Single-rotor flow measuring, monitoring and alarming system of lithium battery diaphragm spraying equipment

Abstract

The invention relates to the field of industrial automatic measurement and control, and discloses a single-rotor flow measurement monitoring alarm system of lithium battery diaphragm spraying equipment, which comprises a signal acquisition module, a dynamic preprocessing module, a cooperative calculation module, a feature analysis module, an intelligent decision module and a grading alarm module, synchronously acquiring rotor pulse, three-axis vibration and pipeline pressure difference signals through a multi-mode signal fusion technology; estimating equivalent viscosity and flow deviation based on a multi-physics field coupling model, and extracting fault sensitive features in combination with high-dimensional tensor decomposition; performing fault mode matching by using the space-time association graph convolutional network, and generating a graded alarm instruction; and dynamically optimizing the adjusting quantity of the feeding pump through reinforcement learning and a hybrid control strategy. According to the method, multi-source interference under complex working conditions can be effectively inhibited, the early diagnosis capability of hidden faults is improved, and full-process closed-loop management of measurement, diagnosis and control is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation measurement and control, and in particular to a single-rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment. Background Art

[0002] In the wet spraying process of lithium battery diaphragms, single-rotor flowmeters are widely used due to their simple structure and low cost, but they face severe challenges brought by complex working conditions in actual industrial scenarios.

[0003] Traditional measurement methods rely on single-dimensional signal analysis of pulse frequency, making it difficult to effectively isolate multiple interference sources such as mechanical vibration and pressure pulsation, resulting in significant fluctuations in flow rate indications. Existing technologies typically employ fixed-threshold filtering and linear compensation strategies, which lack adaptability to nonlinear factors such as the time-varying characteristics of slurry viscosity and rotor-fluid coupling effects, and are prone to cumulative errors during high-speed spraying. Fault diagnosis modules, often based on isolated signal threshold judgments, are unable to capture subtle characteristic changes that indicate early-stage faults, posing the risk of delayed maintenance.

[0004] In terms of control strategies, conventional PID algorithms struggle to cope with dynamic changes in slurry parameters. The rigidity of control parameters can easily lead to overshoot and oscillation during the adjustment process. Furthermore, each functional unit in the system often operates independently, lacking cross-domain data fusion and collaborative decision-making mechanisms. This creates information silos and impacts overall reliability.

[0005] These limitations severely restrict the production quality of lithium battery separators and the safety of equipment operation, and new intelligent monitoring solutions are urgently needed. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention provides a single-rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment, which solves the measurement distortion caused by multi-source interference coupling and time-varying slurry characteristics of traditional single-rotor flowmeters in complex industrial scenarios, as well as the unstable spraying quality caused by fault diagnosis lag and control strategy solidification.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a single-rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment, including a signal acquisition module, a dynamic preprocessing module, a collaborative calculation module, a feature analysis module, a hierarchical alarm module and an intelligent decision-making module;

[0008] The signal acquisition module is connected to the feed pipe of the spraying equipment and is used to obtain the rotor pulse signal, vibration signal and pressure signal;

[0009] The dynamic preprocessing module performs pulse reconstruction and noise suppression on the collected signals;

[0010] The collaborative computing module performs state estimation on the pre-processed signals based on the multi-physics field coupling model and performs high-dimensional feature decomposition on the flow, vibration, and pressure signals;

[0011] The feature analysis module receives the feature decomposition results and calculates the real-time probability of each fault mode through a pattern matching algorithm;

[0012] The hierarchical alarm module triggers a multi-level alarm response based on the real-time probability data of each fault mode;

[0013] The intelligent decision-making module generates dynamic control instructions based on state estimation parameters and switches control modes according to multi-level alarm responses.

[0014] Preferably, the signal acquisition module includes:

[0015] A single-rotor pulse counter installed at the outlet of the feed pipeline is used to detect the slurry flow pulse frequency;

[0016] A three-axis vibration sensor array is installed on the rotor bearing seat to collect XYZ axial vibration acceleration;

[0017] The differential pressure transmitter connected across the inlet and outlet of the feed pipeline is used to measure the pressure difference change in real time.

[0018] Preferably, the dynamic preprocessing module comprises:

[0019] A pulse train reconstruction unit converts a discrete pulse train into a continuous waveform signal;

[0020] Multi-source noise suppression unit eliminates mechanical vibration interference through the coupling equation of vibration signal and flow signal;

[0021] The pressure compensation correction unit dynamically adjusts the flow signal gain coefficient according to the pressure difference change.

[0022] Preferably, the collaborative computing module includes:

[0023] Multi-field coupled state observer to establish a joint differential equation model including flow, pressure and equivalent viscosity;

[0024] The tensor decomposition unit performs high-dimensional tensor decomposition on flow signals, vibration signals, and pressure signals to extract core features.

[0025] Preferably, the feature analysis module includes:

[0026] Fault feature fingerprint library, which stores standard tensor features of typical fault modes;

[0027] The spatiotemporal correlation analysis unit constructs a correlation map between state parameters and fault characteristics;

[0028] Real-time pattern matching unit, calculates the similarity index between the current feature and the standard feature.

[0029] Preferably, the hierarchical alarm module includes:

[0030] Multi-dimensional fusion trigger unit, which generates alarm levels by integrating static thresholds and dynamic trend indicators;

[0031] The collaborative response execution unit triggers sound and light warnings, control mode switching and shutdown protection according to the alarm level.

[0032] Preferably, the intelligent decision-making module includes:

[0033] Dynamic strategy optimization unit, which generates feed pump control instructions based on reinforcement learning algorithm;

[0034] Adaptive mode switching unit switches between precise control mode and safety mode according to the alarm level.

[0035] The present invention also provides a single-rotor flow measurement, monitoring and alarm method for lithium battery diaphragm spraying equipment, comprising the following steps:

[0036] Real-time acquisition of rotor pulse, three-axis vibration and pipeline pressure difference signals;

[0037] Reconstruct the pulse waveform and suppress multi-source interference of the original signal;

[0038] Estimate equivalent viscosity and flow deviation through multi-physics coupling model;

[0039] Perform high-dimensional tensor decomposition on multimodal signals and extract core features;

[0040] Calculate real-time alarm level based on fault feature matching;

[0041] Dynamically adjust the feed pump control strategy according to the alarm level.

[0042] The present invention provides a single-rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment. It has the following beneficial effects:

[0043] 1. The present invention synchronously collects rotor pulses, three-axis vibrations and pipeline pressure difference signals, and establishes a multi-physical field coupling model to effectively suppress the influence of complex interferences such as mechanical vibrations and pressure pulsations on flow measurement.

[0044] 2. This invention uses spatiotemporal tensor decomposition to extract multidimensional joint features from flow, vibration, and pressure signals, overcoming the limitations of traditional single-dimensional threshold detection. A spatiotemporal correlation graph convolutional network captures the dynamic propagation of fault characteristics, enabling early identification of subtle anomalies and providing early warning of hidden faults such as rotor eccentricity and slurry agglomeration.

[0045] 3. The online calibration algorithm in the dynamic preprocessing module of this invention updates the vibration transfer coefficient and viscosity correction factor in real time, automatically adapting to changes in fluid properties caused by different slurry formulations and temperature fluctuations. The pressure compensation unit nonlinearly adjusts the gain based on the rate of change of the pressure difference, eliminating the coupling effect of spray backpressure fluctuations on flow measurement and ensuring measurement consistency across a wide range of operating conditions.

[0046] 4. Through the dynamic integration of reinforcement learning and classic PID control, this invention achieves precise tracking of feed flow under normal operating conditions and seamlessly switches to conservative control mode under abnormal conditions. A multi-level alarm linkage mechanism responds according to fault probability thresholds and change rates, preventing production interruptions caused by false alarms while providing interlocking protection in emergency situations.

[0047] 5. The present invention's feature analysis module interacts with real-time data from the control decision module to form a self-optimizing loop, continuously correcting state estimation deviations and model parameter drift. Dual-redundant hardware design and security protocols ensure fault tolerance across signal acquisition, transmission, and execution links. Combined with ECC checksum storage and policy rollback mechanisms, this system meets the stringent system reliability requirements of lithium battery production. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 Schematic diagram of the system architecture of the present invention;

[0049] Figure 2 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0050] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments 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 creative efforts are within the scope of protection of the present invention.

[0051] Please see the attached Figure 1 The present invention provides a single-rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment. The system realizes precise control of spraying flow and real-time warning of abnormal conditions through multimodal signal fusion and intelligent decision-making mechanism.

[0052] like Figure 1 As shown, the single rotor flow measurement monitoring and alarm system of the lithium battery diaphragm spraying equipment can include a signal acquisition module, a dynamic preprocessing module, a collaborative calculation module, a feature analysis module, a hierarchical alarm module and an intelligent decision-making module.

[0053] The following is a detailed description of each module in the system of the present invention, which comprehensively explains the specific implementation principles, technical details and processes of each module.

[0054] Signal acquisition module

[0055] The signal acquisition module of the present invention consists of three sensing units, each used to acquire rotor pulse signals, triaxial vibration signals, and pipeline pressure differential signals. Each sensing unit achieves synchronous data transmission via a bus architecture, providing raw input for subsequent signal processing.

[0056] The single-rotor pulse counter is installed outside the discharge flange of the feed pipeline, maintaining a preset distance from the rotor blades. When the slurry pushes the rotor to rotate, the blades periodically cut the magnetic flux lines, generating an electrical pulse signal. The pulse frequency is linearly related to the slurry flow rate, specifically meeting the following requirements:

[0057]

[0058] in, Indicates the pulse frequency, is the volume flow rate, The pulse signal is converted into a standard square wave sequence by the shaping circuit, and the rising edge of the square wave corresponds to the moment when the blade passes the detection point.

[0059] A triaxial vibration sensor array, located on the rotor bearing housing, consists of three orthogonally spaced acceleration sensing units. Each unit detects in the radial, axial, and tangential directions parallel to the rotor. When collecting vibration signals, they first pass through an analog filter to eliminate high-frequency noise. The filter's cutoff frequency is set to within ten times the rotor's fundamental frequency. The filtered signal is then scaled to the ADC range using a programmable gain amplifier and ultimately output as a signed digital quantity.

[0060] The differential pressure transmitter is connected between the feed line inlet and the nozzle tip. Its diaphragm senses the inlet static pressure and nozzle back pressure on either side, respectively. The differential pressure signal is converted to a differential voltage using a Wheatstone bridge circuit and then converted to a digital value using a 24-bit Σ-Δ ADC. The transmitter also features built-in temperature compensation circuitry to eliminate zero-point drift caused by ambient temperature fluctuations.

[0061] The output signals of each sensor are connected to the bus system via time-division multiplexing. The bus controller allocates a separate time slot for each data frame. The frame structure includes a start bit, a sensor identification code, a data payload, and a CRC checksum. The data transmission cycle is synchronized with the rotor rotation cycle, ensuring that the sampling time strictly corresponds to the mechanical motion state.

[0062] The pulse counter and vibration sensor use a common ground connection, with a ground loop resistance of less than 0.1Ω to suppress common-mode interference. The differential pressure transmitter is powered by an isolated power supply module, with a magnetic bead filter placed between the power ground and the signal ground to block conducted interference. Signal cables are installed using twisted-pair shielded cable, with the shield grounded at a single point to the device's metal casing.

[0063] The signal acquisition module has a self-test function that sends a handshake command to each sensor upon power-up. A hardware anomaly alarm is triggered if the pulse counter detects no valid pulses for three consecutive cycles, or if the vibration sensor output exceeds the measuring range. The differential pressure transmitter has a built-in over-limit detection circuit that directly cuts off the feed pump power supply when the instantaneous pressure differential exceeds the safety threshold.

[0064] Dynamic preprocessing module

[0065] The dynamic preprocessing module of the present invention receives the raw pulse, vibration, and pressure differential signals from the signal acquisition module and generates a purified flow compensation signal through multi-stage processing. This module comprises three core processing units: pulse sequence reconstruction, multi-source noise suppression, and pressure compensation correction. These units are connected in series to form a serial processing chain.

[0066] The pulse sequence reconstruction unit first converts the discrete pulse sequence into a continuous waveform. Specifically, it identifies the rising and falling edge moments of the original pulse signal. , establish a timestamp sequence. Then generate continuous flow estimates through Gaussian kernel convolution:

[0067]

[0068] in, The kernel width parameter is related to the rotor inertia and is set by default to one-quarter of the rotor rotation period. This parameter can be dynamically adjusted based on the measured pulse interval to accommodate signal smoothing requirements at different slurry flow rates.

[0069] The multi-source noise suppression unit is used to eliminate the coupling interference of mechanical vibration on the flow signal. First, the differential component of the three-axis vibration acceleration signal is extracted. , establish the vibration-flow coupling equation:

[0070]

[0071] Where, is the vibration transmission coefficient, which reflects the influence weight of different axial vibrations on flow measurement. It is calibrated online by recursive least squares method. The specific process is: is the observed value, For input variables, construct an overdetermined system of equations and iteratively solve the coefficient matrix.

[0072] The pressure compensation correction unit dynamically adjusts the flow gain according to the real-time pressure difference change. Calculate the first-order derivative of the pressure difference signal , introduce a nonlinear correction factor:

[0073]

[0074] in, The viscosity-related parameter is extracted from the historical data by using a sliding window regression algorithm. The sliding window length is set to an integer multiple of the rotor rotation period to ensure the integrity of the data period. Transmitted to the collaborative computing module via the SPI bus.

[0075] Data transfer between processing units utilizes a double-buffering mechanism. Pre-processing results are written to a circular buffer, which is then read by subsequent processing units in a fixed time slice. The buffer depth is set to twice the maximum signal delay time to prevent data overflow. Intermediate variables generated during processing are stored in non-volatile memory for state reconstruction upon system recovery.

[0076] The signal transmission bus utilizes a differential signaling protocol. Data frames consist of a start synchronization header, data payload, and a cyclic redundancy check (CRC) code. The synchronization header consists of a square wave sequence at a specific frequency and is used for clock synchronization at the receiving end. The CRC-16 algorithm is used to generate the checksum, which can detect and correct single-bit transmission errors. The bus controller integrates a hardware watchdog circuit that triggers a system reset signal when continuous data frame loss exceeds a preset threshold.

[0077] The module incorporates an anomaly detection unit that generates a hardware fault flag when the reconstructed flow signal exceeds the physical measurement range or when the vibration signal spectrum energy is significantly missing at the rotor fundamental frequency. This flag is transmitted via a separate signal line to the graded alarm module, triggering the sensor calibration process. During calibration, the pulse reconstruction unit switches to fixed-frequency test mode, verifying channel integrity by injecting simulated pulses of known amplitude.

[0078] Collaborative computing module

[0079] The collaborative computing module, described in this invention, receives the purified flow, vibration, and differential pressure signals output by the dynamic preprocessing module and performs multi-field coupled state estimation and high-dimensional eigendecomposition. This module consists of a multi-field coupled state observer and a tensor decomposition unit, which exchange intermediate data via shared memory, forming a closed-loop computing architecture.

[0080] The multi-field coupled state observer first establishes a joint differential equation model. Define the state vector ,in is the instantaneous value of flow, is the flow rate change rate, is the equivalent viscosity estimate. According to the mass conservation and momentum theorem, the nonlinear state equation is constructed:

[0081]

[0082] Where, is the rotor moment of inertia, is the cross-sectional area of ​​the flow channel, is the time-varying coefficient of viscosity, is the process noise vector. The observation matrix Select as , so that the system output is consistent with the flow measurement value.

[0083] State estimation is achieved using the unscented Kalman filter algorithm. First, select Sigma points (n=3 is the state dimension), and the predicted mean and covariance are calculated through nonlinear transformation:

[0084]

[0085]

[0086]

[0087] in, and is the weight coefficient, is the process noise covariance matrix. In the update phase, the Kalman gain is calculated by measuring the residual error and the optimal state estimate is output. .

[0088] The tensor decomposition unit extracts high-order features from multimodal signals. First, a three-dimensional tensor is constructed. ,in is the length of the time series, and the three channels correspond to flow, vibration energy and pressure difference signals respectively. is the sliding window length. The original tensor is decomposed into the product of the kernel tensor and the factor matrix by Tucker decomposition:

[0089]

[0090] Preserve core tensors As a feature fingerprint. Factor matrix Corresponding time mode, Corresponding signal type mode, The decomposition process is implemented using the high-order orthogonal iteration algorithm (HOOI), which optimizes each factor matrix through alternating least squares.

[0091] Core Tensor is further processed to extract fault-sensitive features. Perform first-order difference operations along the time dimension:

[0092]

[0093] Generate differential feature tensor . Concatenate the original core tensor and the difference tensor in the channel dimension to form the final feature vector , the calculation formula is:

[0094]

[0095] in, Represents vector concatenation operation, is a tensor vectorization function.

[0096] The state observer interacts with the tensor decomposition unit through the data bus. is written to the shared memory area for use by the tensor decomposition unit as an additional feature channel when constructing the tensor. At the same time, the feature vector output by the tensor decomposition Feedback to the state observer for online updating of the process noise covariance matrix This feedback mechanism enables state estimation to adapt to changes in working conditions and improves model tracking accuracy.

[0097] Feature analysis module

[0098] The feature analysis module of the present invention receives the core tensor features output by the collaborative computing module and calculates fault probability through pattern matching and spatiotemporal correlation analysis. This module consists of a fault feature fingerprint library, a spatiotemporal correlation analysis unit, and a real-time pattern matching unit. These units interact with each other through a shared database, forming a layered processing architecture.

[0099] The fault feature fingerprint library stores the standard tensor features of five typical faults, including pump wear, pipe blockage, rotor eccentricity, slurry agglomeration and sensor failure. Each type of fault stores 100 sets of historical operating condition data, each set of data contains the core tensor Data storage uses a columnar database structure, with tables indexed by fault type, supporting timestamp-based range queries and batch reads.

[0100] The spatiotemporal correlation analysis unit first constructs a state-feature correlation map. The state estimation value output by the collaborative calculation module is converted into With core tensors Perform a tensor product operation:

[0101]

[0102] Where, represents the first-order expansion matrix of the tensor, Represents the Kronecker product. The resulting three-dimensional correlation tensor is reshaped into a 2D matrix , as the input of the graph convolutional network.

[0103] The graph convolutional network contains three layers of convolution operations, and the update formula for each layer of nodes is:

[0104]

[0105] in, To add a self-connected adjacency matrix, is the degree matrix, is the trainable weight matrix. The network output node feature matrix , after global average pooling, the fault probability vector is generated .

[0106] The real-time pattern matching unit calculates the similarity index between the current feature and the standard feature. The similarity calculation formula for class faults is:

[0107]

[0108] Where, represents the Frobenius inner product, is the time difference between the current time and the time when the historical fault occurred, is the decay time constant. The final probability value is normalized by the Softmax function:

[0109]

[0110] The module sets up a cache queue inside to store the core tensor data of the last 60 seconds. When the threshold is exceeded three times in a row, the data replay mechanism is triggered, writing the tensor data for the associated time period to non-volatile memory. The replayed data includes timestamps, original signal segments, and state estimates for offline fault reproduction analysis.

[0111] The fault feature fingerprint library is updated using an incremental learning strategy. When a new fault type is confirmed, the current core tensor is extracted. The PCA principal components of the dataset are inserted into the database and the index is rebuilt. Historical data is eliminated according to the LRU strategy, and the latest 100 sets of data are retained to ensure storage efficiency. During the update process, the Sherman-Morrison correction term of the covariance matrix is ​​calculated:

[0112]

[0113] This accelerates feature space reconstruction and avoids full recalculation.

[0114] The module interacts with the hierarchical alarm module through interrupt signals. When the threshold is exceeded for the first time, a low priority interrupt request is sent, triggering the early warning log record. If the threshold is exceeded for a specified period of time, a high-priority interrupt is sent, directly activating the alarm execution unit. A semaphore lock is set in the interrupt service routine to prevent multi-threaded access conflicts.

[0115] Hierarchical alarm module

[0116] The hierarchical alarm module of this invention receives the fault probability vector output by the feature analysis module, generates an alarm level through multi-dimensional fusion decision-making, and triggers the corresponding execution action. This module is composed of a multi-dimensional fusion trigger unit and a coordinated response execution unit. The two units interact with each other in real time through an interrupt controller, forming a closed-loop alarm control chain.

[0117] The multi-dimensional fusion trigger unit first establishes the composite alarm condition. Class fault, define static threshold and dynamic trend thresholds , calculate the current time Alarm trigger indicators:

[0118]

[0119] in, is the length of the sliding time window, which is an integer multiple of the rotor rotation period. Generates a primary alarm flag when It is upgraded to a medium alarm when three consecutive sampling cycles A high-level alarm is triggered.

[0120] The coordinated response execution unit executes a preset combination of actions based on the alarm level. In the primary alarm state, the HMI interface's color alert and data logging functions are activated. Specifically, this involves sending the alarm code 0x01 to the HMI via the Modbus TCP protocol and simultaneously opening a ring buffer to store the last 60 seconds of fault probability data. In the intermediate alarm state, the frequency-adjustable audible and visual alarm system is triggered, with the buzzer generating a 2 kHz pulse modulated signal and the LED indicator flashing at a 50% duty cycle square wave.

[0121] In the advanced alarm state, the actuator linkage control includes the following steps: First, send a deceleration command to the feed pump driver to make it The speed is linearly reduced to a safe value within a specified timeframe. The solenoid valve power is then disconnected, and physical isolation is achieved via a hard-wired relay. Finally, a shutdown report message containing a timestamp, fault code, and key parameters is sent to the host computer. Priority arbitration logic is implemented between each action. When multiple alarms are triggered simultaneously, they are handled in the order of "advanced > intermediate > elementary."

[0122] The module incorporates dual redundant state machines, with the primary and backup machines maintaining synchronization through a heartbeat packet mechanism. Every 10 ms, the primary sends a synchronization frame to the backup machine, containing the current alarm level, execution status, and a checksum. If five consecutive heartbeat packets are lost, the backup machine automatically assumes control and sends a failover notification to the intelligent decision-making module via an independent signal line. During the failover process, any unfinished actions continue to execute, preventing loss of control instructions.

[0123] Intelligent decision-making module

[0124] The intelligent decision-making module, described in this invention, receives state estimation parameters from the collaborative computing module and feedback signals from the hierarchical alarm module, generating dynamic control instructions using a reinforcement learning algorithm. This module consists of a dynamic policy optimization unit and an adaptive mode switching unit, which synchronize parameters through a shared policy library, forming a closed-loop control architecture.

[0125] The dynamic policy optimization unit is designed using the Deep Deterministic Policy Gradient (DDPG) algorithm. The state space is defined as , the action space is the feed pump speed adjustment amount The actor network outputs a deterministic policy:

[0126]

[0127] in, is the network parameter matrix. The critic network evaluates the state-action value function:

[0128]

[0129] Target network parameters and Gradual synchronization via soft update strategy:

[0130]

[0131] The reward function is designed as:

[0132]

[0133] Experience replay buffer stores transfer samples , the priority sampling probability is proportional to the absolute value of the TD error.

[0134] The adaptive mode switching unit establishes a three-level control strategy mapping relationship. When the feedback level of the hierarchical alarm module is 0, the precise control mode is activated:

[0135]

[0136] in, , When the alarm level is ≥1, it switches to safe mode:

[0137]

[0138] Saturation function Limit the output to the range of [0,100%]. Lock the actuator output in emergency mode:

[0139]

[0140] in, is the alarm triggering time, is the system inertia time constant.

[0141] A strategy version controller is set inside the module. When the cumulative reward value drops continuously for more than a preset number of times, the strategy rollback mechanism is triggered. The optimal strategy parameters are loaded from the historical strategy library, and the parameter compatibility is checked through cosine similarity:

[0142]

[0143] Meet the similarity threshold The strategy switching is completed when , otherwise the parameter interpolation transition process is started.

[0144] In general, the system of the present invention includes signal acquisition, dynamic preprocessing, collaborative computing, feature analysis, hierarchical alarm, and intelligent decision-making modules. It uses multi-physics field coupling modeling, high-dimensional tensor decomposition, and reinforcement learning algorithms to achieve noise suppression of flow signals, fault feature extraction, and dynamic optimization of control strategies. Through complex domain signal perturbation injection, spatiotemporal correlation graph convolutional networks, and hybrid control mode switching technology, a closed-loop control and collaborative alarm mechanism is formed. It has the ability to suppress multi-source interference, adaptively adjust working conditions, and respond in a hierarchical manner, meeting the stringent measurement accuracy and system reliability requirements of lithium battery separator production.

[0145] like Figure 2 As shown, the present invention also provides a single rotor flow measurement, monitoring and alarm method for lithium battery diaphragm spraying equipment based on the above system, comprising the following steps:

[0146] S1, real-time acquisition of rotor pulse, three-axis vibration and pipeline pressure difference signals;

[0147] The signal acquisition module captures rotor pulse signals, triaxial vibration acceleration, and pipeline differential pressure data in real time, synchronously transmitting these three signals to the dynamic preprocessing module via a bus protocol. The pulse signal undergoes shaping and filtering to generate a standard square wave sequence. The vibration signal undergoes anti-aliasing filtering to retain the valid frequency band, and the differential pressure signal is compensated for temperature drift.

[0148] S2, reconstructing the pulse waveform and suppressing multi-source interference on the original signal;

[0149] The dynamic preprocessing module performs multi-level processing on the original signal: first, the discrete pulse sequence is reconstructed into a continuous flow waveform to eliminate the quantization error caused by signal discretization; then a vibration-flow coupling model is established to remove the mechanical vibration interference component through the online calibrated vibration transfer coefficient; finally, the flow gain is dynamically corrected according to the pressure difference change rate to compensate for the nonlinear deviation caused by slurry viscosity fluctuations.

[0150] S3, estimating equivalent viscosity and flow deviation through multi-physics coupling model;

[0151] The collaborative computing module receives the purified multi-source signals and performs two-way parallel computing:

[0152] State estimation channel: Based on the rotor dynamics and fluid mechanics coupling model, it can calculate the equivalent viscosity, flow deviation and dynamic parameter error in real time;

[0153] Feature extraction channel: Flow, vibration, and pressure signals are constructed into a three-dimensional space-time tensor, and the core feature vectors representing the health status of the system are extracted through high-order decomposition.

[0154] S4, perform high-dimensional tensor decomposition on multimodal signals and extract core features;

[0155] The feature analysis module performs spatiotemporal correlation matching of core features with a pre-defined fault fingerprint library, calculating the real-time probability distribution of each fault mode. The graded alarm module generates three levels of alarm instructions based on probability thresholds and changing trends: a primary warning triggers data recording, an intermediate alarm activates audible and visual warnings, and a high-level alarm interlocks and shuts down the system for protection. Alarm levels are fed back to the intelligent decision-making module in real time via an independent communication link.

[0156] S5. Calculate the real-time alarm level based on the fault feature matching degree;

[0157] The intelligent decision-making module switches control modes based on alarm levels. Under normal operating conditions, a deep reinforcement learning algorithm is used to optimize the feed pump speed online. A PID parameter self-tuning mechanism is introduced for primary alarms. For intermediate or high-level alarms, a conservative control strategy is adopted, gradually reducing the feed flow rate and locking the actuator. Control commands are transmitted to the execution terminal via the industrial bus, forming a closed loop of perception, diagnosis, and control.

[0158] S6. Dynamically adjust the feed pump control strategy according to the alarm level;

[0159] The feed pump's real-time adjustment is fed back into the collaborative computing module, which updates the multiphysics model parameters. The feature analysis module continuously monitors the control effectiveness. When the failure probability drops to a safe threshold, it sends a state recovery command to the alarm module, gradually returning the system to precise control mode, completing the complete abnormal condition processing cycle.

[0160] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A single rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment, characterized in that: It includes signal acquisition module, dynamic pre-processing module, collaborative computing module, feature analysis module, hierarchical alarm module and intelligent decision-making module; The signal acquisition module is connected to the feed pipe of the spraying equipment and is used to obtain the rotor pulse signal, vibration signal and pressure signal; The dynamic preprocessing module performs pulse reconstruction and noise suppression on the collected signals; The collaborative computing module performs state estimation on the pre-processed signals based on the multi-physics field coupling model and performs high-dimensional feature decomposition on the flow, vibration, and pressure signals; The feature analysis module receives the feature decomposition results and calculates the real-time probability of each fault mode through a pattern matching algorithm; The hierarchical alarm module triggers a multi-level alarm response based on the real-time probability data of each fault mode; The intelligent decision-making module generates dynamic control instructions based on state estimation parameters and switches control modes according to multi-level alarm responses.

2. The single rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment according to claim 1 is characterized in that: The signal acquisition module includes: A single-rotor pulse counter installed at the outlet of the feed pipeline is used to detect the slurry flow pulse frequency; A three-axis vibration sensor array is installed on the rotor bearing seat to collect XYZ axial vibration acceleration; The differential pressure transmitter connected across the inlet and outlet of the feed pipeline is used to measure the pressure difference change in real time.

3. The single rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment according to claim 1 is characterized in that: The dynamic preprocessing module comprises: A pulse train reconstruction unit converts a discrete pulse train into a continuous waveform signal; Multi-source noise suppression unit eliminates mechanical vibration interference through the coupling equation of vibration signal and flow signal; The pressure compensation correction unit dynamically adjusts the flow signal gain coefficient according to the pressure difference change.

4. The single rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment according to claim 1 is characterized in that: The collaborative computing module includes: Multi-field coupled state observer to establish a joint differential equation model including flow, pressure and equivalent viscosity; The tensor decomposition unit performs high-dimensional tensor decomposition on flow signals, vibration signals, and pressure signals to extract core features.

5. The single rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment according to claim 1 is characterized in that: The feature analysis module includes: Fault feature fingerprint library, which stores standard tensor features of typical fault modes; The spatiotemporal correlation analysis unit constructs a correlation map between state parameters and fault characteristics; Real-time pattern matching unit, calculates the similarity index between the current feature and the standard feature.

6. The single rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment according to claim 1 is characterized in that: The hierarchical alarm module includes: Multi-dimensional fusion trigger unit, which generates alarm levels by integrating static thresholds and dynamic trend indicators; The collaborative response execution unit triggers sound and light warnings, control mode switching and shutdown protection according to the alarm level.

7. The single rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment according to claim 1 is characterized in that: The intelligent decision-making module includes: Dynamic strategy optimization unit, which generates feed pump control instructions based on reinforcement learning algorithm; Adaptive mode switching unit switches between precise control mode and safety mode according to the alarm level.

8. A single rotor flow measurement, monitoring and alarm method for lithium battery diaphragm spraying equipment, based on the single rotor flow measurement, monitoring and alarm system for lithium battery diaphragm spraying equipment according to any one of claims 1 to 7, characterized in that: The following steps are involved: Real-time acquisition of rotor pulse, three-axis vibration and pipeline pressure difference signals; Reconstruct the pulse waveform and suppress multi-source interference of the original signal; Estimate equivalent viscosity and flow deviation through multi-physics coupling model; Perform high-dimensional tensor decomposition on multimodal signals and extract core features; Calculate real-time alarm level based on fault feature matching; Dynamically adjust the feed pump control strategy according to the alarm level.

Images