional wastewater treatment operation; when the water level is the level for visual diagnosis of the biofilm status, its liquid level is lower than the normal operating level and is formed by active precipitation. This aims to expose part of the MABR biofilm on the spiral rotary MABR membrane module 5 to the air, thereby completely avoiding interference from water on the imaging light path such as refraction and scattering, as well as the decrease in imaging quality due to wastewater color and suspended solids. This creates the necessary physical conditions for subsequently obtaining high-definition, distortion-free images of the biofilm surface. A water quality detection sensor 21 and an ultrasonic level gauge 23 are installed in the membrane reaction zone 2 to detect the real-time process parameters of the reactor.

[0035] This invention provides an exemplary embodiment of a method for visual diagnosis of biofilm status in a spiral rotary MABR autotrophic denitrification reactor, comprising the following steps: S1. Water level control and membrane surface exposure: The water level in the membrane reaction zone 2 is lowered to the biofilm visual diagnosis control level, so that the spiral rotating MABR membrane module 5 is exposed to the water surface at a predetermined ratio, thus exposing its surface biofilm. S2, Image Acquisition: The drive system drives the spiral rotary MABR membrane module 5 to rotate at a preset low speed, while simultaneously acquiring a high-definition image sequence of the biofilm surface exposed to the air through the image acquisition unit. S3. Image processing: Preprocess and segment the acquired high-definition image sequence, and extract features from the segmented sensitive regions to quantitatively characterize the macroscopic morphology of the biofilm. S4. Multi-source data fusion and evaluation: Obtain the real-time process parameters of the reactor operation and fuse them with the features used to quantitatively characterize the macroscopic morphology of the biofilm. Input the data into a pre-trained intelligent evaluation model to output the biofilm health status index and / or the performance risk level of the integrated autotrophic denitrification process. S5. Diagnosis and Early Warning: Based on the output of the intelligent assessment model, a structured biofilm status diagnosis report is generated, and an early warning signal of the corresponding level is triggered according to the preset risk threshold, and corresponding control instructions are executed.

[0036] In this embodiment, the water level control and membrane surface exposure in step S1 aim to create a stable and interference-free physical environment for subsequent high-definition image acquisition. The core operation involves precisely controlling the water level in the membrane reaction zone 2 to partially expose the spiral rotating MABR membrane module 5 above the water surface at a predetermined ratio. By actively and controllably lowering the water level, the biofilm surface, originally submerged in wastewater, is directly exposed to still air. This completely eliminates the interference of water refraction and scattering, as well as the color of wastewater, suspended solids, and bubbles on the imaging light during underwater imaging, laying a crucial physical foundation for the industrial cameras 20a and 20b to acquire high-contrast, distortion-free, and uniformly illuminated high-definition images of the biofilm surface.

[0037] Furthermore, the specific operation process of step S1 is as follows: ① Operation Trigger and Influent Adjustment: The process begins when the system initiates biofilm status diagnosis according to a preset cycle (e.g., every 24 hours), or when a status check is required based on real-time water quality parameters (e.g., a decrease in total nitrogen removal rate, a decrease in nitrite accumulation rate, or an increase in nitrate nitrogen formation rate). First, the PLC controller 27 adjusts the influent flow rate to half of the normal operating flow rate. Simultaneously, the spiral rotary MABR membrane module 5 is continuously rotated at a low speed of 0.5~1.0 rpm to maintain basic mixing within the tank.

[0038] ② Suspension of water intake and activation of dewatering: PLC controller 27 sequentially closes the water inlet valve 12 of membrane reaction zone 2, temporarily stopping the membrane reaction zone 2 from receiving water. Subsequently, PLC controller 27 opens the dewatering valve 14. Under the action of gravity, the wastewater in the upper part of membrane reaction zone 2 is quickly discharged into the external wastewater equalization tank through the dewatering pipe 13, thereby causing the water level in membrane reaction zone 2 to drop continuously.

[0039] ③ Precise Water Level Control and Status Maintenance: The ultrasonic level gauge 23 in the membrane reaction zone monitors water level changes in real time and feeds the data back to the PLC controller 27. The PLC controller 27 controls the opening and closing of the downcomer valve 14 to precisely stabilize the water level at the "biofilm visual diagnostic control level". Furthermore, at this level, the radial exposure ratio of the spiral rotary MABR membrane module 5 (i.e., the ratio of the height of the air exposed at the top of the spiral rotor to the outer diameter of the rotor) is controlled within a predetermined range of 40% to 45%. After the water level stabilizes, the spiral rotary MABR membrane module 5 is kept stationary for 1 to 2 minutes to allow the water surface fluctuations to subside, providing a calm interface for image acquisition.

[0040] In this embodiment, the image acquisition in step S2 aims to achieve standardized, panoramic, and high-efficiency image acquisition of the biofilm state of the spiral rotary MABR membrane module 5. Its core is to systematically acquire a high-definition image sequence covering the entire surface of the rotary disk (including the inner and outer rings, and both front and back sides) through a strategy combining rotational synchronous triggering and axial segmented movement, without introducing excessive hydraulic shear force or prolonged interruption of process operation. Preferably, during image acquisition, the drive system drives the spiral rotary MABR membrane module 5 to rotate at a uniform speed of 0.5 rpm. This low speed setting ensures minimal disk movement within the industrial camera's exposure time, thereby obtaining clear images without motion blur, while avoiding excessive and unexpected hydraulic shear force on the biofilm due to increased rotation speed, which could interfere with its natural state or cause abnormal detachment.

[0041] Further, in step S2, the rotary encoder on the drive system monitors the rotation angle of the helical disc MABR membrane assembly 5 in real time; the rotation synchronization trigger module 26, based on the rotary encoder signal, triggers the image acquisition unit to synchronously capture images of the two helical surfaces of the helical disc MABR membrane assembly 5 every time the helical disc MABR membrane assembly 5 rotates by a preset equal division angle, so as to obtain a high-definition image sequence.

[0042] Specifically, when the image acquisition unit reaches the designated position, the PLC controller 27 sends a command to the frequency converter 10 of the drive motor reducer 8 of the drive system, controlling the spiral rotary MABR membrane module 5 to rotate at a constant speed of 0.5 rpm. The signal from the rotary encoder 9 is simultaneously connected to both the PLC controller 27 and the rotation synchronization trigger module 26. Based on the encoder signal, the rotation synchronization trigger module 26 triggers the image acquisition unit to synchronously photograph the two spiral surfaces of the spiral rotary MABR membrane module 5 every 10 degrees of rotation. Preferably, to achieve a complete scan of the two spiral surfaces of the spiral rotary MABR membrane module 5 without omissions, an interpolation method of "moving one step, rotating one revolution" can be used. That is, during the period when the spiral rotary MABR membrane module 5 rotates a full revolution (360 degrees) to complete the shooting of 36 angle points at that axial position, the industrial camera of the image acquisition unit moves at a constant speed along the axial direction for exactly one pitch. Thus, when one revolution ends, the industrial camera has also moved one pitch in the axial direction, ensuring that a complete "thread" on the spiral surface is continuously covered by the industrial camera's field of view. During the two-round shooting process, the industrial camera moves a total of two pitches axially, achieving double coverage of the spiral surface of that section and further ensuring the integrity of the image.

[0043] In some other exemplary embodiments, the image acquisition in step S2 can employ a segmented acquisition strategy: the axial length of the spiral rotary MABR membrane module 5 is divided into multiple segments, and the image acquisition unit sequentially acquires images within each segment. In this way, segmenting the axial length of the spiral rotary MABR membrane module 5 avoids the time-consuming continuous imaging of the entire spiral rotary MABR membrane module 5, which would affect the continuous operation of the reactor.

[0044] Furthermore, the spiral rotary MABR membrane module 5 is divided into three representative sections in the axial direction: the front section, the middle section, and the rear section. Image acquisition is carried out sequentially in these three sections.

[0045] After the water surface ripples subside, the image acquisition process begins: First, the PLC controller 27 coordinates with the industrial camera of the image acquisition unit to move to the initial shooting position. Then, the spiral rotary MABR membrane module 5 is started to rotate at a constant speed of 0.5 rpm. The industrial camera of the image acquisition unit moves axially at a predetermined speed (matching the rotation speed). Utilizing the pulse signal generated by the high-precision rotary encoder 9 mounted on the drive system (such as the output shaft of the drive motor reducer 8), the industrial camera is synchronously triggered to take one shot every 10 degrees (i.e., 36 equal division points per revolution) of the spiral rotary MABR membrane module 5. Within each segment, the spiral rotary MABR membrane module 5 rotates two revolutions for shooting. Within each segment, the spiral rotary MABR membrane module 5 rotates one revolution. The linear module of the image acquisition unit drives at least two industrial cameras arranged on both sides of the spiral rotary MABR membrane module 5 to move synchronously along the axial direction of the spiral rotary MABR membrane module 5 by one pitch. During this period, the industrial camera group moves a total of two pitches axially. This 10-degree interval strikes a balance between sharpness and coverage efficiency, ensuring sufficient overlap between adjacent images for subsequent stitching or analysis while acquiring a sufficient number of sample points within a limited number of rotations. The mechanism triggered by the rotary encoder 9 ensures the accuracy and repeatability of the shooting angle position.

[0046] After completing the "front-end" shooting, the spiral rotary MABR membrane module 5 pauses rotation. The PLC controller 27 controls the industrial camera group of the image acquisition unit to quickly move to the starting position of the "middle section". Following the same shooting method as the "front-end", the "middle section" and "rear section" shooting are completed sequentially. After all shooting is completed, the industrial camera group returns to the standby position, and the spiral rotary MABR membrane module 5 can resume its normal operating speed according to process requirements.

[0047] During image acquisition, the core mechanism involves the coordinated control of two industrial cameras in the image acquisition unit for simultaneous shooting, encoder angle triggering, low-speed rotation, axial segmentation, and precise movement. This allows for the systematic acquisition of high-definition image sequences covering all key areas of the turntable (front, middle, and rear sections; inner, middle, and outer rings; front and back surfaces) within a very short time. Specifically, the drive system drives the helical turntable MABR membrane module 5 to maintain a low-speed uniform rotation of 0.5 rpm, while simultaneously controlling the industrial cameras equipped with parallel illumination sources to move along the turntable axis in the three representative sections (front, middle, and rear). Utilizing signals generated by the high-precision rotary encoder 9, the two industrial cameras located on both sides of the turntable are simultaneously triggered to capture images every 10 degrees of rotation, ensuring that images of two full revolutions (a total of 72 equally divided angle points) are acquired at each axial position. In particular, by controlling the industrial cameras to move synchronously along the axial direction by one pitch during one revolution of the helical turntable MABR membrane module 5, continuous and seamless scanning of the helical surface is achieved. The structure and motion characteristics of the spiral rotary MABR membrane module 5 allow its inner membrane filaments to be sequentially exposed to the industrial camera's field of view during rotation. Combined with the axial movement of the industrial camera, this completely solves the industry problem of traditional fixed MABR membrane modules where images of the inner membrane filaments cannot be acquired due to structural obstruction. This integrated acquisition method not only overcomes the blind spots of fixed-point shooting, achieving visual coverage of the entire biofilm state, including the innermost ring, but also minimizes disturbance to the biofilm and minimizes interruption of process operation. Therefore, it provides a comprehensive, reliable, and efficient engineering solution for the timely detection and targeted resolution of hidden problems such as excessively thick internal biofilms.

[0048] In this embodiment, the image processing in step S3 is used to perform a series of processing operations on the image sequence acquired by the image acquisition unit. Specifically, image processing includes preprocessing such as denoising, contrast enhancement, and color correction to optimize image quality. Subsequently, based on the image classification model and anomaly detection algorithm, the "sensitive regions" containing key morphological features of biofilms in the image are automatically located and accurately segmented. The purpose of this step is to improve the usability of the original image and focus on the subsequent analysis targets. Preprocessing eliminates some noise introduced by the shooting and enhances features; the intelligent algorithm can automatically identify local areas in the image that may reflect abnormal biofilm conditions (such as significant nodular protrusions and abnormal color spots), thereby concentrating the subsequent more refined feature extraction and analysis work on these most valuable "sensitive regions," significantly improving the overall analysis efficiency and targeting.

[0049] Subsequently, for each segmented sensitive region, multiple dimensions of features capable of quantitatively characterizing the macroscopic morphology of the biomembrane are extracted in parallel. Further, the features used to quantitatively characterize the macroscopic morphology of the biomembrane include at least one or more of the following: color features (such as the mean and standard deviation of each color channel, and the proportion of specific color pixels), texture features (such as energy and entropy calculated based on the gray-level co-occurrence matrix), morphological features (such as the area, perimeter, roundness, and distribution density of protrusions within the region), and the visibility of membrane filament outlines.

[0050] This step completes the transformation from visual information to computable data. By quantifying multiple complementary dimensions such as color (indirectly reflecting the composition and metabolic state of the microbial community), texture (reflecting the roughness and density of the biofilm surface structure), and morphology (directly related to thickness and growth uniformity), a comprehensive and objective digital profile of the biofilm's health status is established, providing detailed input data for subsequent intelligent assessment models.

[0051] In this embodiment, the multi-source data fusion and evaluation in step S4 is the core intelligent decision-making step of the entire method. Features extracted from the image processing step to quantitatively characterize the macroscopic morphology of the biofilm are fused with real-time process parameters from reactor operation, and input into a machine learning or deep learning state assessment model pre-trained on the image processing server 24. This model, after training, can understand the complex mapping relationship between the features used to quantitatively characterize the macroscopic morphology of the biofilm and process performance and real-time process parameters, ultimately outputting two key results: a biofilm health status index (e.g., "normal," "thickened," "excessively thick / abnormal") and an integrated process performance risk level (e.g., "low risk," "medium risk," "high risk"). This step integrates the "phenotypic" information of the organism and the "functional" information of the system through an algorithmic model, achieving diagnosis of the biofilm's own health and predictive, collaborative assessment of the risk to the stable operation of the entire process system. Furthermore, this step automatically generates a structured diagnostic report, which not only includes the status index and risk level but also indicates the specific problem location (e.g., "excessively thick inner biofilm"). Simultaneously, the system automatically triggers different levels of early warning signals (such as "prompt", "warning", and "alarm") based on preset risk thresholds. This step completes the transformation from data analysis to engineering application. The structured diagnostic report provides display content for the human-machine interface 25; while the graded early warning mechanism can provide clear, timely, and differentiated intervention guidance according to the severity of the problem.

[0052] Finally, the PLC controller 27 receives the early warning signal generated by the system and, based on the specific results in the diagnostic report (e.g., "Inner biofilm too thick, high risk"), automatically generates and executes corresponding adaptive control commands. For example, it automatically triggers the reactor's existing "biofilm enhancement and reduction program" and may adjust the speed of the drive motor reducer 8 via the frequency converter 10, or control the downspout valve 14, etc. The significance of this step lies in precisely linking "diagnosis" with "treatment." It transforms biofilm management from a passive operation relying on fixed cycles or experience into an active, intelligent intervention based on real-time, precise status assessment, thereby greatly improving the accuracy, timeliness, and automation level of process control.

[0053] In this process, the acquisition of real-time process parameters is carried out simultaneously with image acquisition and processing. Real-time process parameters are obtained from sensors and monitoring systems in the reactor, such as the water quality sensor 21 in membrane reaction zone 2, the secondary water quality meter 22 in membrane reaction zone, and the ultrasonic level gauge 23. A series of key parameters reflecting the operational performance of the coupled short-cut nitrification and anaerobic ammonia oxidation process are calculated. Furthermore, the real-time process parameters include at least one or more of the following: nitrite nitrogen accumulation rate, influent ammonia nitrogen load, total nitrogen removal load, dissolved oxygen concentration in membrane reaction zone 2, pH value, and redox potential. This step aims to simultaneously acquire "physicochemical index" data of the process operation. Among these, the nitrite nitrogen accumulation rate is the most critical indicator of the efficiency and stability of the short-cut nitrification process. Acquiring these real-time process parameters provides indispensable process-side data support for subsequent fusion analysis with visual morphological characteristics, enabling the evaluation to combine the "appearance" of the biofilm with its actual "function."

[0054] This application also provides an embodiment of a biofilm state visual diagnostic system for a spiral disc type MABR autotrophic denitrification reactor. The biofilm state visual diagnostic system for the spiral disc type MABR autotrophic denitrification reactor is integrated on the spiral disc type MABR autotrophic denitrification reactor. The spiral disc type MABR membrane module 5 and water quality detection sensor (21) are provided in the membrane reaction zone 2 of the spiral disc type MABR autotrophic denitrification reactor.

[0055] A visual diagnostic system for biofilm status in a spiral rotary MABR autotrophic denitrification reactor includes a liquid level control device, an image acquisition unit, a data processing and early warning unit, and a linkage control unit. The PLC controller 27 of the linkage control unit is signal-connected to the liquid level control device, image acquisition unit, data processing and early warning unit, and the reactor's actuators. The PLC controller 27 receives early warning signals and executes corresponding control commands. The various units work collaboratively through signal lines and data streams to achieve intelligent sensing, analysis, and feedback control of the biofilm status. Specifically, the liquid level control device controls the liquid level within the membrane reaction zone 2; the image acquisition unit performs standardized, panoramic image capture of the biofilm surface exposed to air; the data processing and early warning unit is responsible for processing, fusing, analyzing, and making decisions based on multi-source information; and the linkage control unit, with the PLC controller 27 at its core, is responsible for coordinating and controlling the system's diagnostic process and executing commands based on the diagnostic results.

[0056] In some embodiments, the image acquisition unit includes at least two acquisition subunits disposed on both sides of the spiral rotary MABR membrane assembly 5 along its axial direction and a rotation synchronization trigger module 26. The rotation synchronization trigger module 26 receives the angle signal from the rotary encoder 9 on the drive system and triggers the acquisition subunits to acquire images when the spiral rotary MABR membrane assembly 5 rotates to a preset angle. At this time, through the cooperation of the acquisition subunits and the synchronization trigger module 26, at least two acquisition subunits simultaneously photograph the biofilm on the spiral surfaces of the front and back of the spiral rotary MABR membrane assembly 5, thereby achieving strict synchronization between image acquisition and the spatial position of the rotary table, ensuring the systematic nature and repeatability of the image sequence. Further, the rotation synchronization trigger module 26 generates a trigger signal based on the angle signal from the rotary encoder 9, controlling the acquisition subunits to acquire images every time the spiral rotary MABR membrane assembly 5 rotates by a fixed equal division angle. The fixed equal division angle is 10 degrees or an integer multiple thereof. Preferably, the fixed equal division angle is 10 degrees. The rotary encoder 9 detects the rotation angle of the spiral disc MABR membrane module 5 in real time. Every time the spiral disc MABR membrane module 5 rotates 10 degrees, the industrial cameras of the two sets of acquisition subunits are triggered to take pictures, thereby realizing the capture of image sequences of both sides of the rotating spiral disc with standardized positions.

[0057] Furthermore, the two sets of acquisition subunits are symmetrically arranged on both sides of the spiral rotary MABR membrane module 5 via independent camera brackets. Each acquisition subunit includes an industrial camera 20, a light source providing illumination for the industrial camera 20, and a linear module driving the industrial camera to move along the axial direction of the spiral rotary MABR membrane module 5. Furthermore, the industrial cameras 20 in the two sets of acquisition subunits are respectively positioned facing the front and back spiral surfaces of the spiral rotary MABR membrane module 5. The industrial cameras are responsible for capturing high-definition images of the biofilm surface. Each industrial camera 20 integrates or is coaxially equipped with a high-brightness parallel LED illumination source to provide sufficient, uniform, and stable frontal illumination to the shooting area, highlighting the surface texture and color characteristics of the biofilm and ensuring the stability and comparability of image quality. Using two industrial cameras allows for the simultaneous acquisition of images of both sides of the spiral rotary MABR membrane module 5 during a single rotation, avoiding the inefficiency and operational complexity caused by a single camera needing to be moved to the opposite side for secondary shooting.

[0058] Specifically, the first industrial camera 20a, the first industrial lens 19a, and the first parallel light source 18a in the front-side acquisition subunit are mounted on the first camera bracket 17a on the front side and driven by the first linear module 16a on the front side; the second industrial camera 20b, the second industrial lens 19b, and the second parallel light source 18b in the back-side acquisition subunit are mounted on the second camera bracket 17b on the back side and driven by the second linear module 16b on the back side. The two sets of linear modules, one for the front side and one for the back side, are respectively mounted on both sides above the top of the tank along the length direction of the reactor (i.e., the direction of the spiral turntable axis). The PLC controller 27 drives the two sets of linear modules respectively through the servo motors in the linear modules, which can accurately and collaboratively control the synchronous movement and positioning of the front and back camera groups between three preset axial positions: front, middle, and rear.

[0059] Preferably, two MV-HP1400GC color high-performance industrial cameras are selected, with a resolution of 4608×3288 pixels, equipped with a global shutter, and transmitting data via a GigEVision interface. They are mounted on opposite sides of the reactor tank top via camera brackets, with the lens optical axis aligned with the center of the two helical surfaces of the rotating disc, at a shooting distance of approximately 1.5m. The industrial lens is a BT-11C2528MP20-A fixed-focus lens with a focal length of 25mm, ensuring complete coverage of the rotating disc on one side. Two sets of white LED strip light sources, each with a 2.5m long luminous surface, are coaxially mounted with the cameras, providing uniform frontal illumination of over 30,000 Lux to the shooting area. The axial movement mechanism uses two sets of precision ball screw linear modules with an effective stroke of 2.5m and a repeatability of ±0.02mm, driving the camera brackets on the front and back sides respectively. The PLC controller 27 controls the rotation synchronous trigger module 26 through a servo driver; the rotation synchronous trigger module 26 uses a custom trigger board 26 developed based on an STM32 microcontroller to receive pulse signals from the rotary encoder 9 (model: Omron E6B2-CWZ6C, 2000 pulses per revolution) to achieve indexing triggering.

[0060] In some embodiments, the data processing and early warning unit includes an image processing server configured to receive and process image data acquired by the image acquisition unit and real-time process parameter data from the reactor, fuse the two data, and input them into a pre-trained intelligent evaluation model to generate a diagnostic report and an early warning signal. Specifically, the image processing server has a built-in intelligent evaluation model and integrates an early warning signal generation module. Its workflow begins with receiving high-definition raw image data from the image acquisition unit and performing preprocessing (such as noise reduction, contrast enhancement, image cropping, and pixel segmentation); simultaneously, the image processing server obtains online monitoring data of water quality and quantity of the reactor system (such as influent flow rate, ammonia nitrogen, nitrite nitrogen concentration, nitrate nitrogen concentration, pH, dissolved oxygen, ORP, etc.) through a data interface. Subsequently, the image processing server fuses the preprocessed image feature data with the real-time process parameter data and inputs them into the pre-trained intelligent evaluation model for in-depth analysis and intelligent evaluation. Finally, the analysis results of the intelligent evaluation model drive the early warning signal generation module to generate an early warning signal that can quantitatively or qualitatively reflect the health status of the biofilm (such as normal, thickened, excessively thickened) and the integrated process operation risk level (such as low, medium, high risk).

[0061] The image processing server is configured as a Dell Precision T5860 workstation (24), equipped with an Intel Xeon W7-2495x processor, 128GB of memory and an RTX PRO 5000 Blackwell 48G professional graphics card, for running image processing algorithms and intelligent evaluation models.

[0062] In some embodiments, the PLC controller is a Siemens S7-1200 series PLC, which serves as the core of the system control. Its digital / analog modules are connected to the inlet valve 12, the drain valve 14, the drive motor frequency converter 10, the linear module servo driver, the ultrasonic level gauge 23, and the secondary water quality detection instrument 22 in the membrane reaction zone, etc., to receive commands and feedback signals.

[0063] Furthermore, this system is used as an example when applied to an integrated spiral rotary MABR autotrophic denitrification reactor with a treatment capacity of 100 m³ / d. The spiral rotary MABR autotrophic denitrification reactor has an effective volume of 15 m³, housing a set of horizontal spiral rotary MABR membrane modules 5, with an outer diameter of 1.2 m, an axial length of 2.0 m, and a pitch of 0.3 m. The drive motor reducer 8 has a rated power of 1.5 kW. The membrane reaction zone 2 is equipped with an ultrasonic level gauge 23 (model: E+H FMU90) for level monitoring, and an online water quality monitoring system (integrating sensors for ammonia nitrogen, nitrate nitrogen, nitrite nitrogen, dissolved oxygen, pH, and oxidation-reduction potential). The specific steps for running the program are as follows (see...) Figure 1 ): Step S1, Water Level Control and Membrane Exposure: Trigger: The PLC controller 27 automatically initiates a routine diagnostic every 24 hours, or triggers an early warning diagnostic when the real-time monitored nitrite nitrogen accumulation rate (NAR) is below 80% for 2 consecutive hours.

[0064] Regulation and Dewatering: PLC controller 27 adjusts the opening of inlet valve 12 to 50% to slow down the water inflow. Then, it sequentially closes the inlet valves of each section of the membrane area and fully opens the dewatering valve 14, causing the water level in the reactor to drop through the dewatering pipe 13.

[0065] Water level control: The ultrasonic level gauge 23 provides real-time feedback on the water level. The PLC controller 27 controls the dewatering valve 14 through a PID algorithm to precisely stabilize the water level at a height that exposes approximately 42% radially of the spiral disc MABR membrane module 5 (corresponding to a water level drop of approximately 1.0m). After the water level stabilizes, the spiral disc MABR membrane module 5 stops rotating and remains stationary for 90 seconds to allow the water surface to calm down.

[0066] Step S2, Image Acquisition: Positioning: The PLC controller (27) controls the synchronous movement of the linear modules (16a, 16b) on the front and back sides to position the two camera groups to the starting position of the axial "front section".

[0067] Segmented synchronous acquisition: The spiral rotary MABR membrane module 5 is axially divided into front, middle and rear sections, and the front, middle and rear sections are photographed in sequence.

[0068] PLC controller 27 starts drive motor 8, causing the helical rotary MABR membrane module 5 to rotate at a constant low speed of 0.5 rpm. Simultaneously, the linear module is activated to move axially at a speed matching the rotation (0.15 m / min). Every 50 pulses output by rotary encoder 9 correspond to a 9-degree rotation of the rotary disk, and synchronous trigger module 26 simultaneously sends a trigger signal to both cameras for synchronized image capture. During two full rotations (720 degrees) of the helical rotary MABR membrane module 5, a total of 80 images are captured per side, while the linear module drives the camera group to move axially a total of 0.6 m (two pitches).

[0069] Steps S3-S4: Image processing: Image processing server 24 performs batch preprocessing on 480 acquired raw images (3 segments × 2 circles × 80 images), including dark field correction, white balance, and median filtering for noise reduction. Subsequently, a pre-trained U-Net convolutional neural network model is used to segment each image into pixels, automatically identifying and segmenting "sensitive regions" with color anomalies (such as blackening or whitening) and significant textures (such as nodular protrusions). For each segmented region, the following calculations are performed: Color characteristics: the mean and standard deviation of the a (red-green) and b (yellow-blue) components in the LAB color space.

[0070] Texture features: contrast, correlation, and energy of the gray-level co-occurrence matrix (GLCM).

[0071] Morphological characteristics: area, perimeter, circularity, and diameter of the tangent circle within the region (as a proxy feature for local thickness).

[0072] Ultimately, a feature vector set containing hundreds of quantized feature values ​​is generated.

[0073] Step S5: Synchronous acquisition of process parameters: During the same time window of image acquisition, the image processing server 24 obtains and caches the real-time process parameters within that time period from the PLC controller 27 via the OPC protocol, including: ammonia nitrogen, nitrite nitrogen, nitrate nitrogen concentration (used to calculate NAR), dissolved oxygen, pH, influent flow rate, and liquid level.

[0074] Step S6: Multi-source data fusion and intelligent evaluation: Image processing server 24 fuses the statistical values ​​(such as mean, maximum, and quantile) of all feature vectors extracted from the images in this batch with synchronous process parameters (focusing on the NAR time-series variation trend) to form a comprehensive feature vector, which is then input into a pre-trained XGBoost ensemble learning evaluation model. This model outputs two results: Biofilm health status index: a score from 0 to 100, <60 points is "too thick / abnormal", 60-80 points is "thickening warning", and >80 points is "normal".

[0075] Process risk level: "High risk" (requires immediate intervention), "Medium risk" (requires attention), "Low risk".

[0076] Steps S7-S8, Diagnostic Output and Closed-Loop Control: The assessment results generate a structured report and are displayed through the human-machine interface 25. For example: "Diagnosis conclusion: Biofilm health index 65 (thickening warning), process risk is medium; main problem: dark brown nodular protrusions appear in local areas of the inner circle, and the average protrusion density has increased by 15% compared with last week; related process parameters: the current NAR is 75%, showing a slow downward trend."

[0077] If the assessment result is "too thick / abnormal" or "high risk," the system automatically executes closed-loop control: PLC controller 27, based on the problem location in the diagnostic report (e.g., "inner circle"), initiates the existing "biofilm enhancement and reduction program." This program will lower the water level again (this step differs from S1, aiming to utilize shear force) and increase the rotation speed of the spiral rotary MABR membrane module 5 to 3.0 rpm, running for 20 minutes to perform targeted enhanced flushing of the inner circle biofilm. After flushing is completed, the system automatically returns to normal process operation.

[0078] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described visual diagnostic method for biofilm status in a spiral rotary MABR autotrophic denitrification reactor.

[0079] By deploying this system, routine and automated inspections of the reactor biofilm status were achieved three times a week. Compared to the previous method of relying entirely on daily manual offline water quality testing and weekly tank opening checks, the system can detect biofilm thickening trends 24-48 hours in advance and provide early warnings and automatically reduce biofilm thickness before excessive biofilm thickness leads to substandard effluent quality. This has improved the stability of the total nitrogen removal rate of the treatment system (measured by the monthly average coefficient of variation) by approximately 40% and reduced the number of unplanned maintenance operations by more than 60%.

[0080] In summary, this invention provides a complete solution specifically tailored for the "horizontal spiral rotary MABR membrane module" and the "PN-Anammox coupling process." Its image acquisition method (dual-camera synchronization, rotation triggering) and evaluation model (integrating process parameters such as nitrite accumulation rate) are directly targeted at this specific object and process. The output of this invention can directly and accurately reflect the macroscopic physical state of the biofilm (such as excessive thickness, nodular growth) and assess its direct impact on the performance of a specific autotrophic nitrogen removal process, solving the problem of general solutions being "unsuitable for local conditions."

[0081] Furthermore, this invention solves the problems of image blurring caused by refraction and air bubbles in conventional underwater imaging, and the inability to observe the inner membrane filaments of fixed MABR components. It employs a pre-operation method of "lowering the water level to expose part of the turntable to air," combined with a collaborative acquisition strategy of "low-speed rotation, encoder indexing triggering, dual-camera synchronization, and axial segmented movement." This completely eliminates water interference, yielding high-definition images with no optical distortion and uniform illumination, and systematically covering the entire area of ​​the turntable (including the front and back sides, inner and outer rings). In particular, it overcomes the industry challenge of making the inner biofilm "invisible," providing a high-quality visual data foundation for accurate diagnosis.

[0082] This invention utilizes an intelligent assessment model that integrates multi-source "visual-process" data to achieve online, real-time collaborative assessment of biofilm health status and process risks. The system can issue early warnings based solely on early changes in biofilm morphology (such as localized thickening or abnormal color) before significant water quality deterioration, and can accurately pinpoint problem areas (such as "excessive inner layer thickness"). Ultimately, based on this precise diagnostic result, the system can automatically trigger adaptive control commands (such as targeted reduction), overcoming the problems of existing control methods that rely on "post-hoc" feedback from effluent water quality deterioration, resulting in significant lag, ambiguous attribution, and unforeseen issues. This revolutionizes biofilm management from a passive "remedial" mode to a proactive intelligent management mode of "prevention and precise control," significantly improving process stability, treatment efficiency, and intelligence.

[0083] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0084] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. 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 visual diagnostic method for biofilm status in a spiral-rotor MABR autotrophic denitrification reactor, applied to a spiral-rotor MABR autotrophic denitrification reactor including a membrane reaction zone (2) and a spiral-rotor MABR membrane module (5); Its features are, Includes the following steps: S1. Water level control and membrane surface exposure: The water level in the membrane reaction zone (2) is lowered to the biofilm visual diagnosis control level, so that the spiral rotating MABR membrane module (5) is exposed on the water surface at a predetermined ratio, exposing its surface biofilm. S2, Image Acquisition: The drive system drives the spiral turntable MABR membrane module (5) to rotate at a preset low speed, and at the same time, the image acquisition unit synchronously acquires a high-definition image sequence of the biofilm surface exposed to the air; S3. Image processing: Preprocess and segment the acquired high-definition image sequence, and extract features from the segmented sensitive regions to quantitatively characterize the macroscopic morphology of the biofilm. S4. Multi-source data fusion and evaluation: Obtain the real-time process parameters of the reactor operation and fuse them with the features used to quantitatively characterize the macroscopic morphology of the biofilm. Input the data into a pre-trained intelligent evaluation model to output the biofilm health status index and / or the performance risk level of the integrated autotrophic denitrification process. The real-time process parameters include at least one or more of the following: nitrite nitrogen accumulation rate, influent ammonia nitrogen load, total nitrogen removal load, dissolved oxygen concentration, pH value and redox potential in the membrane reaction zone (2); The features used to quantitatively characterize the macroscopic morphology of biological membranes include at least one or more of the following: color features, texture features, morphological features, and membrane filament outline visibility. The pre-trained intelligent evaluation model is a machine learning or deep learning evaluation model that can understand the mapping relationship between the features used to quantitatively characterize the macroscopic morphology of biofilms and real-time process parameters. S5. Diagnosis and Early Warning: Based on the output of the intelligent assessment model, a structured biofilm status diagnosis report is generated, and an early warning signal of the corresponding level is triggered according to the preset risk threshold, and corresponding control instructions are executed.

2. The method for visual diagnosis of biofilm status in a spiral rotary MABR autotrophic denitrification reactor according to claim 1, characterized in that, In step S1: the biofilm visual diagnostic control liquid level is set to make the radial exposure ratio of the spiral disc type MABR membrane module (5) 40%~50%. After the biofilm visual diagnostic control liquid level is reached, the spiral disc type MABR membrane module (5) is kept from rotating for a preset time so that the water surface ripples are calmed.

3. The method for visual diagnosis of biofilm status in a spiral rotary MABR autotrophic denitrification reactor according to claim 1, characterized in that, In step S2: The rotary encoder (9) on the drive system monitors the rotation angle of the helical rotary MABR membrane module (5) in real time; The rotation synchronization trigger module (26) triggers the image acquisition unit to synchronously capture images of the two spiral surfaces of the spiral rotary MABR membrane assembly (5) every time the spiral rotary MABR membrane assembly (5) rotates by a preset equal division angle, according to the signal from the rotary encoder (9), so as to obtain a high-definition image sequence.

4. The method for visual diagnosis of biofilm status in a spiral rotary MABR autotrophic denitrification reactor according to claim 3, characterized in that, In step S2: The image acquisition adopts a segmented acquisition strategy: the axial length of the spiral rotary MABR membrane module (5) is divided into multiple segments, and the image acquisition unit sequentially acquires images in each of the segments.

5. The method for visual diagnosis of biofilm status in a spiral rotary MABR autotrophic denitrification reactor according to claim 3, characterized in that, As the spiral rotary MABR membrane module (5) rotates one revolution, the linear module of the image acquisition unit drives at least two industrial cameras arranged on both sides of the spiral rotary MABR membrane module (5) to move synchronously by one pitch along the axis of the spiral rotary MABR membrane module (5).

6. A visual diagnostic system for biofilm status in a spiral rotary MABR autotrophic denitrification reactor, characterized in that, The system is integrated into the spiral disc type MABR autotrophic denitrification reactor. The membrane reaction zone (2) of the spiral disc type MABR autotrophic denitrification reactor is equipped with a spiral disc type MABR membrane module (5) and a water quality detection sensor (21). The system employs the visual diagnostic method for the biofilm status of the spiral disc type MABR autotrophic denitrification reactor as described in any one of claims 1-5. The biofilm condition visual diagnostic system includes: A liquid level regulating device is used to control the liquid level in the membrane reaction zone (2); The image acquisition unit includes at least two acquisition sub-units located on both sides of the spiral turntable MABR membrane assembly (5) along the axial direction and a rotation synchronization trigger module (26). The rotation synchronization trigger module (26) is used to receive the angle signal from the rotary encoder (9) on the drive system and trigger the acquisition sub-units to acquire images when the spiral turntable MABR membrane assembly (5) rotates to a preset angle. The data processing and early warning unit includes an image processing server (24), which is configured to receive and process image data acquired by the image acquisition unit and real-time process parameter data from the reactor, fuse the two and input them into a pre-trained intelligent evaluation model to generate a diagnostic report and an early warning signal. The PLC controller (27) is connected to the actuators of the liquid level regulating device, image acquisition unit, data processing and early warning unit and reactor, and is used to receive early warning signals and execute corresponding control commands.

7. The biofilm state visual diagnostic system for a spiral rotary MABR autotrophic denitrification reactor according to claim 6, characterized in that, The acquisition subunit includes an industrial camera, a light source that provides illumination for the industrial camera, and a linear module that drives the industrial camera to move axially along the spiral rotary MABR membrane assembly (5). The two acquisition subunits are symmetrically arranged on both sides of the spiral rotary MABR membrane assembly (5) via independent camera brackets.

8. The biofilm state visual diagnostic system for a spiral rotary MABR autotrophic denitrification reactor according to claim 6, characterized in that, The rotation synchronization trigger module (26) generates a trigger signal according to the angle signal of the rotary encoder (9) when the helical turntable MABR membrane assembly (5) rotates by a fixed equal division angle, and controls the acquisition subunit to perform image acquisition. The fixed equal division angle is 10 degrees or an integer multiple thereof.

9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the visual diagnostic method for biofilm status in a spiral rotary MABR autotrophic denitrification reactor as described in any one of claims 1-5.

Images