Zebrafish breeding intelligent monitoring and early warning system and method based on BIM and multi-terminal cooperation

Abstract

The application discloses a zebrafish breeding intelligent monitoring and early warning system and method based on BIM and multi-terminal cooperation, belongs to the technical field of laboratory animal breeding and intelligent monitoring, and comprises a multi-parameter data acquisition module, a BIM three-dimensional modeling and space mapping module, a multi-sensor space-time data fusion module, a growth stage adaptive early warning module, a multi-terminal cooperative intelligent scheduling module, a three-dimensional visual operation and maintenance module and a historical data and traceability module. The application solves the poor adaptability of traditional fixed models by automatically identifying the growth stage and dynamically adjusting the weight, greatly reduces the false negative rate and the false positive rate, introduces a multivariate time series prediction model to realize early warning of future water quality parameters, adjusts the original post-warning to active intervention, saves valuable time window for abnormal disposal, realizes parameter deduction of positions where sensors are not installed and sensor fault self-diagnosis based on the parameter field reconstruction technology of BIM water route topological relationship, and eliminates the monitoring blind area.

Description

Technical Field

[0001] This invention relates to the field of laboratory animal breeding and intelligent monitoring technology, and in particular to an intelligent monitoring and early warning system and method for zebrafish breeding based on BIM and multi-terminal collaboration. Background Technology

[0002] Zebrafish, as a model organism, is widely used in life sciences, drug development, environmental toxicology, and other fields. Laboratory zebrafish farming has extremely high requirements for environmental factors such as water quality parameters (water temperature, pH, dissolved oxygen, ammonia nitrogen, nitrite, etc.), photoperiod, and water flow velocity, and the sensitivity of different growth stages (embryonic stage, larval stage, juvenile stage, adult stage, and reproductive stage) to various parameters varies significantly.

[0003] Existing zebrafish farming monitoring systems have the following shortcomings: (a) The early warning mechanism is rigid and lacks adaptability to different stages: The existing system uses a weighted scoring model with fixed thresholds or fixed weights for anomaly detection, which cannot adapt to the dynamic changes in parameter sensitivity of zebrafish at different growth stages, resulting in a high false alarm rate in the juvenile stage and a high false alarm rate in the adult stage.

[0004] (ii) Lack of spatial perception and predictive ability: Existing systems treat each aquaculture tank as an isolated monitoring point, ignoring the physical coupling caused by the water supply and drainage pipe connections within the aquaculture racks. This prevents the use of upstream tank data to predict downstream tank conditions and hinders self-calibration during sensor drift. Furthermore, existing systems can only issue alerts after the fact, failing to predict future water quality deterioration trends and often missing the optimal intervention window.

[0005] (iii) Low efficiency of multi-terminal collaboration: Existing systems mostly use manual notification or order-grabbing modes. Task allocation does not take into account the matching degree between fault type and maintenance personnel skills, nor does it take into account personnel location and task load, resulting in frequent problems such as response timeout, duplicate processing, and skill mismatch.

[0006] Therefore, designing a zebrafish farming monitoring system capable of adaptive early warning during growth stages, spatial topology deduction and prediction, and intelligent multi-terminal collaboration has significant practical application value. Based on this, a zebrafish farming intelligent monitoring and early warning system and method based on BIM and multi-terminal collaboration is proposed. Summary of the Invention

[0007] The purpose of this invention is to solve the problems existing in the prior art, and to propose an intelligent monitoring and early warning system and method for zebrafish farming based on BIM and multi-terminal collaboration.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: The intelligent monitoring and early warning system for zebrafish farming based on BIM and multi-terminal collaboration includes a multi-parameter data acquisition module, a BIM 3D modeling and spatial mapping module, a multi-sensor spatiotemporal data fusion module, a growth stage adaptive early warning module, a multi-terminal collaborative intelligent scheduling module, a 3D visualization operation and maintenance module, and a historical data and traceability module. The multi-parameter data acquisition module is used to collect water quality parameters, environmental parameters, equipment status parameters, and image and video data of the aquaculture tank in real time. The BIM 3D modeling and spatial mapping module is used to construct a high-precision BIM model of the zebrafish farming rack, establish the spatial topology relationship of the farming tank, sensors, and pipelines, and assign a unique code to each farming tank and sensor to realize the dynamic binding of data with BIM spatial nodes. The multi-sensor spatiotemporal data fusion module is used to perform spatial interpolation fusion of the measured data of the current cylinder block sensor with the data of adjacent spatial nodes in the BIM model to generate a three-dimensional dynamic parameter field, and to perform self-verification and virtual sensing compensation on the sensor data. The growth stage adaptive early warning module is used to dynamically adjust the weight coefficients of the early warning scoring model according to the current growth stage of the zebrafish, and output the predicted values ​​of key parameters for future periods based on the multivariate time series prediction model. The combination of the current parameter values ​​and the predicted values ​​of future parameters realizes a dual-scale composite early warning. The multi-terminal collaborative intelligent scheduling module is used to achieve intelligent task distribution, progress synchronization and conflict resolution through edge computing nodes based on early warning information and fault root cause analysis results, combined with the location, skill tags and current task load of maintenance personnel; The three-dimensional visualization operation and maintenance module is used to integrate and display real-time monitoring data, parameter field interpolation results, early warning information and BIM model, and provide a three-dimensional navigation path to the abnormal cylinder block; The historical data and traceability module is used to store all monitoring data, operation records and abnormal events, and supports traceability analysis and experimental data correlation mining.

[0009] As a preferred embodiment, the BIM 3D modeling and spatial mapping module is used for: Extract the waterway topology from the BIM model, including the connection sequence of each cylinder, the direction of water flow, and the pipe diameter, and construct a directed graph structure representing the transmission relationship from the water source to the upstream cylinder and then to the downstream cylinder. Establish a four-dimensional mapping relationship between equipment codes, spatial locations, data interfaces, and waterway topology adjacency relationships.

[0010] As a preferred embodiment, the multi-sensor spatiotemporal data fusion module is specifically used for: Spatial domain fusion: Based on the spatial coordinates and waterway topology provided by the BIM model, Kriging interpolation or inverse distance weighted interpolation methods are used to calculate the parameter estimates of the current cylinder or nodes without installed sensors based on the measured data of adjacent cylinders and upstream and downstream cylinders, and to construct a three-dimensional dynamic parameter field. Sensor self-calibration: The measured value of the current cylinder is compared with the interpolated estimated value of the upstream and downstream. If the deviation exceeds the preset threshold, it is determined that the current sensor is drifting or malfunctioning, and a sensor calibration warning is triggered. Time-domain fusion: Perform moving average filtering on multiple time-point data of the same cylinder block and the same parameter to remove peak noise; Multi-source alignment: Align water quality parameters, environmental parameters, and image recognition feature values ​​with the same timestamp to form a sample vector for input to the prediction model.

[0011] As a preferred embodiment, the growth stage adaptive early warning module includes: Growth Stage Identification and Weight Adaptive Unit: Establish a zebrafish growth stage-early warning weight mapping knowledge base that includes five stages: embryonic stage, larval stage, juvenile stage, adult stage, and reproductive stage. By recording the introduction date of zebrafish in each tank and combining it with the average pixel area of ​​the fish body output by the image recognition module, the current growth stage is automatically determined, and the corresponding weight coefficient is automatically loaded according to the identified stage. Multivariate time series prediction unit: Construct and pre-train a multivariate time series prediction model based on Long Short-Term Memory Network (LSTM) or Temporal Convolutional Network (TCN). The model input is multi-parameter time series data including water temperature, pH, dissolved oxygen, conductivity, ammonia nitrogen, nitrite, light intensity, and feeding amount within a past time window. The output is the predicted values ​​of key parameters for the next 15 to 60 minutes. Composite deviation score calculation unit: calculates the current deviation score separately. and future deviation score The calculation formula is: in Rate the health of the equipment. For the parameter abnormal volatility, This represents the historical average abnormal volatility. This is an image recognition anomaly index. These are the weighting coefficients; Hierarchical composite early warning and root cause analysis unit: based on and The values ​​and preset thresholds trigger different levels of warnings, and the attention mechanism weights or SHAP values ​​of the prediction model are called to analyze the contribution of each input parameter to the prediction results, output the main causal parameters, and generate warning information with root cause hints and handling suggestions.

[0012] As a preferred embodiment, the warning levels set by the hierarchical composite early warning and root cause analysis unit include: Early warning: ; General warning: ; Important Warning: ; Emergency Warning: ; The advance warning is used to issue an early warning when the predicted value of a parameter is about to exceed the standard in a future period, while the current parameter is still within the normal range.

[0013] As a preferred embodiment, the multi-terminal collaborative intelligent scheduling module is specifically used for: After receiving the early warning information, extract the abnormal location, abnormal type, root cause analysis results, and required skill tags; Query the list of online maintenance personnel terminals to obtain the current location, skill tags, and current task load of each terminal; A weighted matching algorithm is used to calculate the task suitability of each operations and maintenance personnel: Select the terminal with the highest compatibility to push the task; A weighted matching algorithm is used to calculate the task suitability of each operations and maintenance personnel: Select the terminal push task with the highest compatibility; If no response is received from the terminal within the preset time, the task will be automatically pushed to the terminal with the second highest compatibility. Establish a global synchronization mechanism for operation timestamps and processing progress. When multiple terminals are detected responding to the same task simultaneously, retain the terminal that responded earliest, cancel the task of the remaining terminals, and send a conflict notification.

[0014] As a preferred embodiment, the 3D visualization operation and maintenance module is specifically used for: Develop a 3D visualization interface based on Unity or Three.js engine, and mark the operating status of each aquaculture tank on the BIM model with different colors. The colors include green for normal, blue for early warning, yellow for general warning, orange for important warning, and red for emergency warning. Supports overlay display of parameter field thermal maps, intuitively showing the spatial distribution and diffusion trend of temperature field, dissolved oxygen field, and ammonia nitrogen concentration field on the breeding rack; Clicking on any aquaculture tank will bring up a detailed information panel, displaying real-time parameters, future time period prediction curves, current growth stage, current weight coefficient, historical warning records, and maintenance logs; After an abnormal task is dispatched, the system automatically plans the optimal 3D path from the current location of the maintenance personnel to the abnormal cylinder, highlights the navigation on the BIM model, and synchronizes it to the mobile device.

[0015] A smart monitoring and early warning method for zebrafish farming based on BIM and multi-terminal collaboration includes the following steps: Step S1: Construct a BIM model of the zebrafish farming rack, deploy sensors, cameras and edge computing nodes, and configure the early warning weight coefficients and maintenance personnel information for each growth stage; Step S2: Real-time collection of water quality parameters, environmental parameters, equipment status parameters, and image and video data for each aquaculture tank; Step S3: Preprocess and align the collected data spatiotemporally to form a sample vector; Step S4: Map the preprocessed data to the corresponding spatial nodes of the BIM model, construct a three-dimensional dynamic parameter field based on the BIM waterway topology using spatial interpolation, and diagnose the sensor status by cross-verifying upstream and downstream sensor data. Step S5: Based on the date the fry were introduced and the average size of the fish in the image, the growth stage of the zebrafish in the tank is automatically determined, and the corresponding warning weight coefficient is loaded from the knowledge base; Step S6: Extract multivariate time series data from a past time window, input it into the multivariate time series prediction model, output the predicted values ​​of key parameters for future periods, and calculate the current deviation score and the future deviation score respectively; Step S7: Determine whether to trigger an alert and the alert level based on the current deviation score, the future deviation score and the preset threshold. If an alert is triggered, output the main cause parameters and generate an alert message with root cause prompts and handling suggestions. Step S8: Assign tasks based on the abnormal location, required skill tags, location of maintenance personnel, and load calculation task suitability in the early warning information, and track task progress in real time; Step S9: Plan the optimal three-dimensional path to the abnormal cylinder for the maintenance personnel receiving the task and perform navigation. After handling the abnormality on-site, record the processing process and results through the terminal. Step S10: Verify whether the anomaly has been resolved. If it has, close the task and update the BIM model status. If it has not been resolved, upgrade the warning level and reassign the task. Store all data in the database for source analysis and model iteration optimization.

[0016] As a preferred embodiment, the multivariate temporal prediction model in step S6 employs a Long Short-Term Memory (LSTM) network or a Temporal Convolutional Network (TCN), and the model input tensor shape is: ,in Take 12 to 24 as the corresponding timeframe for the past 1 to 2 hours. It includes 8 to 12 relevant variables such as water temperature, pH, dissolved oxygen, conductivity, ammonia nitrogen, nitrite, light intensity, and feeding amount. The model output is the predicted values ​​of ammonia nitrogen concentration and dissolved oxygen concentration for the next 15 to 60 minutes. The current deviation score and the future deviation score are calculated according to the following formula: in Rate the health of the equipment. For the parameter abnormal volatility, This represents the historical average abnormal volatility. This is an image recognition anomaly index. The weighting coefficients are dynamically applied based on the growth stage.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention solves the problem of poor adaptability of traditional fixed models by automatically identifying growth stages and dynamically adjusting weights, thereby significantly reducing the false alarm rate and the missed alarm rate. At the same time, it introduces a multivariate time series prediction model to achieve early warning of future water quality parameters, changing the original post-event alarm to proactive intervention, and gaining valuable time window for anomaly handling.

[0018] 2. This solution utilizes parameter field reconstruction technology based on BIM waterway topology to achieve parameter extrapolation and sensor fault self-diagnosis at locations where sensors are not installed, thus eliminating monitoring blind spots. Combined with multi-sensor spatiotemporal fusion and image recognition enhancement, it significantly improves system robustness and monitoring reliability. Attached Figure Description

[0019] Figure 1 This is a framework diagram of the intelligent monitoring and early warning system for zebrafish farming based on BIM and multi-terminal collaboration proposed in this invention. Figure 2 This is a schematic diagram of the adaptive and composite early warning process for the growth stage of the present invention; Figure 3 This is a flowchart of the intelligent monitoring and early warning method for zebrafish farming based on BIM and multi-terminal collaboration proposed in this invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0022] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0023] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0024] Example, refer to Figures 1 to 3 A smart monitoring and early warning system for zebrafish farming based on BIM and multi-terminal collaboration, comprising: The multi-parameter data acquisition module is used to collect water quality parameters, environmental parameters, equipment status parameters, and image and video data of the aquaculture tank in real time. The BIM 3D modeling and spatial mapping module is used to construct a high-precision BIM model of the zebrafish breeding rack, establish the spatial topology relationship of the breeding tank, sensors, and pipelines, and assign a unique code to each breeding tank and sensor to realize the dynamic binding of data with BIM spatial nodes. The multi-sensor spatiotemporal data fusion module is used to perform spatial interpolation fusion of the measured data of the current cylinder block sensor with the data of adjacent spatial nodes in the BIM model to generate a three-dimensional dynamic parameter field, and to perform self-verification and virtual sensing compensation on the sensor data. The growth stage adaptive early warning module is used to dynamically adjust the weight coefficients of the early warning scoring model according to the current growth stage of the zebrafish, and output the predicted values ​​of key parameters for future periods based on the multivariate time series prediction model. It combines the current parameter values ​​and the predicted values ​​of future parameters to achieve dual-scale composite early warning. The multi-terminal collaborative intelligent scheduling module is used to intelligently distribute tasks, synchronize progress, and resolve conflicts by using edge computing nodes, based on early warning information and root cause analysis results, combined with the location, skill tags, and current task load of maintenance personnel, and in conjunction with the location of maintenance personnel and skill tags. The 3D visualization operation and maintenance module is used to integrate and display real-time monitoring data, parameter field interpolation results, early warning information and BIM model, and provide a 3D navigation path to the abnormal cylinder. The historical data and traceability module is used to store all monitoring data, operation records and abnormal events, and supports traceability analysis and experimental data correlation mining.

[0025] Furthermore, the BIM 3D modeling and spatial mapping module is also used to: extract the waterway topology in the BIM model, including the connection sequence of each cylinder, the direction of water flow, and the pipe diameter; construct a directed graph structure representing the transmission relationship of "water source → upstream cylinder → downstream cylinder"; and establish a four-dimensional mapping relationship of "equipment code - spatial location - data interface - waterway topology adjacency relationship".

[0026] Furthermore, the multi-sensor spatiotemporal data fusion module is specifically used for: Spatial domain fusion: Based on the spatial coordinates and waterway topology provided by the BIM model, Kriging interpolation or inverse distance weighted interpolation methods are used to calculate the parameter estimates of the current cylinder or nodes without installed sensors based on the measured data of adjacent cylinders and upstream and downstream cylinders, and to construct a three-dimensional dynamic parameter field. Sensor self-calibration: The measured value of the current cylinder is compared with the interpolated estimated value of the upstream and downstream. If the deviation exceeds the preset threshold (e.g., 15%), it is determined that the current sensor is drifting or malfunctioning, and a sensor calibration warning is triggered. Time-domain fusion: Perform moving average filtering on multiple time-point data of the same cylinder block and the same parameter to remove peak noise; Multi-source alignment: Align water quality parameters, environmental parameters, and image recognition feature values ​​with the same timestamp to form a sample vector for input to the prediction model.

[0027] Furthermore, the growth stage adaptive early warning module includes the following sub-units: Growth Stage Recognition and Weight Adaptive Unit: A knowledge base for zebrafish growth stage-early warning weight mapping is established, encompassing five stages: embryonic stage (0-3 days), larval stage (4-30 days), juvenile stage (1-3 months), adult stage (3-12 months), and reproductive stage. The system automatically determines the current growth stage by recording the introduction date of zebrafish in each tank and combining this with the average pixel area of ​​the fish body output by the image recognition module, while also allowing manual correction by the experimenter. The system automatically loads the corresponding weight coefficients based on the identified stage. An example weight configuration is: embryonic period. Larval stage Juvenile stage Adult fish period Breeding season .

[0028] Multivariate Temporal Prediction Unit: Constructs and pre-trains a multivariate temporal prediction model based on a Long Short-Term Memory (LSTM) network or a Temporal Convolutional Network (TCN); the model input tensor shape is... ,in Take 12-24 (corresponding to the past 1-2 hours, with a sampling interval of 5 minutes). It includes 8-12 relevant variables such as water temperature, pH, dissolved oxygen, conductivity, ammonia nitrogen, nitrite, light intensity, and feeding amount; the model output is the predicted value of key parameters for the next 15 minutes, 30 minutes, and 60 minutes, including at least ammonia nitrogen concentration and dissolved oxygen concentration.

[0029] Composite deviation score calculation unit: calculates the current deviation score separately. and future deviation score The calculation formula is: in: The equipment health score is calculated based on sensor calibration status, equipment runtime, and historical fault records. The higher the score, the healthier the equipment. This represents the abnormal volatility of the current parameters (calculated based on the ratio of the standard deviation of each parameter to its suitable range over the past hour). This is the historical average abnormal volatility (historical statistical value of the same cylinder block at the same time period, used for normalization). The image recognition anomaly index (with a value range of 0-1, a weighted comprehensive score based on three types of abnormal behaviors: floating, clustering, and death, output by deep learning models such as YOLOv8). Let be the weighting coefficient, satisfying .

[0030] Hierarchical composite early warning and root cause analysis unit: based on and Different levels of alerts are triggered by values ​​and preset thresholds: Advanced warning (blue): ; General warning (yellow): ; Important Warning (Orange): ; Emergency Warning (Red): .

[0031] When an alert is triggered, the attention mechanism weights or SHAP values ​​of the prediction model are invoked to analyze the contribution of each input parameter to the prediction result. The top three input features with the highest contribution are output as the main causal parameters, and disposal suggestions are generated based on a pre-set expert rule base to form a complete alert information with root cause indications.

[0032] Furthermore, the multi-terminal collaborative intelligent scheduling module is specifically used for: receiving early warning information, extracting the abnormal location (BIM spatial node), abnormal type, root cause analysis results, and required skill tags (such as "water quality control," "equipment maintenance," and "fish disease treatment"); querying the online maintenance personnel terminal list to obtain the current location (via indoor positioning or manual check-in), skill tags, and current task load (number of incomplete tasks) for each terminal; and using a weighted matching algorithm to calculate the task suitability of each maintenance personnel. The system selects the terminal with the highest compatibility to push the task; if no response is received from the terminal within a preset time (e.g., 30 seconds), the task is automatically pushed to the terminal with the second highest compatibility; a global synchronization mechanism for operation timestamps and processing progress is established. When multiple terminals are detected responding to the same task at the same time, the terminal that responded earliest is retained, the task of the other terminals is canceled and a conflict prompt is sent.

[0033] Furthermore, the 3D visualization operation and maintenance module is specifically used for: developing a 3D visualization interface based on the Unity or Three.js engine, marking the operating status of each aquaculture tank on the BIM model with different colors, including: green (normal), blue (advanced warning), yellow (general warning), orange (important warning), and red (emergency warning); supporting the overlay display of parameter field heat maps, intuitively showing the spatial distribution and diffusion trend of temperature field, dissolved oxygen field, and ammonia nitrogen concentration field on the aquaculture rack; clicking on any aquaculture tank will pop up a detailed information panel, displaying real-time parameters, the prediction curve for the next 30 minutes, the current growth stage, the current weight coefficient, historical warning records, and maintenance logs; after an abnormal task is dispatched, it will automatically plan the optimal 3D path from the current location of the operation and maintenance personnel to the abnormal tank, highlight the navigation on the BIM model, and synchronize it to the mobile terminal.

[0034] Furthermore, the historical data and traceability module is specifically used for: storing sensor data using a time-series database (with a retention period of 3 years), and storing device ledgers, maintenance records, and user operation logs using a relational database; supporting the querying of historical data by time, tank number, parameter type, warning level, and other conditions; providing an "anomaly rewind" function: it can replay the change curves of all parameters, aquarium video snapshots, and all warning and scheduling decision records made by the system within 1 hour before and after the anomaly occurred, and supports the correlation analysis of experimental data and aquaculture environment data.

[0035] A smart monitoring and early warning method for zebrafish farming based on BIM and multi-terminal collaboration includes the following steps: Step S1: System Initialization and BIM Modeling Construct a BIM model of the zebrafish farming rack, labeling all tanks, sensors, pipelines, and water flow directions; deploy all sensors, cameras, and edge computing nodes; and configure the warning weight coefficients and maintenance personnel information for each growth stage.

[0036] Step S2: Real-time multi-source data acquisition: Water quality parameters, environmental parameters, equipment status, and video images of each tank are collected at a preset frequency. Water quality parameter collection frequency: water temperature, pH, dissolved oxygen, and conductivity are collected once every 1 minute, while ammonia nitrogen and nitrite are collected once every 30 minutes. Image data collection frequency: one frame every 10 seconds, which is automatically increased to one frame per second in abnormal conditions.

[0037] Step S3: Data preprocessing and spatiotemporal alignment: Numerical data is filtered for noise reduction and missing value interpolation is performed; image data is dehazed, de-reflected, and enhanced; all data are aligned with the same timestamp to form a sample vector.

[0038] Step S4: BIM spatial modeling and parametric field reconstruction: The preprocessed data is mapped to the corresponding spatial nodes of the BIM model; based on the BIM waterway topology, the Kriging interpolation method is used to construct the three-dimensional dynamic parameter field of the entire structure; through cross-verification of upstream and downstream sensor data, the current sensor is diagnosed to determine whether there is drift or failure.

[0039] Step S5: Growth Stage Identification and Weight Adaptation: The system automatically determines the growth stage of the zebrafish in the tank based on the date the fry were introduced and the average size of the fish recognized from images; it also loads the corresponding warning weight coefficients from the knowledge base. .

[0040] Step S6: Multivariate time series prediction and composite deviation scoring: Extract multivariate time-series data from the past hour (12 sampling points, one every 5 minutes), input them into the LSTM / TCN prediction model; output the predicted values ​​of key parameters (ammonia nitrogen, dissolved oxygen) for the next 30 minutes; calculate the current deviation score for each. and future deviation score .

[0041] Step S7: Composite Early Warning Judgment and Root Cause Analysis: according to and The system uses the value and preset threshold to determine whether an alert is triggered and the alert level. If an alert is triggered, it calls the model attention weight or SHAP value, outputs the top three parameters with the highest contribution as the main causes, and generates an alert message with root cause hints and handling suggestions.

[0042] Step S8: Multi-terminal collaborative task scheduling: Based on the abnormal location, required skill tags, and the location and load of maintenance personnel in the early warning information, calculate the task suitability and assign tasks; track task progress in real time and handle timeouts and conflicts.

[0043] Step S9: 3D Visualization Navigation and On-site Processing: The system plans and navigates the optimal 3D path to the abnormal cylinder for the maintenance personnel receiving the task; it handles the abnormality on-site, records the handling process and results through the terminal, and uploads photos / videos before and after the handling.

[0044] Step S10: Task closure and data storage: The system verifies whether the anomaly has been resolved; if resolved, the task is closed and the BIM model status is updated; if not resolved, the warning level is upgraded and the task is reassigned; all data (raw monitoring data, prediction results, warning records, scheduling records, and processing records) are stored in the database for subsequent source analysis and model iteration optimization.

[0045] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A zebrafish breeding intelligent monitoring and early warning system based on BIM and multi-terminal collaboration, characterized in that, It includes a multi-parameter data acquisition module, a BIM 3D modeling and spatial mapping module, a multi-sensor spatiotemporal data fusion module, a growth stage adaptive early warning module, a multi-terminal collaborative intelligent scheduling module, a 3D visualization operation and maintenance module, and a historical data and traceability module; The multi-parameter data acquisition module is used to collect water quality parameters, environmental parameters, equipment status parameters, and image and video data of the aquaculture tank in real time. The BIM 3D modeling and spatial mapping module is used to construct a high-precision BIM model of the zebrafish farming rack, establish the spatial topology relationship of the farming tank, sensors, and pipelines, and assign a unique code to each farming tank and sensor to realize the dynamic binding of data with BIM spatial nodes. The multi-sensor spatiotemporal data fusion module is used to perform spatial interpolation fusion of the measured data of the current cylinder block sensor with the data of adjacent spatial nodes in the BIM model to generate a three-dimensional dynamic parameter field, and to perform self-verification and virtual sensing compensation on the sensor data. The growth stage adaptive early warning module is used to dynamically adjust the weight coefficients of the early warning scoring model according to the current growth stage of the zebrafish, and output the predicted values ​​of key parameters for future periods based on the multivariate time series prediction model. The combination of the current parameter values ​​and the predicted values ​​of future parameters realizes a dual-scale composite early warning. The multi-terminal collaborative intelligent scheduling module is used to achieve intelligent task distribution, progress synchronization and conflict resolution through edge computing nodes based on early warning information and fault root cause analysis results, combined with the location, skill tags and current task load of maintenance personnel; The three-dimensional visualization operation and maintenance module is used to integrate and display real-time monitoring data, parameter field interpolation results, early warning information and BIM model, and provide a three-dimensional navigation path to the abnormal cylinder block; The historical data and traceability module is used to store all monitoring data, operation records and abnormal events, and supports traceability analysis and experimental data correlation mining.

2. The intelligent monitoring and early warning system for zebrafish farming based on BIM and multi-terminal collaboration as described in claim 1, characterized in that, The BIM 3D modeling and spatial mapping module is used for: Extract the waterway topology from the BIM model, including the connection sequence of each cylinder, the direction of water flow, and the pipe diameter, and construct a directed graph structure representing the transmission relationship from the water source to the upstream cylinder and then to the downstream cylinder. Establish a four-dimensional mapping relationship between equipment codes, spatial locations, data interfaces, and waterway topology adjacency relationships.

3. The intelligent monitoring and early warning system for zebrafish farming based on BIM and multi-terminal collaboration as described in claim 1, characterized in that, The multi-sensor spatiotemporal data fusion module is specifically used for: Spatial domain fusion: Based on the spatial coordinates and waterway topology provided by the BIM model, Kriging interpolation or inverse distance weighted interpolation methods are used to calculate the parameter estimates of the current cylinder or nodes without installed sensors based on the measured data of adjacent cylinders and upstream and downstream cylinders, and to construct a three-dimensional dynamic parameter field. Sensor self-calibration: The measured value of the current cylinder is compared with the interpolated estimated value of the upstream and downstream. If the deviation exceeds the preset threshold, it is determined that the current sensor is drifting or malfunctioning, and a sensor calibration warning is triggered. Time-domain fusion: Perform moving average filtering on multiple time-point data of the same cylinder block and the same parameter to remove peak noise; Multi-source alignment: Align water quality parameters, environmental parameters, and image recognition feature values ​​with the same timestamp to form a sample vector for input to the prediction model.

4. The intelligent monitoring and early warning system for zebrafish farming based on BIM and multi-terminal collaboration as described in claim 1, characterized in that, The growth stage adaptive early warning module includes: Growth Stage Identification and Weight Adaptive Unit: Establish a zebrafish growth stage-early warning weight mapping knowledge base that includes five stages: embryonic stage, larval stage, juvenile stage, adult stage, and reproductive stage. By recording the introduction date of zebrafish in each tank and combining it with the average pixel area of ​​the fish body output by the image recognition module, the current growth stage is automatically determined, and the corresponding weight coefficient is automatically loaded according to the identified stage. Multivariate time series prediction unit: Construct and pre-train a multivariate time series prediction model based on Long Short-Term Memory Network (LSTM) or Temporal Convolutional Network (TCN). The model input is multi-parameter time series data including water temperature, pH, dissolved oxygen, conductivity, ammonia nitrogen, nitrite, light intensity, and feeding amount within a past time window. The output is the predicted values ​​of key parameters for the next 15 to 60 minutes. Composite deviation score calculation unit: calculates the current deviation score separately. and future deviation score The calculation formula is: in Rate the health of the equipment. For the parameter abnormal volatility, This represents the historical average abnormal volatility. This is an image recognition anomaly index. These are the weighting coefficients; Hierarchical composite early warning and root cause analysis unit: based on and The values ​​and preset thresholds trigger different levels of warnings, and the attention mechanism weights or SHAP values ​​of the prediction model are called to analyze the contribution of each input parameter to the prediction results, output the main causal parameters, and generate warning information with root cause hints and handling suggestions.

5. The intelligent monitoring and early warning system for zebrafish farming based on BIM and multi-terminal collaboration as described in claim 1, characterized in that, The warning levels set by the hierarchical composite early warning and root cause analysis unit include: Early warning: ; General warning: ; Important Warning: ; Emergency Warning: ; The advance warning is used to issue an early warning when the predicted value of a parameter is about to exceed the standard in a future period, while the current parameter is still within the normal range.

6. The intelligent monitoring and early warning system for zebrafish farming based on BIM and multi-terminal collaboration as described in claim 1, characterized in that, The multi-terminal collaborative intelligent scheduling module is specifically used for: After receiving the early warning information, extract the abnormal location, abnormal type, root cause analysis results, and required skill tags; Query the list of online maintenance personnel terminals to obtain the current location, skill tags, and current task load of each terminal; A weighted matching algorithm is used to calculate the task suitability of each operations and maintenance personnel: Select the terminal with the highest compatibility to push the task; A weighted matching algorithm is used to calculate the task suitability of each operations and maintenance personnel: Select the terminal push task with the highest compatibility; If no response is received from the terminal within the preset time, the task will be automatically pushed to the terminal with the second highest compatibility. Establish a global synchronization mechanism for operation timestamps and processing progress. When multiple terminals are detected responding to the same task simultaneously, retain the terminal that responded earliest, cancel the task of the remaining terminals, and send a conflict notification.

7. The intelligent monitoring and early warning system for zebrafish farming based on BIM and multi-terminal collaboration as described in claim 1, characterized in that, The 3D visualization operation and maintenance module is specifically used for: Develop a 3D visualization interface based on Unity or Three.js engine, and mark the operating status of each aquaculture tank on the BIM model with different colors. The colors include green for normal, blue for early warning, yellow for general warning, orange for important warning, and red for emergency warning. Supports overlay display of parameter field thermal maps, intuitively showing the spatial distribution and diffusion trend of temperature field, dissolved oxygen field, and ammonia nitrogen concentration field on the breeding rack; Clicking on any aquaculture tank will bring up a detailed information panel, displaying real-time parameters, future time period prediction curves, current growth stage, current weight coefficient, historical warning records, and maintenance logs; After an abnormal task is dispatched, the system automatically plans the optimal 3D path from the current location of the maintenance personnel to the abnormal cylinder, highlights the navigation on the BIM model, and synchronizes it to the mobile device.

8. A method for intelligent monitoring and early warning of zebrafish farming based on BIM and multi-terminal collaboration, employing the system described in any one of claims 1 to 7, characterized in that, Includes the following steps: Step S1: Construct a BIM model of the zebrafish farming rack, deploy sensors, cameras and edge computing nodes, and configure the early warning weight coefficients and maintenance personnel information for each growth stage; Step S2: Real-time collection of water quality parameters, environmental parameters, equipment status parameters, and image and video data for each aquaculture tank; Step S3: Preprocess and align the collected data spatiotemporally to form a sample vector; Step S4: Map the preprocessed data to the corresponding spatial nodes of the BIM model, construct a three-dimensional dynamic parameter field based on the BIM waterway topology using spatial interpolation, and diagnose the sensor status by cross-verifying upstream and downstream sensor data. Step S5: Based on the date the fry were introduced and the average size of the fish in the image, the growth stage of the zebrafish in the tank is automatically determined, and the corresponding warning weight coefficient is loaded from the knowledge base; Step S6: Extract multivariate time series data from a past time window, input it into the multivariate time series prediction model, output the predicted values ​​of key parameters for future periods, and calculate the current deviation score and the future deviation score respectively; Step S7: Determine whether to trigger an alert and the alert level based on the current deviation score, the future deviation score and the preset threshold. If an alert is triggered, output the main cause parameters and generate an alert message with root cause prompts and handling suggestions. Step S8: Assign tasks based on the abnormal location, required skill tags, location of maintenance personnel, and load calculation task suitability in the early warning information, and track task progress in real time; Step S9: Plan the optimal three-dimensional path to the abnormal cylinder for the maintenance personnel receiving the task and perform navigation. After handling the abnormality on-site, record the processing process and results through the terminal. Step S10: Verify whether the anomaly has been resolved. If it has, close the task and update the BIM model status. If it has not been resolved, upgrade the warning level and reassign the task. Store all data in the database for source analysis and model iteration optimization.

9. A method for intelligent monitoring and early warning of zebrafish farming based on BIM and multi-terminal collaboration, characterized in that, In step S6, the multivariate temporal prediction model uses a Long Short-Term Memory (LSTM) network or a Temporal Convolutional Network (TCN). The shape of the model input tensor is: ,in Take 12 to 24 as the corresponding timeframe for the past 1 to 2 hours. It includes 8 to 12 relevant variables such as water temperature, pH, dissolved oxygen, conductivity, ammonia nitrogen, nitrite, light intensity, and feeding amount. The model output is the predicted values ​​of ammonia nitrogen concentration and dissolved oxygen concentration for the next 15 to 60 minutes. The current deviation score and the future deviation score are calculated according to the following formula: Where H represents the equipment health score. For the parameter abnormal volatility, This represents the historical average abnormal volatility. This is an image recognition anomaly index. The weighting coefficients are dynamically applied based on the growth stage.

Images