Real-time visual monitoring system for operation data of centrifugal machine

Abstract

The invention relates to the technical field of industrial automation and intelligent control, in particular to a centrifuge operation data real-time visualization monitoring system which comprises a distributed data acquisition unit, a data analysis and processing unit, a real-time monitoring and visualization unit, a decentralized control system unit and a remote access and support unit. According to the method, on-line incremental learning and hybrid modeling are fused with a physical mechanism and data-driven analysis, so that dynamic prediction, anomaly recognition and causal reasoning of the running state of the centrifugal machine are realized, and lightweight deployment of the edge side is supported; based on the multi-physics coupling real-time simulation and augmented reality technology, multi-phase flow state high-precision dynamic visualization and operation risk intelligent space early warning in the centrifugal process are achieved, and multi-parameter dynamic optimization regulation and control in the centrifugal process are achieved by means of a fuzzy self-adaptive control strategy and a distributed intelligent cooperation mechanism; and the intelligent analysis, real-time monitoring and accurate control capability and the complex working condition adaptability of the system to the operation of the centrifugal machine are obviously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation and intelligent control, and in particular to a real-time visual monitoring system for centrifuge operation data. Background Art

[0002] As a core device for separating mixtures, centrifuges have been upgraded from traditional centrifugal separation to high-precision biopreparation in the pharmaceutical and biological fields. For example, the STC series of fully automatic tubular centrifuges achieve nanoscale material separation through ultra-high-speed rotation at 100,000 rpm and are widely used in the production of biological products such as mRNA vaccines and monoclonal antibodies. However, as bioprocesses place increasing demands on separation accuracy and safety, existing technologies have limited the ability to perform real-time analysis, low dynamic visualization accuracy, poor adaptability to parameter control, and low efficiency in multi-device collaboration.

[0003] Based on this, the present invention provides a real-time visual monitoring system for centrifuge operation data to solve the above-mentioned technical problems. Summary of the Invention

[0004] The purpose of the present invention is to provide a real-time visualization monitoring system for centrifuge operation data. The present invention integrates online incremental learning and hybrid modeling with physical mechanisms and data-driven analysis to achieve dynamic prediction, anomaly identification and causal reasoning of the centrifuge operation status and support lightweight deployment on the edge side. Based on multi-physical field coupling real-time simulation and augmented reality technology, high-precision dynamic visualization of multiphase flow in the centrifugal process and intelligent spatial early warning of operational risks are achieved. With the help of fuzzy adaptive control strategy and distributed intelligent collaborative mechanism, dynamic optimization and regulation of multiple parameters of the centrifugal process are realized, which significantly improves the system's intelligent analysis, real-time monitoring, precise control capabilities of centrifuge operation and adaptability to complex working conditions.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The present invention provides a real-time visual monitoring system for centrifuge operation data, comprising a distributed data acquisition unit, a data analysis and processing unit, a real-time monitoring and visualization unit, a distributed control system unit, and a remote access and support unit, wherein:

[0007] The distributed data acquisition unit: based on the edge computing architecture, collects multi-dimensional operating parameters such as temperature, speed, vibration acceleration, current load, pressure gradient, material flow and separation efficiency in real time through multiple types of sensors;

[0008] The data analysis and processing unit is used to dynamically predict the centrifuge's operating status, identify anomalies, and perform causal reasoning through online incremental learning and hybrid modeling methods, integrating physical mechanisms with data-driven analysis, and supporting lightweight deployment on the edge.

[0009] The real-time monitoring and visualization unit: Based on multi-physics field coupling real-time simulation and augmented reality technology, it can realize high-precision dynamic visualization of multiphase flow in the centrifugal process and intelligent spatial warning of operation risks;

[0010] The distributed control system unit: based on fuzzy adaptive control strategy and distributed intelligent coordination mechanism, performs dynamic optimization and control of multiple parameters of the centrifugal process;

[0011] The remote access and support unit is used to adopt the HTTPS encrypted communication protocol to build a cloud-edge collaborative architecture, supporting technicians to perform remote system configuration, parameter adjustment and firmware upgrades through Web terminals or mobile applications.

[0012] The distributed data acquisition unit includes an edge computing node module, a multi-type sensor interface module, a data synchronization acquisition and timestamp marking module, and a communication transmission module, wherein:

[0013] The edge computing node module is used to perform local data preprocessing and real-time response decision-making;

[0014] The multi-type sensor interface module is used to connect to multiple types of sensors such as temperature, speed, and vibration;

[0015] The data synchronization acquisition and timestamp marking module is used to achieve μs-level clock synchronization using the PTPv2 protocol, and to synchronize the high-precision acquisition and time alignment of multi-channel data;

[0016] The communication transmission module is used to upload the collected data to the analysis platform via an industrial bus or a wireless network.

[0017] The multi-type sensors include a fiber grating vibration sensor deployed on the surface of the centrifuge rotor, a Hall current sensor at the motor end, and a Coriolis mass flowmeter on the material pipeline.

[0018] The data analysis and processing unit includes an online incremental learning module, a hybrid modeling module, a causal reasoning engine, and a lightweight deployment module, wherein:

[0019] The online incremental learning module: performs real-time anomaly detection on the time series data of vibration acceleration based on a streaming decision tree algorithm;

[0020] The hybrid modeling module is used to integrate the centrifugal force field theoretical model and the LSTM neural network to establish a nonlinear mapping relationship between material separation efficiency and process parameters, supporting dynamic process optimization;

[0021] The causal reasoning engine: based on the Bayesian network algorithm, analyzes the causal relationship between multiple parameters;

[0022] The lightweight deployment module is used to compress the trained AI model to edge devices using the TensorFlow Lite framework for operation.

[0023] The online incremental learning module performs real-time anomaly detection on vibration acceleration time series data based on a streaming decision tree algorithm. The specific operations are as follows:

[0024] A1: Normalize and segment the received vibration acceleration signal using a sliding window;

[0025] A2: Uses the Hoeffding Tree structure and updates the node splitting strategy online;

[0026] A3: Construct a comprehensive anomaly index based on the change in information gain ratio and the residual sum of squares;

[0027] A4: Dynamically adjust model parameters through sliding window strategy and weight decay function;

[0028] Among them, the abnormal index A t The specific calculation formula at time step t is:

[0029]

[0030] Where, IGR(X|S t ) is the current sample set S t The information gain ratio of the feature X; w is the sliding window length; r i is the prediction residual of the i-th step; α∈[0,1] is the balance factor, which is used to adjust the trade-off between model stability and sensitivity;

[0031] The streaming decision tree model uses the Hoeffding Tree criterion to determine the timing of node splitting, and its judgment conditions are as follows:

[0032]

[0033] Where ΔG is the gain difference between the suboptimal split attribute and the optimal attribute; R is the value range of information gain; n is the number of samples under the current node; δ is the confidence threshold, which is used to control the robustness of the split.

[0034] The hybrid modeling module is used to integrate the centrifugal force field theoretical model and the LSTM neural network to establish a nonlinear mapping relationship between material separation efficiency and process parameters, and support dynamic process optimization. The specific operations are as follows:

[0035] B1: Calculate the ideal separation efficiency η based on the centrifugal force field dynamics equation theory :

[0036]

[0037] Where N is the rotation speed; ρ P ,ρ f is the particle / fluid density; d P is the particle size; L is the length of the centrifugal chamber; μ is the fluid dynamic viscosity; r o / r i is the ratio of the outer diameter to the inner diameter of the drum;

[0038] B2: Use Attention-LSTM network to process real-time process parameters:

[0039] h t =LSTM(T t ,μ t ,Q t ,P t-1 ;h t-1 )

[0040] α t =softmax(W a [h t ;η theory ])

[0041] η pred =W o [α t ⊙h t ]+b o

[0042] Where, T t ,μ t ,Q t ,P t is the real-time temperature, viscosity, flow rate, and pressure; η pred is the output efficiency prediction value;

[0043] B3: Integrate theoretical value and predicted value through DS evidence theory:

[0044]

[0045] B4: Solve the optimal speed N based on the gradient descent method opt :

[0046]

[0047] Where γ = 0.01.

[0048] The real-time monitoring and visualization unit includes a multi-physics field simulation engine, an AR augmented reality module, a risk warning visualization module, and a historical data backtracking module, wherein:

[0049] The multi-physics field simulation engine: Based on CFD and FEA technology, it simulates the flow field distribution, temperature gradient and rotor stress changes in the centrifugal chamber in real time to generate a three-dimensional dynamic simulation model;

[0050] The AR augmented reality module is used to superimpose equipment parameters, fault locations, and maintenance instructions in the operator's field of view through smart glasses and spatial positioning technology, supporting contactless remote collaboration;

[0051] The risk warning visualization module is used to present the equipment status in the form of heat maps and three-dimensional vector maps through WebGL technology. Abnormal parameters automatically trigger red, yellow and green warnings and mark the spatial location;

[0052] The historical data backtracking module: based on the time series database, supports multi-dimensional data query and generates traceable audit reports.

[0053] The distributed control system unit includes a fuzzy adaptive control module, a multi-agent collaboration module, a dynamic parameter optimization module, and a safety interlock control module, wherein:

[0054] The fuzzy adaptive control module is used to dynamically adjust the PID parameters according to the material characteristic parameters;

[0055] The multi-agent collaboration module: Based on the blockchain consensus mechanism, it performs task scheduling, load balancing and fault redundancy among multiple centrifuges, supporting clustered production;

[0056] The dynamic parameter optimization module is used to integrate the genetic algorithm and the expert knowledge base to automatically optimize the centrifugal speed and temperature curve, thereby improving separation efficiency and reducing energy consumption;

[0057] The safety interlock control module: based on the fail-safe principle, automatically triggers emergency shutdown and valve closing protection actions when parameters exceed the limit.

[0058] The multi-agent collaboration module: Based on the blockchain consensus mechanism, it performs task scheduling, load balancing and fault redundancy among multiple centrifuges, and supports clustered production. The specific operations are as follows:

[0059] C1: Intelligent agent identity management: Each centrifuge acts as an independent intelligent agent and generates a public and private key pair through the elliptic curve digital signature algorithm: Signature = Sign SK (m), Verify PK (signature, m), where SK is the private key; PK is the public key; m is the message;

[0060] C2: Task allocation mechanism: Using the improved contract network protocol, automatic task allocation is achieved based on blockchain smart contracts: Where: C iis the current load of agent i; C max is the maximum load threshold; E i is the current energy consumption; E i,avg is the historical average energy consumption; P i is the processing accuracy; α, β, γ are weight coefficients;

[0061] C3: Load balancing algorithm: Based on the dynamic weighted Byzantine fault-tolerant consensus mechanism, each agent calculates its weight using the following formula: Where: T i is the temperature; T max -T min is the temperature threshold range; R i is the reliability score; ω1, ω2, ω3 are weight coefficients;

[0062] C4: Fault detection and redundancy processing: Fault detection using majority voting mechanism: When the failure probability exceeds the threshold of 0.67, the smart contract is triggered to perform task migration: migration target = argmax j (W j Compatibility j,i ), where: Compatibility j,i Score the task compatibility between agent j and agent i;

[0063] C5: Consensus reaching process: An improved consensus protocol based on PBFT, including the following steps:

[0064] ①Preparation phase: the master node broadcasts the preparatory message<PRE-PREPARE,v,n,d,m> ;

[0065] ② Preparation phase: Slave node verifies and broadcasts preparation message<PREPARE,v,n,d,i> ;

[0066] ③ Commit phase: After receiving 2f+1 prepare messages, broadcast the commit message<COMMIT,v,n,d,i> ;

[0067] ④Execution phase: After receiving 2f+1 submission messages, execute the task and update the status;

[0068] Where: v is the view number; n is the sequence number; d is the message digest; i is the node ID; f is the number of faulty nodes that can be tolerated;

[0069] C6: Blockchain data structure: Using improved Merkle tree to store task execution records: H i,j =Hash(H i,2j ||H i,2j+1 ), where: H i,jis the hash value of the jth node in the i-th layer; || is the connection operation.

[0070] The remote access and support unit includes a secure communication protocol module, a cloud-edge collaborative architecture module, a Yuanda configuration and configuration module, and a firmware remote upgrade module, wherein:

[0071] The secure communication protocol module is used to use HTTPS and SSL encryption protocols to ensure remote connection security;

[0072] The cloud-edge collaboration architecture module is used for efficient collaboration between cloud management and edge control;

[0073] The Yuanda configuration and configuration module is used by technical personnel to configure system parameters and adjust policies through the terminal;

[0074] The firmware remote upgrade module is used for remote downloading and automatic updating of device programs.

[0075] Compared with the prior art, the present invention has the following beneficial effects:

[0076] The present invention integrates physical mechanisms and data-driven analysis through online incremental learning and hybrid modeling to achieve dynamic prediction, anomaly identification and causal reasoning of the centrifuge's operating status and support lightweight deployment on the edge. It also realizes high-precision dynamic visualization of multiphase flow in the centrifugal process and intelligent spatial warning of operational risks based on multi-physical field coupling real-time simulation and augmented reality technology. It uses fuzzy adaptive control strategy and distributed intelligent collaborative mechanism to realize dynamic optimization and regulation of multiple parameters of the centrifugal process, significantly improving the system's intelligent analysis, real-time monitoring, precise control capabilities of centrifuge operation and adaptability to complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 This is a system diagram of a real-time visual monitoring system for centrifuge operation data according to the present invention.

[0078] Figure 2 This is a hybrid modeling workflow diagram for a real-time visualization monitoring system for centrifuge operation data according to the present invention.

[0079] Figure 3 This is a multi-physical field simulation flow chart of a real-time visual monitoring system for centrifuge operation data according to the present invention.

[0080] Description of Figure Numbers:

[0081] 100. Distributed data acquisition unit; 101. Edge computing node module; 102. Multi-type sensor interface module; 103. Data synchronization acquisition and timestamp marking module; 104. Communication transmission module; 200. Data analysis and processing unit; 201. Online incremental learning module; 202. Hybrid modeling module; 203. Causal reasoning engine; 204. Lightweight deployment module; 300. Real-time monitoring and visualization unit; 301. Multi-physics field simulation engine; 302. AR augmented reality module; 303. Risk warning visualization module; 304. Historical data backtracking module; 400. Distributed control system unit; 401. Fuzzy adaptive control module; 402. Multi-agent collaboration module; 403. Dynamic parameter optimization module; 404. Safety interlock control module; 500. Remote access and support unit; 501. Security communication protocol module; 502. Cloud-edge collaborative architecture module; 503. Yuanda configuration and configuration module; 504. Firmware remote upgrade module. DETAILED DESCRIPTION

[0082] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0083] Example:

[0084] like Figure 1-Figure 3As shown, this embodiment provides a real-time visualization monitoring system for centrifuge operation data, including a distributed data acquisition unit 100, a data analysis and processing unit 200, a real-time monitoring and visualization unit 300, a distributed control system unit 400, and a remote access and support unit 500, wherein: the distributed data acquisition unit 100: based on the edge computing architecture, collects multi-dimensional operating parameters such as temperature, speed, vibration acceleration, current load, pressure gradient, material flow rate and separation efficiency in real time through multiple types of sensors; the data analysis and processing unit 200: is used to integrate physical mechanism and data-driven analysis through online incremental learning and hybrid modeling methods to monitor centrifuge operation. Dynamic prediction of operation status, anomaly identification and causal reasoning, and support for lightweight deployment on the edge side; Real-time monitoring and visualization unit 300: Based on multi-physics field coupling real-time simulation and augmented reality technology, high-precision dynamic visualization of multiphase flow in the centrifugal process and intelligent spatial warning of operation risks; Distributed control system unit 400: Based on fuzzy adaptive control strategy and distributed intelligent collaborative mechanism, dynamic optimization and regulation of multiple parameters of the centrifugal process; Remote access and support unit 500: Used to adopt HTTPS encrypted communication protocol to build a cloud-edge collaborative architecture, supporting technicians to perform remote system configuration, parameter adjustment and firmware upgrades through Web terminals or mobile applications.

[0085] Among them, it should be noted that the multi-dimensional operating parameters obtained by the distributed data acquisition unit 100 are subjected to hybrid modeling and decision generation by the data analysis and processing unit 200, and the optimization parameters output by the data analysis and processing unit 200 are transmitted to the distributed control system unit 400 through the OPCUA protocol to perform dynamic regulation. The transmission delay is <10ms, and the real-time monitoring and visualization unit 300 synchronously presents the multi-physical field status and risk warning. The remote access and support unit 500 provides secure interconnection and remote operation and maintenance support throughout the entire process.

[0086] In this embodiment, it should also be noted that the distributed data acquisition unit 100 includes an edge computing node module 101, a multi-type sensor interface module 102, a data synchronization acquisition and timestamp module 103, and a communication transmission module 104. The edge computing node module 101 is used to perform local data preprocessing and real-time response decisions. The multi-type sensor interface module 102 is used to connect to multiple sensors for temperature, speed, and vibration. These sensors include fiber Bragg grating vibration sensors deployed on the surface of the centrifuge rotor, Hall effect current sensors at the motor end, and Coriolis mass flow meters on the material pipeline. The data synchronization acquisition and timestamp module 103 uses the PTPv2 protocol to achieve μs-level clock synchronization, enabling high-precision synchronous acquisition and time alignment of multi-channel data. The communication transmission module 104 is used to upload the collected data to the analysis platform via an industrial bus or wireless network.

[0087] It should be noted that the edge computing node module 101 performs local preprocessing on the heterogeneous sensor data accessed by the multi-type sensor interface module 102. After the data synchronization acquisition and timestamp marking module 103 completes the μs-level time alignment, it is uploaded to the cloud analysis platform by the communication transmission module 104 through a wired / wireless hybrid channel.

[0088] Furthermore, it should be noted that the PTPv2 protocol achieves a synchronization accuracy of ±100ns in a 1Gbps industrial Ethernet environment through hardware timestamps and the best master clock algorithm. Specifically, it includes: Clock hierarchy calculation: I. Determine the Grandmaster Clock based on clock accuracy and stability; II. Delay measurement: Use the Pee Delay mechanism to compensate for link asymmetry; III. Clock correction: Use the Sync / Follow Up Phase calibration is performed on the message pairs.

[0089] In this embodiment, it should also be noted that the data analysis and processing unit 200 includes an online incremental learning module 201, a hybrid modeling module 202, a causal reasoning engine 203, and a lightweight deployment module 204, wherein: the online incremental learning module 201: performs real-time anomaly detection on the time series data of vibration acceleration based on the streaming decision tree algorithm; the specific operations are as follows: A1: normalizes and segments the received vibration acceleration signal into sliding windows; A2: adopts the HoeffdingTree structure to update the split node strategy online; A3: constructs a comprehensive anomaly index based on the change of information gain ratio and the residual square sum; A4: dynamically adjusts the model parameters through the sliding window strategy and the weight attenuation function; wherein, the anomaly index A t The specific calculation formula at time step t is:

[0090]

[0091] Where, IGR(X|S t ) is the current sample set S t The information gain ratio of the feature X; w is the sliding window length; r i is the prediction residual of the i-th step; α∈[0,1] is a balancing factor used to adjust the trade-off between model stability and sensitivity; the streaming decision tree model uses the Hoeffding Tree criterion to determine the timing of node splitting, and its judgment conditions are as follows:

[0092]

[0093] Where ΔG is the gain difference between the suboptimal splitting attribute and the optimal attribute; R is the value range of information gain; n is the number of samples under the current node; δ is the confidence threshold used to control the robustness of the split. Hybrid modeling module 202: used to integrate the centrifugal force field theoretical model and the LSTM neural network to establish a nonlinear mapping relationship between material separation efficiency and process parameters, supporting dynamic process optimization; the specific operations are as follows: B1: Calculate the ideal separation efficiency η based on the centrifugal force field dynamics equation theory :

[0094]

[0095] Where N is the rotation speed; ρ P ,ρ f is the particle / fluid density; d P is the particle size; L is the length of the centrifugal chamber; μ is the fluid dynamic viscosity; r o / r i is the ratio of the outer diameter to the inner diameter of the drum; B2: Attention-LSTM network is used to process real-time process parameters:

[0096] h t =LSTM(T t ,μ t ,Q t ,P t-1 ;h t-1 )

[0097] α t =softmax(W a [h t ;η theory ])

[0098] η pred =W o [α t ⊙h t ]+b o

[0099] Where, T t ,μ t ,Q t ,P t is the real-time temperature, viscosity, flow rate, and pressure; η pred Output efficiency prediction value; B3: Integrate theoretical value and prediction value through DS evidence theory:

[0100]

[0101] B4: Solve the optimal speed N based on the gradient descent method opt : Where γ = 0.01. Causal inference engine 203: Analyzes causal relationships between multiple parameters based on a Bayesian network algorithm. Lightweight deployment module 204: Uses the TensorFlow Lite framework to compress trained AI models and run them on edge devices.

[0102] Among them, it should be noted that the online incremental learning module 201 performs real-time anomaly detection on vibration signals and outputs warning indicators, the hybrid modeling module 202 integrates physical laws and data-driven models to generate process optimization suggestions, the causal reasoning engine 203 analyzes the causal relationship between multiple parameters to locate the root cause of the fault, and the lightweight deployment module 204 compresses and transplants the optimized model to the edge.

[0103] Furthermore, it should be noted that the hybrid modeling is adapted to the scenario: in the biopharmaceutical scenario, the hybrid modeling module 202 learns the rheological properties of the protein solution through LSTM and dynamically adjusts the correction coefficient of the centrifugal force field theoretical model.

[0104] In this embodiment, it should also be noted that the real-time monitoring and visualization unit 300 includes a multi-physics field simulation engine 301, an AR augmented reality module 302, a risk warning visualization module 303, and a historical data backtracking module 304, wherein: the multi-physics field simulation engine 301: based on CFD and FEA technology, simulates the flow field distribution, temperature gradient and rotor stress changes in the centrifugal cavity in real time to generate a three-dimensional dynamic simulation model; the AR augmented reality module 302: is used to superimpose equipment parameters, fault points and maintenance guides in the operator's field of view through smart glasses and spatial positioning technology, supporting contactless remote collaboration; the risk warning visualization module 303: is used to present the equipment status in the form of heat maps and three-dimensional vector maps through WebGL technology, and abnormal parameters automatically trigger red, yellow and green color warnings and mark the spatial location; the historical data backtracking module 304: based on the time series database, supports multi-dimensional data query and generates traceable audit reports.

[0105] Among them, it should be noted that the real-time dynamic model generated by the multi-physics field simulation engine 301 provides a three-dimensional data base for the AR augmented reality module 302, the risk warning visualization module 303 integrates the simulation results with the real-time data and converts them into graded alarm signals, and the historical data backtracking module 304 synchronously archives the complete process data to form a closed-loop verification chain.

[0106] Furthermore, it should be noted that the parameter mapping of multi-physics field simulation: CFD simulation is based on the ANSYS Fluent solver, which divides the centrifugal chamber into more than 2 million grids, and calculates the flow field characteristic parameters such as the Reynolds number and the Nusselt number in real time; FEA analysis uses ABAQUS to solve the rotor stress, taking into account material nonlinearity (such as the plastic deformation of 316L stainless steel) and contact nonlinearity (the clearance between the rotor and the bearing), and the deviation between the simulation results and the measured stress is less than 5%. AR visualization interaction mechanism: The operator can use the gesture recognition function of the smart glasses to call up the flow field slice diagram of any cross section of the centrifuge in real time. For example, when viewing the nanocarbon separation process, the gesture sliding can switch to display the particle concentration cloud map at different speeds. The red area automatically marks the agglomeration risk point and displays the optimal speed adjustment suggestion (such as reducing from 10,000 RPM to 8,500 RPM).

[0107] In this embodiment, it should also be noted that the distributed control system unit 400 includes a fuzzy adaptive control module 401, a multi-agent collaboration module 402, a dynamic parameter optimization module 403, and a safety interlock control module 404, wherein: the fuzzy adaptive control module 401 is used to dynamically adjust the PID parameters according to the material characteristic parameters of protein thermal stability and cell shear sensitivity; the multi-agent collaboration module 402 is based on the blockchain consensus mechanism to perform task scheduling, load balancing and fault redundancy among multiple centrifuges to support clustered production; the specific operations are as follows: C1: Agent identity management: Each centrifuge is an independent agent, and generates a public-private key pair through the elliptic curve digital signature algorithm: Signature = Sign SK (m), Verify PK (signature, m), where: SK is the private key; PK is the public key; m is the message; C2: Task allocation mechanism: Using the improved contract network protocol, automatic task allocation is achieved based on blockchain smart contracts: Where: C i is the current load of agent i; C max is the maximum load threshold; E i is the current energy consumption; E i,avg is the historical average energy consumption; P i is the processing accuracy; α, β, γ are weight coefficients; C3: Load balancing algorithm: Based on the Byzantine fault-tolerant consensus mechanism with dynamic weights, each agent calculates its weight using the following formula: Where: T i is the temperature; T max -T min is the temperature threshold range; R i is the reliability score; ω1, ω2, ω3 are weight coefficients; C4: Fault detection and redundancy processing: majority voting mechanism is used to detect faults: When the failure probability exceeds the threshold of 0.67, the smart contract is triggered to perform task migration: migration target = argmax j (W j Compatibility j,i ), where: Compatibility j,i Score the task compatibility of agent j and agent i; C5: Consensus reaching process: An improved consensus protocol based on PBFT, which includes the following steps: ① Pre-preparation phase: The master node broadcasts the pre-preparation message<PRE-PREPARE,v,n,d,m> ;② Preparation phase: Slave node verifies and broadcasts preparation message<PREPARE,v,n,d,i> ;③ Commit phase: after receiving 2f+1 prepare messages, broadcast the commit message<COMMIT,v,n,d,i> ;④Execution phase: After receiving 2f+1 commit messages, execute the task and update the status; where: v is the view number; n is the sequence number; d is the message digest; i is the node ID; f is the number of tolerable fault nodes; C6: Blockchain data structure: using the improved Merkle tree to store task execution records: H i,j =Hash(H i,2j ||H i,2j+1 ), where: H i,j is the hash value of the jth node in the i-th layer; || is a join operation. Dynamic parameter optimization module 403 integrates a genetic algorithm with an expert knowledge base to automatically optimize the centrifuge speed and temperature profile, improving separation efficiency and reducing energy consumption. Safety interlock control module 404, based on the fail-safe principle, automatically triggers emergency shutdown and valve closure protection actions when parameters exceed limits.

[0108] Among them, it should be noted that the fuzzy adaptive control module 401 and the dynamic parameter optimization module 403 form a dual-loop adjustment system of "parameter self-tuning-process self-optimization", the multi-agent collaboration module 402 realizes the distributed scheduling and fault-tolerant migration of cluster tasks based on blockchain smart contracts, and the safety interlock control module (404) monitors the parameters of the entire process in real time and implements hierarchical protection.

[0109] Furthermore, it is important to note the construction of the fuzzy adaptive control rule base: For cell culture scenarios, the fuzzy controller establishes a "shear sensitivity-speed adjustment" rule base: When the shear sensitivity index of the CHO cell culture fluid is detected to be greater than 0.8, the speed adjustment gradient is automatically set to 50 RPM / minute (the conventional gradient is 200 RPM / minute) and the temperature compensation coefficient is increased (for every 100 RPM decrease, the temperature increases by 0.5°C) to maintain cell viability above 90%. A blockchain-based evidence storage mechanism for multi-agent collaboration: The task scheduling records of each centrifuge are stored in blocks, containing information such as task ID, allocation time, execution parameters, and efficiency indicators.

[0110] In this embodiment, it should also be noted that the remote access and support unit 500 includes a secure communication protocol module 501, a cloud-edge collaborative architecture module 502, a far-reaching configuration and configuration module 503, and a firmware remote upgrade module 504, wherein: the secure communication protocol module 501: is used to adopt HTTPS and SSL encryption protocols to ensure the security of remote connections; the cloud-edge collaborative architecture module 502: is used for efficient collaboration of cloud management and edge control; the far-reaching configuration and configuration module 503: is used for technicians to configure system parameters and adjust policies through the terminal; the firmware remote upgrade module 504: is used for remote downloading and automatic updating of device programs.

[0111] Among them, it should be noted that the secure communication protocol module 501 establishes an encrypted channel to ensure the security of data transmission, the cloud-edge collaborative architecture module 502 realizes the seamless connection between cloud-based intelligent analysis and edge real-time control, and the Yuanda configuration and configuration module 503 and the firmware remote upgrade module 504 respectively provide "software-defined" parameter adjustment and "over-the-air download" hardware update capabilities.

[0112] Furthermore, it is important to note the load balancing strategy for cloud-edge collaboration: the cloud server is responsible for historical data mining and global optimization (e.g., analyzing the energy consumption data of 100 centrifuges to generate quarterly energy-saving reports), while the edge device handles real-time control (e.g., millisecond-level vibration response). When the cloud load is too high, non-real-time tasks (e.g., monthly performance analysis) are automatically migrated to the edge node to ensure that the response delay of key control instructions is less than 50ms. Differential algorithm for firmware upgrade: The firmware remote upgrade module 504 uses a binary differential algorithm (e.g., bsdiff) to transmit only the code differences before and after the update.

[0113] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these 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 any one or more embodiments or examples.

[0114] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A real-time visual monitoring system for centrifuge operation data, characterized in that: The system comprises a distributed data acquisition unit (100), a data analysis and processing unit (200), a real-time monitoring and visualization unit (300), a distributed control system unit (400), and a remote access and support unit (500), wherein: The distributed data acquisition unit (100) is based on an edge computing architecture and uses multiple types of sensors to collect multi-dimensional operating parameters such as temperature, rotation speed, vibration acceleration, current load, pressure gradient, material flow rate, and separation efficiency in real time; The data analysis and processing unit (200) is used to dynamically predict the centrifuge operating state, identify anomalies and perform causal reasoning by integrating physical mechanisms and data-driven analysis through online incremental learning and hybrid modeling methods, and supports lightweight deployment on the edge side; The real-time monitoring and visualization unit (300) is based on multi-physics field coupling real-time simulation and augmented reality technology to provide high-precision dynamic visualization of the multiphase flow state of the centrifugal process and intelligent spatial warning of operation risks; The distributed control system unit (400) performs dynamic optimization and control of multiple parameters of the centrifugal process based on a fuzzy adaptive control strategy and a distributed intelligent collaborative mechanism; The remote access and support unit (500) is used to adopt the HTTPS encrypted communication protocol to build a cloud-edge collaborative architecture, supporting technicians to perform remote system configuration, parameter adjustment and firmware upgrade through a Web terminal or mobile application.

2. A real-time visual monitoring system for centrifuge operation data according to claim 1, characterized in that: The distributed data acquisition unit (100) comprises an edge computing node module (101), a multi-type sensor interface module (102), a data synchronization acquisition and time stamp marking module (103), and a communication transmission module (104), wherein: The edge computing node module (101) is used to perform local data preprocessing and real-time response decision-making; The multi-type sensor interface module (102) is used to connect to multi-type sensors of temperature, rotation speed, and vibration; The data synchronization acquisition and time stamp marking module (103) is used to achieve μs-level clock synchronization using the PTPv2 protocol, and to synchronize the high-precision synchronization acquisition and time alignment of multi-channel data; The communication transmission module (104) is used to upload the collected data to the analysis platform via an industrial bus or a wireless network.

3. A real-time visual monitoring system for centrifuge operation data according to claim 2, characterized in that: The multi-type sensors include a fiber grating vibration sensor deployed on the surface of the centrifuge rotor, a Hall current sensor at the motor end, and a Coriolis mass flowmeter on the material pipeline.

4. A real-time visual monitoring system for centrifuge operation data according to claim 1, characterized in that: The data analysis and processing unit (200) includes an online incremental learning module (201), a hybrid modeling module (202), a causal reasoning engine (203), and a lightweight deployment module (204), wherein: The online incremental learning module (201) performs real-time anomaly detection on the time series data of vibration acceleration based on a streaming decision tree algorithm; The hybrid modeling module (202) is used to integrate the centrifugal force field theoretical model and the LSTM neural network to establish a nonlinear mapping relationship between material separation efficiency and process parameters, and support dynamic process optimization; The causal reasoning engine (203) analyzes the causal relationship between multiple parameters based on the Bayesian network algorithm; The lightweight deployment module (204) is used to compress the trained AI model to the edge device for operation using the TensorFlow Lite framework.

5. A real-time visual monitoring system for centrifuge operation data according to claim 4, characterized in that: The online incremental learning module (201) performs real-time anomaly detection on the time series data of vibration acceleration based on a streaming decision tree algorithm, and the specific operations are as follows: A1: Normalize and segment the received vibration acceleration signal using a sliding window; A2: Uses the Hoeffding Tree structure and updates the node splitting strategy online; A3: Construct a comprehensive anomaly index based on the change in information gain ratio and the residual sum of squares; A4: Dynamically adjust model parameters through sliding window strategy and weight decay function; Among them, the abnormal index A t The specific calculation formula at time step t is: Where, IGR(X|S t ) is the current sample set S t The information gain ratio of the feature X; w is the sliding window length; r i is the prediction residual of the i-th step; α∈[0,1] is the balance factor, which is used to adjust the trade-off between model stability and sensitivity; The streaming decision tree model uses the Hoeffding Tree criterion to determine the timing of node splitting, and its judgment conditions are as follows: ΔG≤∈, and Where ΔG is the gain difference between the suboptimal split attribute and the optimal attribute; R is the value range of information gain; n is the number of samples under the current node; δ is the confidence threshold, which is used to control the robustness of the split.

6. A real-time visual monitoring system for centrifuge operation data according to claim 4, characterized in that: The hybrid modeling module (202) is used to integrate the centrifugal force field theoretical model and the LSTM neural network, establish a nonlinear mapping relationship between material separation efficiency and process parameters, and support dynamic process optimization. The specific operations are as follows: B1: Calculate the ideal separation efficiency η based on the centrifugal force field dynamics equation theory : Where N is the speed; ρ P ,ρ f is the particle / fluid density; d P is the particle size; L is the length of the centrifugal chamber; μ is the fluid dynamic viscosity; r o / r i is the ratio of the outer diameter to the inner diameter of the drum; B2: Use Attention-LSTM network to process real-time process parameters: h t =LSTM(T t ,μ t ,Q t ,P t-1 ;h t-1 ) a t =softmax(W a [h t ;or theory ]) or pred =W o [a t ⊙h t ]+b o Where, T t ,μ t ,Q t ,P t is the real-time temperature, viscosity, flow rate, and pressure; η pred is the output efficiency prediction value; B3: Integrate theoretical value and predicted value through DS evidence theory: B4: Solve the optimal speed N based on the gradient descent method opt : Where γ = 0.

01.

7. A real-time visual monitoring system for centrifuge operation data according to claim 1, characterized in that: The real-time monitoring and visualization unit (300) includes a multi-physics field simulation engine (301), an AR augmented reality module (302), a risk warning visualization module (303), and a historical data backtracking module (304), wherein: The multi-physics field simulation engine (301) simulates the flow field distribution, temperature gradient and rotor stress change in the centrifugal chamber in real time based on CFD and FEA technology to generate a three-dimensional dynamic simulation model; The AR augmented reality module (302) is used to superimpose and display equipment parameters, fault points and maintenance instructions in the operator's field of view through smart glasses and spatial positioning technology, supporting contactless remote collaboration; The risk warning visualization module (303) is used to present the equipment status in the form of a heat map or a three-dimensional vector map using WebGL technology, and abnormal parameters automatically trigger red, yellow, and green color warnings and mark the spatial location; The historical data backtracking module (304) is based on a time series database, supports multi-dimensional data query, and generates a traceable audit report.

8. A real-time visual monitoring system for centrifuge operation data according to claim 1, characterized in that: The distributed control system unit (400) includes a fuzzy adaptive control module (401), a multi-agent collaboration module (402), a dynamic parameter optimization module (403), and a safety interlock control module (404), wherein: The fuzzy adaptive control module (401) is used to dynamically adjust PID parameters according to material characteristic parameters; The multi-agent collaboration module (402) performs task scheduling, load balancing and fault redundancy among multiple centrifuges based on the blockchain consensus mechanism, supporting clustered production; The dynamic parameter optimization module (403) is used to integrate the genetic algorithm and the expert knowledge base to automatically optimize the centrifugal speed and temperature curve, thereby improving the separation efficiency and reducing energy consumption; The safety interlock control module (404) automatically triggers emergency shutdown and valve closing protection actions when parameters exceed the limit based on the fail-safe principle.

9. A real-time visual monitoring system for centrifuge operation data according to claim 8, characterized in that: The multi-agent collaboration module (402) performs task scheduling, load balancing, and fault redundancy among multiple centrifuges based on the blockchain consensus mechanism, supporting clustered production. The specific operations are as follows: C1: Intelligent agent identity management: Each centrifuge acts as an independent intelligent agent and generates a public and private key pair through the elliptic curve digital signature algorithm: Signature = Sign SK (m), Verify PK (signature, m), where SK is the private key; PK is the public key; m is the message; C2: Task allocation mechanism: Using the improved contract network protocol, automatic task allocation is achieved based on blockchain smart contracts: Where: C i is the current load of agent i; C max is the maximum load threshold; E i is the current energy consumption; E i,avg is the historical average energy consumption; P i is the processing accuracy; α, β, γ are weight coefficients; C3: Load balancing algorithm: Based on the dynamic weighted Byzantine fault-tolerant consensus mechanism, each agent calculates its weight using the following formula: Where: T i is the temperature; T max -T min is the temperature threshold range; R i is the reliability score; ω1, ω2, ω3 are weight coefficients; C4: Fault detection and redundancy processing: Fault detection using majority voting mechanism: When the failure probability exceeds the threshold of 0.67, the smart contract is triggered to perform task migration: migration target = argmax j (W j Compatibility j,i ), where: Compatibility j,i Score the task compatibility between agent j and agent i; C5: Consensus reaching process: An improved consensus protocol based on PBFT, including the following steps: ①Preparation phase: the master node broadcasts the preparatory message<PRE-PREPARE,v,n,d,m> ; ② Preparation phase: Slave node verifies and broadcasts preparation message<PREPARE,v,n,d,i> ; ③ Commit phase: After receiving 2f+1 prepare messages, broadcast the commit message<COMMIT,v,n,d,i> ; ④Execution phase: After receiving 2f+1 submission messages, execute the task and update the status; Where: v is the view number; n is the sequence number; d is the message digest; i is the node ID; f is the number of faulty nodes that can be tolerated; C6: Blockchain data structure: Using improved Merkle tree to store task execution records: H i,j =Hash(H i,2j ||H i,2j+1 ), where: H i,j is the hash value of the jth node in the i-th layer; || is the connection operation.

10. A real-time visual monitoring system for centrifuge operation data according to claim 1, characterized in that: The remote access and support unit (500) includes a secure communication protocol module (501), a cloud-edge collaborative architecture module (502), a Yuanda configuration and configuration module (503), and a firmware remote upgrade module (504), wherein: The secure communication protocol module (501) is used to ensure remote connection security using HTTPS and SSL encryption protocols; The cloud-edge collaborative architecture module (502) is used for efficient collaboration between cloud management and edge control; The Yuanda configuration and configuration module (503) is used for technical personnel to configure system parameters and adjust strategies through the terminal; The firmware remote upgrade module (504) is used for remote downloading and automatic updating of device programs.

Images