Simulation prediction-based industrial vibrating screen data acquisition evaluation and analysis method

Abstract

This invention relates to the field of industrial vibrating screen simulation and testing technology, specifically a simulation-based prediction-based method for industrial vibrating screen data acquisition, evaluation, and analysis. The method includes: constructing a simulation logical topology structure containing the vibrating screen equipment and its physical connections; integrating heterogeneous sensor data from the field to generate a standardized operating parameter sequence; extracting the performance target parameter set for the current evaluation cycle; inputting these three parameters into a multimodal simulation prediction model; using the simulation logical topology as constraints and the performance target parameters as expected convergence conditions; dynamically extrapolating and outputting the future state simulation trajectory of the entire production line process; analyzing the predicted values ​​of key performance indicators at the end of the evaluation cycle; calculating multi-dimensional performance deviation metrics; and integrating these metrics to generate the evaluation results. This method enables continuous simulation extrapolation of the entire screening production line process, accurately matching the performance orientation of the evaluation cycle, and improving the completeness and accuracy of state prediction and evaluation.

Description

Technical Field

[0001] This invention relates to the field of industrial vibrating screen simulation and testing technology, and in particular to a method for data acquisition, evaluation and analysis of industrial vibrating screens based on simulation prediction. Background Technology

[0002] The data acquisition and evaluation of existing industrial vibrating screening production lines mostly rely on on-site physical sensors to obtain heterogeneous data. Only basic standardization processing is performed on the data to form local operating parameters. Single equipment status monitoring is carried out based on the conventional configuration parameters of the production line. A simulation structure covering all vibrating screening equipment and physical connection relationships of the production line has not been built. The evaluation process only refers to historical conventional parameters and does not match the specific performance target parameters for the current evaluation period.

[0003] Existing technologies cannot transform the topological relationships of production line equipment into simulation constraints, nor can they use performance target parameters as convergence conditions for prediction models. They can only achieve status monitoring of discrete points and single equipment, and cannot conduct dynamic simulation and deduction of the entire production line process. They cannot generate continuous future state simulation trajectories, cannot accurately extract key performance indicators at the end of the evaluation cycle, and performance deviations can only be calculated for local parameters, thus failing to form a complete full-process evaluation result.

[0004] It is necessary to construct a simulation logical topology structure covering the entire production line vibrating screening equipment and physical connection topology and use it as the model constraint condition. The performance target parameters of the current evaluation cycle are introduced as the expected convergence condition of the model. The simulation trajectory of the future state of the entire process is deduced through the multimodal simulation prediction model. Based on the simulation trajectory, the key performance indicators at the end of the evaluation cycle are analyzed, and multi-dimensional performance deviation measurement is completed and integrated to generate the evaluation results. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a simulation-based prediction-based method for data acquisition, evaluation, and analysis of industrial vibrating screening.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a simulation-based prediction-based method for data acquisition, evaluation, and analysis of industrial vibrating screening, comprising: Construct a simulation logic topology structure that includes all vibrating screening equipment and their physical connection topology in the target screening production line; The system receives heterogeneous sensor data transmitted from multiple physical sensors installed on-site at the vibrating screening equipment in real time, and generates a standardized operating parameter sequence. Extract the set of performance target parameters for the target screening production line for the current evaluation period from the system configuration library; The standardized sequence of operating parameters, the simulation logic topology, and the set of performance target parameters are input into the trained multimodal simulation prediction model, where the simulation logic topology serves as a constraint and the set of performance target parameters serves as the expected convergence condition. The multimodal simulation prediction model is activated, and under the combined effect of constraints and expected convergence conditions, it dynamically extrapolates the standardized operating parameter sequence and outputs the simulation trajectory of the entire future state of the target screening production line. The simulation trajectory of the future state of the entire process is analyzed to separate the set of predicted values ​​of key performance indicators for the end point of the current evaluation cycle. The set of predicted values ​​of key performance indicators is compared with the set of performance target parameters to calculate a multi-dimensional set of performance deviation measures. The evaluation result is generated by integrating the set of predicted values ​​of key performance indicators, the set of performance target parameters, and the set of multi-dimensional performance deviation measures.

[0007] As a further aspect of the present invention, heterogeneous sensor data transmitted from multiple physical sensors installed on-site at the vibrating screening equipment are accessed in real time to generate a standardized operating parameter sequence, including: The physical connection topology reflects the flow direction of materials in the actual production line; The heterogeneous sensor data includes vibration status data and material flow rate data; The heterogeneous sensor data that is accessed in real time is subjected to heterogeneous data assimilation processing to generate a standardized sequence of operating parameters with a unified timestamp and a unified data format. The heterogeneous data assimilation process for the real-time accessed heterogeneous sensor data specifically includes: A dedicated data parsing adapter is configured for each type of heterogeneous sensor data, and the data parsing adapter extracts valid numerical fields that are directly related to the state of the screening equipment from the raw data stream; Start a global time synchronization service, which provides a reference clock signal for the entire system based on the network time synchronization signal; Each valid numerical field output by the data parsing adapter is marked with a timestamp provided by the global time synchronization service to form a raw data point with a time identifier; All raw data points with time stamps are imported into a unified data format converter. The data format converter linearly maps raw data points with different dimensions and different numerical ranges to a standard numerical range between zero and one, generating standardized data points. All standardized data points within the preset time alignment window are aligned and packaged according to their timestamps to generate a standardized runtime parameter frame containing multiple parameter types and with strictly aligned timestamps. The standardized operating parameter sequence is formed by arranging consecutive standardized operating parameter frames in chronological order.

[0008] As a further aspect of the present invention, the step of extracting the set of performance target parameters for the target screening production line for the current evaluation cycle from the system configuration library includes: Receive the evaluation period definition instruction submitted by the user through the interactive interface, wherein the evaluation period definition instruction includes the start time and end time of the evaluation period; According to the evaluation cycle definition instruction, a search is performed in the system configuration library to find the various performance expectations that are pre-set for the target screening production line within the time range of the evaluation cycle; If multiple independent performance expectations are retrieved, the performance expectations are categorized according to their defined performance index types to form a preliminary subset of performance targets. Read the historical operation records of the target screening production line, and extract the actual average performance value of the same historical period that is similar to the evaluation cycle in terms of season and production plan from the historical operation records; Each expected performance value in the preliminary performance target subset is compared with the corresponding actual average performance value for the same historical period. When a comparison reveals that a certain performance expectation deviates from the actual average performance value of the corresponding historical period, and the degree of deviation exceeds the preset reasonableness threshold, a configuration anomaly warning will be generated. If no configuration anomaly warning is generated, or if the generated configuration anomaly warning is confirmed to be ignored by the user, the preliminary performance target subset will be formally determined as the set of performance target parameters used for the current simulation prediction.

[0009] As a further aspect of the present invention, the multimodal simulation prediction model is activated, and under the combined effect of constraints and expected convergence conditions, it dynamically extrapolates the standardized operating parameter sequence, including the following progressive steps: The standardized operating parameter sequence at the current moment is used as the initial state input of the multimodal simulation prediction model; The multimodal simulation prediction model contains a neural network computation graph that is isomorphic to the simulation logic topology. The neural network computation graph simulates the flow and sieving transformation of materials in the physical connection topology. Within each model's internal simulation step, the simulated state parameters of each vibrating screening device calculated in the previous step are transmitted to the downstream connected device nodes as simulated input parameters, according to the flow direction defined by the simulation logic topology. At each device node, the received analog input parameters are processed through a dedicated analog sub-network, and the expected constraints imposed by the set of performance target parameters are combined to calculate the analog output parameters and analog energy consumption parameters of the device node at the current step size. Collect the simulated output parameters of all device nodes at the current step size, and summarize them into a snapshot of the overall state of the target screening production line at the current step size. The steps of transmission, calculation and summarization are repeated until the model's internal simulation time reaches the end of the current evaluation cycle. All overall state simulation snapshots generated during this process are connected in chronological order to form the full-process future state simulation trajectory.

[0010] As a further aspect of the present invention, the progressive steps of dynamic deduction also include an online correction mechanism for the simulated state, the specific steps of which are as follows: A series of synchronization checkpoints are set to establish the correspondence between the internal extrapolation time of the multimodal simulation prediction model and the actual system time. Whenever the model's internal simulation time reaches a synchronization checkpoint, the system pauses the model simulation and obtains the latest standardized operating parameter frame corresponding to the current system's actual time from the real-time data access stream as the real-state observation value. The simulated state snapshot generated by the model at the current synchronization checkpoint is compared with the actual state observation at the same point, and the state parameter difference vector is calculated. The state parameter difference vector is fed back into the state correction module of the multimodal simulation prediction model; The state correction module dynamically adjusts the internal state parameters on which the model depends for subsequent inferences based on the state parameter difference vector, so that the subsequently generated simulated state snapshots are corrected in the direction of the actual observations, and then the inference of subsequent steps continues.

[0011] As a further aspect of the present invention, the step of calculating the state parameter difference vector specifically includes: For each type of state parameter of the same dimension contained in the real state observation and the simulated state snapshot, perform arithmetic subtraction to obtain a series of original differences; A normalization coefficient is preset for each type of state parameter. The original difference of each type of state parameter is multiplied by its corresponding normalization coefficient and converted into a dimensionless standardized difference. The standardized differences corresponding to all state parameters are weighted and combined according to the preset parameter importance weight vector to obtain a scalarized comprehensive state difference. Only when the overall state difference exceeds the preset difference tolerance threshold will the state parameter difference vector and the standardized differences of the corresponding state parameters be transmitted to the state correction module for processing.

[0012] As a further aspect of the present invention, the simulation trajectory of the future state of the entire process is analyzed to separate the set of predicted values ​​of key performance indicators for the end point of the current evaluation cycle. The steps are as follows: Load a predefined key performance indicator extraction rule base, which defines the calculation logic of various key performance indicators and their parameter dependencies in the full-process future state simulation trajectory; From the full-process future state simulation trajectory, extract the final state simulation snapshot corresponding to the termination time of the current evaluation cycle; The parameters in the final state simulation snapshot are substituted into the calculation according to the calculation logic defined in the rule base of the key performance indicators. The calculated set of numerical results is output as the predicted set of key performance indicators.

[0013] As a further aspect of the present invention, the calculation of the set of predicted key performance indicators also includes referencing historical simulation trajectory trends, specifically: While capturing the final state simulation snapshot, a trajectory segment of a fixed time length before the termination time is also captured from the full-process future state simulation trajectory as the pre-termination state trajectory. Statistical analysis is performed on the trajectory before termination to calculate the rate of change of the main state parameters in the trajectory within the fixed time length. From the key performance indicator extraction rule base, find the specific performance indicators whose calculation logic is affected by the parameter change rate; For the specific performance index, the calculation result based on the final state simulation snapshot is fused with the change rate of the state parameter corresponding to the specific performance index. The original calculation result based on the final state simulation snapshot is replaced with the fused correction value, thereby updating the set of predicted values ​​for the key performance index.

[0014] As a further aspect of the present invention, the calculation yields a multi-dimensional set of performance deviation metrics, including: From the set of predicted values ​​for key performance indicators and the set of performance target parameters, one pair of predicted values ​​and target values ​​for the same performance indicator are matched one by one. For each pair of predicted and target values, calculate the absolute deviation of the predicted value relative to the target value; Read the preset deviation range reference value for the performance index from the system configuration library; Dividing the absolute deviation by the deviation range reference value yields a dimensionless relative deviation coefficient. The relative deviation coefficients corresponding to all performance indicators are organized according to their indicator categories to form the multi-dimensional performance deviation measurement set, where each dimension corresponds to the relative deviation coefficient of a performance indicator.

[0015] As a further aspect of the present invention, a deep analysis step of the performance deviation metric set is also included: Establish a performance index correlation graph, which describes the mutual influence relationships and influence weights among different performance indicators; The multi-dimensional set of performance deviation metrics is input into the performance index correlation graph; The performance index correlation graph calculates the relative deviation coefficient of each dimension and the indirect impact value on other related dimensions based on the internally defined mutual influence relationships and influence weights. By summing the direct relative deviation coefficient of each dimension and the indirect influence values ​​passed from all other dimensions, the overall influence strength of each dimension is obtained. All performance indicators are sorted in descending order of their overall impact intensity to generate a sorted list of key deviation dimensions. The key deviation dimension sorting list is integrated with the multi-dimensional performance deviation measurement set, key performance indicator prediction value set, and performance target parameter set to generate a final evaluation result containing in-depth analysis information.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: A simulation logical topology structure containing all vibrating screening equipment and their physical connection topology in the target screening production line is constructed. This simulation logical topology structure is directly used as the constraint condition of the multimodal simulation prediction model. The performance target parameter set of the target screening production line for the current evaluation period is extracted from the system configuration library. This performance target parameter set is used as the expected convergence condition of the multimodal simulation prediction model. The standardized operating parameter sequence, simulation logical topology structure, and performance target parameter set are synchronously input into the multimodal simulation prediction model, so that the model derivation process matches the overall physical connection architecture of the production line, the model calculation process follows the actual equipment correlation of the production line, and the model convergence direction fits the performance setting of the current evaluation period. This reduces the derivation deviation caused by the disconnect between local parameters and the overall production line architecture in conventional data evaluation, and avoids the prediction direction deviation problem caused by not combining the performance target of the specific evaluation period.

[0017] Under the combined influence of constraints and expected convergence conditions, the multimodal simulation prediction model dynamically extrapolates the standardized operating parameter sequence, outputting the simulation trajectory of the entire future state of the target screening production line. It then performs analytical operations on this trajectory, separating the set of predicted key performance indicators (KPIs) corresponding to the end of the current evaluation cycle. By comparing this set with the set of performance target parameters, a multi-dimensional performance deviation measurement set is obtained. Finally, the KPIs, performance target parameters, and multi-dimensional performance deviation measurement set are integrated to form a complete evaluation result. This enables continuous simulation extrapolation of the entire production line's state, accurately obtaining performance indicator values ​​at the end of the evaluation cycle, generating multi-dimensional performance deviation quantification results, and constructing an evaluation system covering the entire production line. This approach abandons the conventional method of analyzing only discrete points and single equipment states, forming a complete evaluation process from simulation extrapolation to indicator analysis and deviation calculation. Attached Figure Description

[0018] Figure 1 This is a flowchart of the simulation prediction-based industrial vibration screening data acquisition, evaluation, and analysis method described in this invention. Figure 2 A flowchart for extracting the set of performance target parameters; Figure 3 This is a comprehensive state difference monitoring map for the entire lifecycle; Figure 4 A dual-axis trend chart for the simulation and deduction stage of an industrial vibrating screen production line; Figure 5 This is a diagram for calculating the performance deviation of an industrial vibrating screen production line. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 This invention provides a simulation-based prediction-based method for data acquisition, evaluation, and analysis of industrial vibrating screening. The specific method includes: A simulation logical topology structure is constructed, encompassing all vibrating screening equipment and their physical connections within the target screening production line. This structure defines the virtual connection relationships between the equipment. Real-time access is provided to heterogeneous sensor data transmitted from multiple physical sensors installed on-site at the vibrating screening equipment. This data is processed to generate a standardized sequence of operating parameters. A set of performance target parameters set for the target screening production line in the current evaluation cycle is extracted from the system configuration library. The standardized operating parameter sequence, the simulation logical topology structure, and the performance target parameter set are input into a pre-trained multimodal simulation prediction model. In this process, the simulation logical topology structure serves as a constraint on model operation, while the performance target parameter set serves as the expected convergence condition for model deduction. The multimodal simulation prediction model is activated, allowing it to dynamically deduce the input standardized operating parameter sequence under the combined effect of the constraints and the expected convergence condition. Ultimately, it outputs a simulation trajectory describing the entire future state of the target screening production line from the current moment to the end of the evaluation cycle. The simulation trajectory is analyzed to separate the set of predicted key performance indicators (KPIs) at the end of the evaluation period. This set of predicted values ​​is compared with the previously extracted set of performance target parameters to calculate a multi-dimensional performance deviation metric set. The KPI predicted values, performance target parameters, and multi-dimensional performance deviation metric set are then integrated to generate a complete evaluation results report.

[0022] In one embodiment of the present invention, heterogeneous sensor data transmitted from multiple physical sensors installed on-site at the vibrating screening equipment are accessed in real time to generate a standardized operating parameter sequence. The process includes: the physical connection topology reflects the flow direction of materials in the actual production line; the accessed heterogeneous sensor data mainly includes vibration state data and material flow data. Heterogeneous data assimilation processing is performed on the real-time accessed heterogeneous sensor data to generate a standardized operating parameter sequence with a unified timestamp and data format. The specific steps of the heterogeneous data assimilation processing are: configuring a dedicated data parsing adapter for each type of heterogeneous sensor data, which extracts valid numerical fields directly related to the screening equipment status from its corresponding raw data stream; initiating a global time synchronization service, which provides a reference clock signal for the entire system based on network time synchronization signals; and marking each valid numerical field output by the data parsing adapter with a unified timestamp provided by the global time synchronization service, thereby forming raw data points with time identifiers. All raw data points with time stamps are imported into a unified data format converter. This converter linearly maps raw data points of different dimensions and numerical ranges to a standard numerical range between zero and one, generating standardized data points. All standardized data points falling within a preset time alignment window are aligned and packaged according to their timestamps, generating standardized runtime parameter frames containing multiple parameter types with strictly aligned timestamps. Arranging consecutive standardized runtime parameter frames in chronological order constitutes the required standardized runtime parameter sequence.

[0023] Extracting the set of performance target parameters for the target screening production line for the current evaluation period from the system configuration library includes the following steps: (See below) Figure 2 The system receives an evaluation cycle definition command submitted by the user through an interactive interface. This command includes the start and end times of the evaluation cycle. Based on the evaluation cycle definition command, the system searches the system configuration library for the various performance expectation values ​​pre-set for the target screening production line within the evaluation cycle timeframe. If multiple independent performance expectation values ​​are found, they are categorized according to their defined performance index types to form a preliminary performance target subset. The system reads the historical operation records of the target screening production line and extracts the actual average performance values ​​of the same historical period that are similar to the current evaluation cycle in terms of season and production plan. Each performance expectation value in the preliminary performance target subset is compared with the corresponding actual average performance value of the same historical period. When the comparison finds that a performance expectation value deviates from the corresponding actual average performance value of the same historical period, and the deviation exceeds a preset reasonableness threshold, a configuration anomaly warning is generated. If no configuration anomaly warning is generated, or if the generated configuration anomaly warning is ignored by the user, the preliminary performance target subset is formally determined as the performance target parameter set for the current simulation prediction.

[0024] In practical implementation, real-time data access and processing, as well as the extraction of performance target parameters, are illustrated in a specific scenario: a mineral processing production line containing two series-connected vibrating screens, the first being a feed screen and the second a finished product screen. The physical connection topology clearly reflects the direction of material flow from the feed screen to the finished product screen. Physical sensors installed on the feed screen transmit vibration acceleration and motor current data, while physical sensors installed on the finished product screen transmit instantaneous flow rate data of the material under the screen; these together constitute heterogeneous sensor data. To process the vibration acceleration data, a dedicated data parsing adapter is configured, which extracts the effective acceleration value and peak frequency as valid numerical fields from the raw data stream. To process the motor current data, another dedicated data parsing adapter is configured, which extracts the root mean square value of the current as a valid numerical field. To process the instantaneous flow rate data of the material, a third dedicated data parsing adapter is configured, which extracts the volumetric flow rate value as a valid numerical field. A global time synchronization service is initiated, which generates a reference clock signal with millisecond-level accuracy based on the network time synchronization signal. Each data parsing adapter immediately marks a valid numeric field with a precise timestamp provided by the global time synchronization service when it outputs the field, forming a raw data point with a time identifier.

[0025] In practice, all raw data points with time stamps are imported into a unified data format converter. This converter pre-stores the dimensions and preset numerical ranges for each parameter. For the effective value of vibration acceleration, the preset range is 0 to 10 m / s². A raw data point of 5.2 m / s² is linearly mapped, and the calculation formula is as follows:

[0026] in: Represents the standardized data point values. Represents the original data point values. This represents the minimum value within the preset range of values ​​for this parameter. This represents the maximum value within the preset range of this parameter. Substituting 5.2 m / s², the standardized data point value is calculated to be 0.52. For the root mean square value of the motor current, the preset range is 20A to 100A; a raw data point of 60A is converted to a standardized data point value of 0.5. For the material volumetric flow rate, the preset range is 0 m³ / h to 50 m³ / h; a raw data point of 25 m³ / h is converted to a standardized data point value of 0.5. All standardized data points within the preset 100-millisecond time alignment window are aligned according to their timestamps. Multiple standardized data points under the same timestamp are packaged to generate a standardized operating parameter frame containing various parameter types such as acceleration, current, and flow rate. Continuous standardized operating parameter frames are arranged in chronological order to form a standardized operating parameter sequence.

[0027] In one embodiment of the present invention, a multimodal simulation prediction model is activated, enabling it to dynamically extrapolate a standardized operating parameter sequence under the combined effects of constraints and desired convergence conditions. This process includes the following progressive steps: using the standardized operating parameter sequence at the current moment as the initial state input of the multimodal simulation prediction model. The multimodal simulation prediction model internally constructs a neural network computation graph isomorphic to the simulation logic topology, which is used to simulate the flow and screening transformation of materials within the physical connection topology. Within each model's extrapolation step, the simulated state parameters of each vibrating screening device calculated in the previous step are transmitted to its downstream connected device nodes as simulated input parameters, according to the flow direction defined by the simulation logic topology. At each device node, the received simulated input parameters are processed through a dedicated device simulation sub-network, and combined with the desired constraints imposed by the performance target parameter set, the simulated output parameters and simulated energy consumption parameters of the device node at the current step are calculated. The simulated output parameters of all device nodes at the current step are collected and summarized into a snapshot of the overall state simulation of the target screening production line at the current step. Repeat the above steps of transmission, calculation and summarization until the model's internal extrapolation time reaches the end of the current evaluation cycle. All the overall state simulation snapshots generated during this process are connected in chronological order to form the full-process future state simulation trajectory.

[0028] In practical implementation, a multimodal simulation prediction model is activated for dynamic deduction. A specific example scenario is a target screening production line containing vibrating screening equipment 1 and vibrating screening equipment 2. The simulation logic topology clearly indicates that the material flows from vibrating screening equipment 1 to vibrating screening equipment 2. The standardized operating parameter sequence at the current moment is used as the initial state input of the multimodal simulation prediction model. The standardized operating parameter sequence includes the vibration frequency parameter of vibrating screening equipment 1 (0.52), the amplitude parameter (0.61), and the feed flow rate parameter of vibrating screening equipment 2 (0.73). The performance target parameter set includes the screening efficiency target value (0.95) and the throughput target value (0.80). The multimodal simulation prediction model contains a neural network computation graph isomorphic to the simulation logic topology. The neural network computation graph has two nodes corresponding to vibrating screening equipment 1 and vibrating screening equipment 2 respectively. The connection direction between the nodes is consistent with the physical connection topology. The neural network computation graph simulates the transformation process of material flowing from vibrating screening equipment 1 to vibrating screening equipment 2 after screening.

[0029] In some embodiments, within each model's internal simulation step, the simulated state parameters of each vibrating screen device calculated in the previous step are passed to its downstream connected device nodes as simulated input parameters according to the flow direction defined by the simulation logic topology. For example, in the initial step, the simulated state parameters of vibrating screen device one are derived from the standardized operating parameter sequence. After calculation, the simulated output parameters of vibrating screen device one are obtained as the material particle size distribution vector [0.15, 0.25, 0.60]. This material particle size distribution vector is passed to vibrating screen device two as simulated input parameters according to the flow direction. In a specific implementation, at each device node, the received simulated input parameters are processed by a dedicated device simulation sub-network. Combined with the expected constraints imposed by the performance target parameter set, the simulated output parameters and simulated energy consumption parameters of the device node at the current step are calculated. For vibrating screen device two, the dedicated device simulation sub-network receives the simulated input parameter, i.e., the material particle size distribution vector. The screening efficiency target value in the performance target parameter set is used as the expected constraint to influence the calculation. The calculation process of the dedicated device simulation sub-network is described as follows:

[0030] in: This represents the simulated output parameter vector of the device node at the nth step. Represents a non-linear activation function. The weight matrix represents the device-specific analog subnetwork. This represents the vector of analog input parameters passed to the device node at the nth step size. The bias vector representing the device-specific analog subnetwork. The coupling coefficient represents the desired constraint. This represents a vector of target values ​​related to the current device node within the set of performance target parameters. This represents the simulated output parameter vector of the device node at the (n-1)th step.

[0031] In some embodiments, the simulated output parameters of all device nodes at the current step size are collected and summarized into a snapshot of the overall state of the target screening production line at the current step size. For example, at step size n, the simulated output parameter vectors of vibrating screening device one [0.15, 0.25, 0.60] and vibrating screening device two [0.08, 0.12, 0.80] are collected and summarized, and the total system processing capacity parameter 0.75 is added to form an overall state simulation snapshot {device one output: [0.15, 0.25, 0.60], device two output: [0.08, 0.12, 0.80], total processing capacity: 0.75}. In practice, the steps of transmission, calculation and summarization are repeated until the model's internal simulation time reaches the end of the current evaluation period. Assuming the evaluation period is 4 hours in the future and the model's internal simulation step size is set to 5 minutes, the transmission, calculation and summarization steps are repeated 48 times. All the overall state simulation snapshots generated in this process are connected in chronological order to form the full-process future state simulation trajectory. The full-process future state simulation trajectory is a time series array containing 48 overall state simulation snapshots.

[0032] In one embodiment of the present invention, an online correction mechanism for the simulated state is introduced in the progressive steps of dynamic inference. The specific steps are as follows: A series of synchronization checkpoints are set based on the correspondence between the internal inference time of the multimodal simulation prediction model and the actual system time. Whenever the internal inference time of the model reaches a synchronization checkpoint, the system pauses the model inference and obtains the latest standardized operating parameter frame corresponding to the current actual system time from the real-time data access stream as the true state observation value. The simulated state snapshot generated by the model at the current synchronization checkpoint is compared with the true state observation value at the same point to calculate the state parameter difference vector. The state parameter difference vector calculation step specifically includes: performing arithmetic subtraction on each type of state parameter of the same dimension contained in the true state observation value and the simulated state snapshot to obtain a series of original differences. A normalization coefficient is preset for each type of state parameter, and the original difference of each type of state parameter is multiplied by its corresponding normalization coefficient to convert it into a dimensionless standardized difference. The standardized differences corresponding to all state parameters are weighted and combined according to a preset parameter importance weight vector to obtain a scalarized comprehensive state difference degree. Only when the overall state difference exceeds a preset difference tolerance threshold will the state parameter difference vector and the standardized differences of the corresponding state parameters be transmitted to the state correction module for processing. Based on the received state parameter difference vector, the state correction module dynamically adjusts the internal state parameters upon which the model relies for subsequent inferences, correcting the generated simulated state snapshots towards the actual observations before continuing inferences for subsequent steps.

[0033] In practical implementation, an online correction mechanism for the simulated state is introduced during the dynamic simulation process. A specific example scenario is that a multimodal simulation prediction model is simulating the next two hours of a production line containing a primary vibrating screen and a secondary vibrating screen. The simulation logic topology defines the material flow from the primary vibrating screen to the secondary screening equipment. A series of synchronization checkpoints are set to correlate the internal simulation time of the multimodal simulation prediction model with the actual system time. The time interval between these checkpoints is fifteen minutes, meaning that when the internal simulation time reaches fifteen, thirty, forty-five minutes after startup, the synchronization check process will be triggered.

[0034] In practice, whenever the model's internal simulation time reaches a synchronization checkpoint, the system pauses the model simulation and retrieves the latest standardized operating parameter frame corresponding to the current system time from the real-time data access stream as the true state observation value. For example, when the model's internal simulation time reaches a synchronization checkpoint of 45 minutes, the system pauses the simulation and retrieves a standardized operating parameter frame with a timestamp of the system's actual time from the data stream. The true state observation value includes the primary vibrating screen vibration intensity parameter value of 0.68, the secondary screening equipment feed flow rate parameter value of 0.72, and the system total power parameter value of 0.61. Simultaneously, the simulated state snapshot generated by the model at the current synchronization checkpoint includes the corresponding simulated vibration intensity parameter value of 0.65, simulated feed flow rate parameter value of 0.70, and simulated total power parameter value of 0.58. The simulated state snapshot generated by the model at the current synchronization checkpoint is compared with the true state observation value at the same point, and the state parameter difference vector is calculated. The specific steps for calculating the state parameter difference vector include: performing arithmetic subtraction on each type of state parameter of the same dimension contained in the actual state observation and the simulated state snapshot to obtain a series of original differences. The original difference of vibration intensity is 0.68-0.65=0.03, the original difference of feed flow rate is 0.72-0.70=0.02, and the original difference of total power is 0.61-0.58=0.03.

[0035] In some embodiments, a normalization coefficient is preset for each type of state parameter. This normalization coefficient is used to convert the original differences between different dimensions or numerical ranges into dimensionless standardized differences. The normalization coefficient for vibration intensity is set to 2.0, for feed flow rate to 2.5, and for total power to 3.0. The original difference of each type of state parameter is multiplied by its corresponding normalization coefficient to convert it into dimensionless standardized differences. The standardized difference for vibration intensity is 0.03 * 2.0 = 0.06, for feed flow rate it is 0.02 * 2.5 = 0.05, and for total power it is 0.03 * 3.0 = 0.09. The standardized differences corresponding to all state parameters are weighted according to a preset parameter importance weight vector to obtain a scalarized comprehensive state variability. In the parameter importance weight vector, vibration intensity has a weight of 0.4, feed flow rate has a weight of 0.3, and total power has a weight of 0.3. The formula for calculating the comprehensive state variability is:

[0036] in: Represents the degree of difference in overall status. The preset weights represent the parameters of the i-th type of state. The standardized difference represents the parameter of the i-th type of state. The total number of types representing state parameters. Substituting the values ​​into the calculation yields the overall state difference. See Table 1.

[0037] Table 1: Example Table of Comparison and Difference Calculation of Status Parameters of Synchronous Checkpoints State parameter categories Real state observations Simulated state snapshot value Original difference Normalization coefficient Standardized difference Preset weights Weighted contribution Vibration intensity 0.68 0.65 0.03 2.0 0.06 0.4 0.024 Feed flow rate 0.72 0.70 0.02 2.5 0.05 0.3 0.015 Total power 0.61 0.58 0.03 3.0 0.09 0.3 0.027 Overall state difference 0.066 It is understandable that the system's preset difference tolerance threshold is set to, for example, 0.10. The calculated overall state difference of 0.066 does not exceed the difference tolerance threshold of 0.10. Therefore, at this synchronization checkpoint, the system does not pass the state parameter difference vector to the state correction module for processing, but directly continues to execute the derivation of subsequent steps. Optionally, in another data comparison scenario, assuming the calculated overall state difference is 0.15, exceeding the difference tolerance threshold of 0.10, the system will pass the state parameter difference vector [vibration intensity difference: 0.06, feed flow rate difference: 0.05, total power difference: 0.09] and its corresponding standardized differences for various state parameters to the state correction module for processing. The state correction module dynamically adjusts the internal state parameters on which the model depends for subsequent inferences based on the received state parameter difference vector. The state correction module contains a feedback gain matrix, which maps the state parameter difference vector to the adjustment amount of the hidden state vector of the neural network computation graph in the multimodal simulation prediction model, so that the subsequently generated simulation state snapshot is corrected in the direction of the real observation value. After the adjustment is completed, the system continues to execute the inference of the subsequent step size.

[0038] See Figure 3 This is a full-cycle comprehensive state difference monitoring chart. The comprehensive difference remains stable within the range of 0.055 to 0.066, consistently below the tolerance threshold of 0.1, indicating high accuracy in this simulation and eliminating the need for state correction. The curve shows a trend of first rising and then falling with slight fluctuations. The difference increases slowly from 0 to 45 minutes, reaching a peak of 0.066 at 45 minutes; it falls rapidly from 45 to 60 minutes; it rises slightly again from 60 to 90 minutes, reaching a secondary peak of 0.064 at 90 minutes; and it continues to decline from 90 to 120 minutes, eventually converging to 0.055, indicating a gradual improvement in simulation accuracy. The threshold line clearly defines the triggering condition for online correction. The state correction module is only activated when the difference exceeds 0.1, adjusting the model's internal parameters to avoid ineffective correction. As a core monitoring indicator, it evaluates the effectiveness of the simulation in real time, ensuring prediction accuracy.

[0039] In one embodiment of the present invention, the simulation trajectory of the entire future state is analyzed to separate a set of predicted key performance indicators (KPIs) for the end point of the current evaluation cycle. The steps are as follows: A predefined KPI extraction rule base is loaded, which defines the calculation logic of various KPIs and their parameter dependencies in the simulation trajectory of the entire future state. A snapshot of the final state simulation corresponding to the end time of the current evaluation cycle is extracted from the simulation trajectory. The parameters in the final state simulation snapshot are substituted into the calculation logic defined in the KPI extraction rule base for calculation. A set of calculated numerical results is output as the set of predicted KPIs. The calculation of the set of predicted KPIs also includes a reference to historical simulation trajectory trends. Specifically, while extracting the final state simulation snapshot, a trajectory segment of a fixed time length before the end time is extracted from the simulation trajectory of the entire future state as the pre-terminal state trajectory. Statistical analysis is performed on the pre-terminal state trajectory to calculate the rate of change of the main state parameters within the fixed time length. Specific performance indicators whose calculation logic is affected by the rate of change of parameters are identified from the KPI extraction rule base. For these specific performance indicators, the calculation results based on the final state simulation snapshot are fused with the rate of change of the state parameters corresponding to the specific performance indicator. The original calculation results based on the final state simulation snapshot are replaced with the corrected values ​​after fusion calculation, thereby updating the set of predicted values ​​for key performance indicators.

[0040] In practical implementation, the simulation trajectory of the entire process's future state is analyzed to obtain a set of predicted values ​​for key performance indicators (KPIs). A specific example scenario is a ore screening production line containing three series-connected vibrating screens. The simulation trajectory of the entire process's future state has been generated by a multimodal simulation prediction model, with the evaluation cycle ending at the eighth hour in the future. A predefined KPI extraction rule base is loaded, stored in a structured data table format. This rule base defines the calculation logic for three key performance indicators: "screening efficiency," "total throughput," and "average energy consumption per ton of product," as well as their parameter dependencies in the simulation trajectory of the entire process's future state. In practical implementation, a snapshot of the final state simulation at the corresponding evaluation cycle's end time is extracted from the simulation trajectory of the entire process's future state. The final state simulation snapshot is a data structure containing multi-dimensional parameters, such as the flow rate of material undersize from screen #1 (0.45), screen #2 (0.30), raw material feed flow rate (0.85), final output material flow rate (0.72), instantaneous motor power for screen #1 (0.65), screen #2 (0.60), and screen #3 (0.55). The parameters in the final state simulation snapshot are then substituted into the calculation logic defined in the key performance indicator extraction rule base for calculation. For the "screening efficiency" indicator, the calculation is (0.45 + 0.30) / 0.85 = 0.882. For the "total throughput" indicator, the final output material flow rate of 0.72 is directly used. For the "average energy consumption per ton of product" indicator, the power parameter integral required in the calculation logic is obtained by numerical integration from the power parameter time series corresponding to the full-process future state simulation trajectory. Assuming the integral result is 4.2, the "average energy consumption per ton of product" is calculated as 4.2 / 0.72 = 5.833. The calculated set of numerical results is output as the set of predicted values ​​for key performance indicators. The initial set is {screening efficiency: 0.882, total throughput: 0.72, average energy consumption per ton of product: 5.833}.

[0041] In some embodiments, the calculation of the set of predicted key performance indicators (KPIs) also includes referencing historical simulation trajectory trends. While capturing a snapshot of the final state simulation, a trajectory segment of a fixed time length prior to the termination time is extracted from the full-process future state simulation trajectory as the pre-termination state trajectory. This fixed time length is, for example, set to the last forty minutes of the trajectory. Statistical analysis is performed on the pre-termination state trajectory to calculate the rate of change of the main state parameters within the fixed time length. By analyzing the pre-termination state trajectory, the rate of change of the "final output material flow rate parameter" in the last forty minutes is calculated to be +0.002 per minute, and the rate of change of the "total system power parameter" (the sum of the instantaneous power of the three screens) in the last forty minutes is -0.001 per minute. From the KPI extraction rule base, specific performance indicators whose calculation logic is affected by the parameter change rate are identified. The "average energy consumption per ton of product" is marked as a specific performance indicator affected by the change rate in the KPI extraction rule base, and its correction logic integrates the power value at the final moment with the power change trend. For the specific performance indicator "average energy consumption per ton of product", its calculation result based on the final state simulation snapshot is fused with the rate of change of the state parameters corresponding to the specific performance indicator. The formula used for the fusion calculation is:

[0042] in: The correction value representing "average energy consumption per ton of product" The power parameter represents the integral value over the evaluation period. The correction factor representing the rate of change of power. The rate of change representing the "total system power parameter" This represents the estimated time span used for trend extrapolation. The predicted "total processing volume" based on the final state simulation snapshot is 0.72. The correction factor representing the rate of change in flow rate. This represents the rate of change of the "final output material flow rate parameter". Assume the parameter is set as follows: , , (minutes), substitute into the calculation to obtain the correction value. The original calculation result of 5.833 based on the final state simulation snapshot was replaced with the corrected value of 4.020 after fusion calculation, thereby updating the set of predicted values ​​for key performance indicators. The updated set is {screening efficiency: 0.882, total throughput: 0.72, average energy consumption per ton of product: 4.020}. See Table 2.

[0043] Table 2: Example Table of Key Performance Indicator Calculation and Trend Correction

[0044] It is understandable that the fixed time length can be set proportionally based on the total length of the evaluation cycle. For example, the last tenth of the evaluation cycle could be selected as the analysis window for the state trajectory before termination. The parameter change rate can be calculated by fitting the parameter sequence in the state trajectory before termination using linear regression to obtain the slope as the change rate. Optionally, the key performance indicator extraction rule base can define various indicators affected by the change rate and their specific fusion calculation functions. For example, for the "average equipment load rate" indicator, its correction may depend on the changing trend of the vibration intensity parameter. The correction coefficient in the fusion calculation formula... , and estimated time span The rationality of trend extrapolation can be ensured by fitting historical data. In some embodiments, if the calculated rate of change of parameters is close to zero, the fusion calculation can be skipped, and the calculation result based on the final state simulation snapshot can be used directly as the output to simplify the process. Optionally, the analysis of the state trajectory before termination can not only calculate the rate of change, but also perform more complex statistical analyses, such as calculating the variance of parameter fluctuations and introducing the variance of fluctuations as one of the correction factors into the calculation logic of specific performance indicators. It can be understood that introducing the reference of historical simulation trajectory trends allows the predicted values ​​of key performance indicators to reflect the dynamic change trend of the system at the end of the evaluation period, rather than just the static final value. This helps to capture the delayed impact of inertia or gradual change processes on performance indicators.

[0045] See Figure 4 This is a dual-axis trend chart from the simulation stage of an industrial vibrating screen production line, used to illustrate the dynamic changes in material flow rate and system power. The two curves intersect at the 20-minute mark, reflecting the dynamic equilibrium point between flow rate and power. The material flow rate linearly increases from 0.64 to 0.72, a 12.5% ​​increase within 40 minutes, demonstrating the gradual increase in production line capacity. The total system power linearly decreases from 1.840kW to 1.800kW, a 2.17% decrease within 40 minutes, reflecting the effect of system energy efficiency optimization. The two curves show a perfect negative correlation, verifying the optimization objective of "increased capacity and reduced energy consumption" in the simulation model. The decreasing trend of total system power reflects the effect of the "dedicated equipment simulation sub-network + energy consumption parameter calculation" in the model, verifying the energy efficiency optimization of the production line during operation. Simultaneously, quantifying capacity increase and energy consumption reduction provides a visual representation of the production line's operating efficiency.

[0046] In one embodiment of the present invention, a multi-dimensional performance deviation measurement set is calculated. The process includes: matching a pair of predicted values ​​and target values ​​for the same performance indicator from a set of predicted values ​​for key performance indicators and a set of performance target parameters. For each pair of predicted and target values, the absolute deviation of the predicted value relative to the target value is calculated. A preset deviation range benchmark value for the performance indicator is read from the system configuration library. The absolute deviation is divided by the deviation range benchmark value to obtain a dimensionless relative deviation coefficient. The relative deviation coefficients corresponding to all performance indicators are organized according to their indicator categories to form the multi-dimensional performance deviation measurement set, where each dimension corresponds to a relative deviation coefficient for one performance indicator. The method also includes a deep analysis step of the performance deviation measurement set: establishing a performance indicator correlation graph, which describes the mutual influence relationships and influence weights between different performance indicators. The multi-dimensional performance deviation measurement set is input into the performance indicator correlation graph. The performance indicator correlation graph calculates the relative deviation coefficient of each dimension and the indirect influence value on other related dimensions based on its internally defined mutual influence relationships and influence weights. The direct relative deviation coefficients of each dimension, along with the indirect impact values ​​passed down from all other dimensions, are summarized to obtain the comprehensive influence strength of each dimension. All performance indicator dimensions are then sorted in descending order of comprehensive influence strength, generating a key deviation dimension ranking list. This key deviation dimension ranking list is integrated with the multi-dimensional performance deviation measurement set, the key performance indicator predicted value set, and the performance target parameter set to generate a final evaluation result containing in-depth analytical information.

[0047] In practical implementation, this involves calculating and deeply analyzing a set of performance deviation metrics. A specific example scenario is that the set of predicted key performance indicators includes a predicted screening efficiency of 88.0%, a predicted throughput of 0.72, and a predicted energy consumption per ton of product of 5.5 kWh / ton. The set of performance target parameters includes a target screening efficiency of 95.0%, a target throughput of 0.80, and a target energy consumption per ton of product of 5.0 kWh / ton. From the set of predicted key performance indicators and the set of performance target parameters, pairs of predicted and target values ​​for the same performance indicator are matched one by one. For example, the predicted screening efficiency of 88.0% and the target screening efficiency of 95.0% are matched as one pair, the predicted throughput of 0.72 and the target throughput of 0.80 are matched as one pair, and the predicted energy consumption per ton of product of 5.5 and the target energy consumption per ton of product of 5.0 are matched as one pair.

[0048] In practice, for each pair of predicted and target values, the absolute deviation of the predicted value relative to the target value is calculated. The absolute deviation of screening efficiency is |88.0%-95.0%|=7.0%, the absolute deviation of throughput is |0.72-0.80|=0.08, and the absolute deviation of energy consumption per ton of product is |5.5-5.0|=0.5 kWh / ton. From the system configuration library, the preset deviation range benchmark value for each performance indicator is read: the deviation range benchmark value for screening efficiency is 20.0%, for throughput is 0.40, and for energy consumption per ton of product is 2.0 kWh / ton. Dividing the absolute deviation by the deviation range benchmark value yields a dimensionless relative deviation coefficient. The relative deviation coefficient for screening efficiency is 7.0% / 20.0%=0.35, the relative deviation coefficient for throughput is 0.08 / 0.40=0.20, and the relative deviation coefficient for energy consumption per ton of product is 0.5 / 2.0=0.25. All the relative deviation coefficients corresponding to the performance indicators are organized according to their indicator categories to form a multi-dimensional performance deviation measurement set {screening efficiency deviation: 0.35, throughput deviation: 0.20, energy consumption per ton of product deviation: 0.25}, where each dimension corresponds to the relative deviation coefficient of one performance indicator.

[0049] In some embodiments, the method further includes a deep analysis step of the performance deviation metric set to establish a performance index correlation graph. The performance index correlation graph describes the mutual influence relationships and influence weights between different performance indicators in a directed graph structure. In a specific example, the performance index correlation graph defines that screening efficiency has a positive impact on throughput with a weight of 0.3; screening efficiency has a negative impact on energy consumption per ton of product with a weight of 0.4; and throughput has a positive impact on energy consumption per ton of product with a weight of 0.5. The multi-dimensional performance deviation metric set {screening efficiency deviation: 0.35, throughput deviation: 0.20, energy consumption per ton of product deviation: 0.25} is input into the performance index correlation graph. Based on its internally defined mutual influence relationships and influence weights, the performance index correlation graph calculates the indirect influence value of the relative deviation coefficient of each dimension on other related dimensions. For the throughput dimension, it is affected by the screening efficiency dimension; the screening efficiency deviation of 0.35, through an influence weight of 0.3, produces an indirect influence value of 0.35 * 0.3 = 0.105 on the throughput dimension. For the energy consumption per ton of product, it is affected by both the screening efficiency and the throughput. A screening efficiency deviation of 0.35 has an indirect impact of 0.35 * 0.4 = 0.140 on the energy consumption per ton of product through an influence weight of 0.4. A throughput deviation of 0.20 has an indirect impact of 0.20 * 0.5 = 0.100 on the energy consumption per ton of product through an influence weight of 0.5.

[0050] In practice, the direct relative deviation coefficient of each dimension, along with the indirect impact values ​​transmitted from all other dimensions, are summed to obtain the comprehensive impact strength of each dimension. The formula for calculating the comprehensive impact strength is:

[0051] in: This represents the overall influence strength of the i-th dimension. Represents the direct relative deviation coefficient of the i-th dimension. The influence weights from dimension j to dimension i in the performance indicator correlation graph. This represents the direct relative deviation coefficient of dimension j. For the screening efficiency dimension, its direct relative deviation coefficient is 0.35. In the performance index correlation graph example, no other dimensions directly affect screening efficiency; therefore, the overall influence strength of the screening efficiency dimension is... For the throughput dimension, the direct relative deviation coefficient is 0.20, influenced by the screening efficiency dimension; therefore, the overall impact of the throughput dimension is significant. Regarding energy consumption per ton of product, the direct relative deviation coefficient is 0.25. It is also affected by both screening efficiency and throughput dimensions; therefore, the overall impact of energy consumption per ton of product is significant. It is understandable that, according to the order of comprehensive impact intensity from high to low, all performance index dimensions are sorted. The comprehensive impact intensity of energy consumption per ton of product is the highest at 0.49, followed by screening efficiency at 0.35, and the comprehensive impact intensity of throughput is the lowest at 0.305, generating a key deviation dimension ranking list [energy consumption per ton of product, screening efficiency, throughput].

[0052] See Figure 5 This is a performance deviation measurement calculation chart for an industrial vibrating screen production line. The predicted screening efficiency is slightly lower than the target value, with a deviation of approximately 5.4%, but overall close to the target. The predicted throughput is slightly lower than the target value, with a deviation of approximately 16.7%, making it the indicator with the largest deviation. The predicted energy consumption per ton of product is slightly higher than the target value, with a deviation of approximately 9.1%, indicating slightly higher-than-expected energy consumption. It visually presents the achievement status of the three core KPIs, representing a core visualization result of production line performance evaluation. It intuitively displays the gap between the predicted and target values ​​of each core indicator, quickly identifying non-compliant items. Combined with the performance indicator correlation graph, the transmission impact and root causes of deviations can be further analyzed. It provides data support for adjusting equipment parameters and optimizing operating strategies, specifically increasing throughput and reducing energy consumption. The visualization presents the core effects of simulation prediction and performance evaluation, intuitively demonstrating the technological value.

[0053] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for data acquisition, evaluation, and analysis of industrial vibrating screening based on simulation prediction, characterized in that, include: Construct a simulation logic topology structure that includes all vibrating screening equipment and their physical connection topology in the target screening production line; The system receives heterogeneous sensor data transmitted from multiple physical sensors installed on-site at the vibrating screening equipment in real time, and generates a standardized operating parameter sequence. Extract the set of performance target parameters for the target screening production line for the current evaluation period from the system configuration library; The standardized sequence of operating parameters, the simulation logic topology, and the set of performance target parameters are input into the trained multimodal simulation prediction model, where the simulation logic topology serves as a constraint and the set of performance target parameters serves as the expected convergence condition. The multimodal simulation prediction model is activated, and under the combined effect of constraints and expected convergence conditions, it dynamically extrapolates the standardized operating parameter sequence and outputs the simulation trajectory of the entire future state of the target screening production line. The simulation trajectory of the future state of the entire process is analyzed to separate the set of predicted values ​​of key performance indicators for the end point of the current evaluation cycle. The set of predicted values ​​of key performance indicators is compared with the set of performance target parameters to calculate a multi-dimensional set of performance deviation measures. The evaluation result is generated by integrating the set of predicted values ​​of key performance indicators, the set of performance target parameters, and the set of multi-dimensional performance deviation measures.

2. The method for data acquisition, evaluation, and analysis of industrial vibrating screening based on simulation prediction according to claim 1, characterized in that, The system receives heterogeneous sensor data from multiple physical sensors installed on-site at the vibrating screening equipment in real time, and generates a standardized sequence of operating parameters, including: The physical connection topology reflects the flow direction of materials in the actual production line; The heterogeneous sensor data includes vibration status data and material flow rate data; The heterogeneous sensor data that is accessed in real time is subjected to heterogeneous data assimilation processing to generate a standardized sequence of operating parameters with a unified timestamp and a unified data format. The heterogeneous data assimilation process for the real-time accessed heterogeneous sensor data specifically includes: A dedicated data parsing adapter is configured for each type of heterogeneous sensor data, and the data parsing adapter extracts valid numerical fields that are directly related to the state of the screening equipment from the raw data stream; Start a global time synchronization service, which provides a reference clock signal for the entire system based on the network time synchronization signal; Each valid numerical field output by the data parsing adapter is marked with a timestamp provided by the global time synchronization service to form a raw data point with a time identifier; All raw data points with time stamps are imported into a unified data format converter. The data format converter linearly maps raw data points with different dimensions and different numerical ranges to a standard numerical range between zero and one, generating standardized data points. All standardized data points within the preset time alignment window are aligned and packaged according to their timestamps to generate a standardized runtime parameter frame containing multiple parameter types and with strictly aligned timestamps. The standardized operating parameter sequence is formed by arranging consecutive standardized operating parameter frames in chronological order.

3. The method for data acquisition, evaluation, and analysis of industrial vibrating screening based on simulation prediction according to claim 2, characterized in that, The step of extracting the set of performance target parameters for the target screening production line for the current evaluation period from the system configuration library includes: Receive the evaluation period definition instruction submitted by the user through the interactive interface, wherein the evaluation period definition instruction includes the start time and end time of the evaluation period; According to the evaluation cycle definition instruction, a search is performed in the system configuration library to find the various performance expectations that are pre-set for the target screening production line within the time range of the evaluation cycle; If multiple independent performance expectations are retrieved, the performance expectations are categorized according to their defined performance index types to form a preliminary subset of performance targets. Read the historical operation records of the target screening production line, and extract the actual average performance value of the same historical period that is similar to the evaluation cycle in terms of season and production plan from the historical operation records; Each expected performance value in the preliminary performance target subset is compared with the corresponding actual average performance value for the same historical period. When a comparison reveals that a certain performance expectation deviates from the actual average performance value of the corresponding historical period, and the degree of deviation exceeds the preset reasonableness threshold, a configuration anomaly warning will be generated. If no configuration anomaly warning is generated, or if the generated configuration anomaly warning is confirmed to be ignored by the user, the preliminary performance target subset will be formally determined as the set of performance target parameters used for the current simulation prediction.

4. The method for data acquisition, evaluation, and analysis of industrial vibrating screening based on simulation prediction according to claim 3, characterized in that, Activating the multimodal simulation prediction model, and enabling it to dynamically extrapolate the standardized operating parameter sequence under the combined effect of constraints and desired convergence conditions, includes the following progressive steps: The standardized operating parameter sequence at the current moment is used as the initial state input of the multimodal simulation prediction model; The multimodal simulation prediction model contains a neural network computation graph that is isomorphic to the simulation logic topology. The neural network computation graph simulates the flow and sieving transformation of materials in the physical connection topology. Within each model's internal simulation step, the simulated state parameters of each vibrating screening device calculated in the previous step are transmitted to the downstream connected device nodes as simulated input parameters, according to the flow direction defined by the simulation logic topology. At each device node, the received analog input parameters are processed through a dedicated analog sub-network, and the expected constraints imposed by the set of performance target parameters are combined to calculate the analog output parameters and analog energy consumption parameters of the device node at the current step size. Collect the simulated output parameters of all device nodes at the current step size, and summarize them into a snapshot of the overall state of the target screening production line at the current step size. The steps of transmission, calculation and summarization are repeated until the model's internal simulation time reaches the end of the current evaluation cycle. All overall state simulation snapshots generated during this process are connected in chronological order to form the full-process future state simulation trajectory.

5. The method for data acquisition, evaluation, and analysis of industrial vibrating screening based on simulation prediction according to claim 4, characterized in that, The progressive steps of dynamic simulation also include an online correction mechanism for the simulated state, the specific steps of which are as follows: A series of synchronization checkpoints are set to establish the correspondence between the internal extrapolation time of the multimodal simulation prediction model and the actual system time. Whenever the model's internal simulation time reaches a synchronization checkpoint, the system pauses the model simulation and obtains the latest standardized operating parameter frame corresponding to the current system's actual time from the real-time data access stream as the real-state observation value. The simulated state snapshot generated by the model at the current synchronization checkpoint is compared with the actual state observation at the same point, and the state parameter difference vector is calculated. The state parameter difference vector is fed back into the state correction module of the multimodal simulation prediction model; The state correction module dynamically adjusts the internal state parameters on which the model depends for subsequent inferences based on the state parameter difference vector, so that the subsequently generated simulated state snapshots are corrected in the direction of the actual observations, and then the inference of subsequent steps continues.

6. The method for data acquisition, evaluation, and analysis of industrial vibrating screening based on simulation prediction according to claim 5, characterized in that, The step of calculating the state parameter difference vector specifically includes: For each type of state parameter of the same dimension contained in the real state observation and the simulated state snapshot, perform arithmetic subtraction to obtain a series of original differences; A normalization coefficient is preset for each type of state parameter. The original difference of each type of state parameter is multiplied by its corresponding normalization coefficient and converted into a dimensionless standardized difference. The standardized differences corresponding to all state parameters are weighted and combined according to the preset parameter importance weight vector to obtain a scalarized comprehensive state difference. Only when the overall state difference exceeds the preset difference tolerance threshold will the state parameter difference vector and the standardized differences of the corresponding state parameters be transmitted to the state correction module for processing.

7. The method for data acquisition, evaluation, and analysis of industrial vibrating screening based on simulation prediction according to claim 6, characterized in that, The simulation trajectory of the entire process's future state is analyzed to separate the set of predicted values ​​for key performance indicators at the end of the current evaluation period. The steps are as follows: Load a predefined key performance indicator extraction rule base, which defines the calculation logic of various key performance indicators and their parameter dependencies in the full-process future state simulation trajectory; From the full-process future state simulation trajectory, extract the final state simulation snapshot corresponding to the termination time of the current evaluation cycle; The parameters in the final state simulation snapshot are substituted into the calculation according to the calculation logic defined in the rule base of the key performance indicators. The calculated set of numerical results is output as the predicted set of key performance indicators.

8. The method for data acquisition, evaluation, and analysis of industrial vibrating screening based on simulation prediction according to claim 7, characterized in that, The calculation of the predicted set of key performance indicators also includes a reference to the trends of historical simulation trajectories, specifically: While capturing the final state simulation snapshot, a trajectory segment of a fixed time length before the termination time is also captured from the full-process future state simulation trajectory as the pre-termination state trajectory. Statistical analysis is performed on the trajectory before termination to calculate the rate of change of the main state parameters in the trajectory within the fixed time length. From the key performance indicator extraction rule base, find the specific performance indicators whose calculation logic is affected by the parameter change rate; For the specific performance index, the calculation result based on the final state simulation snapshot is fused with the change rate of the state parameter corresponding to the specific performance index. The original calculation result based on the final state simulation snapshot is replaced with the fused correction value, thereby updating the set of predicted values ​​for the key performance index.

9. The method for data acquisition, evaluation, and analysis of industrial vibrating screening based on simulation prediction according to claim 8, characterized in that, The calculation yields a multi-dimensional set of performance deviation metrics, including: From the set of predicted values ​​for key performance indicators and the set of performance target parameters, one pair of predicted values ​​and target values ​​for the same performance indicator are matched one by one. For each pair of predicted and target values, calculate the absolute deviation of the predicted value relative to the target value; Read the preset deviation range reference value for the performance index from the system configuration library; Dividing the absolute deviation by the deviation range reference value yields a dimensionless relative deviation coefficient. The relative deviation coefficients corresponding to all performance indicators are organized according to their indicator categories to form the multi-dimensional performance deviation measurement set, where each dimension corresponds to the relative deviation coefficient of a performance indicator.

10. The method for data acquisition, evaluation, and analysis of industrial vibrating screening based on simulation prediction according to claim 9, characterized in that, It also includes a deep analysis step on the set of performance deviation metrics: Establish a performance index correlation graph, which describes the mutual influence relationships and influence weights among different performance indicators; The multi-dimensional set of performance deviation metrics is input into the performance index correlation graph; The performance index correlation graph calculates the relative deviation coefficient of each dimension and the indirect impact value on other related dimensions based on the internally defined mutual influence relationships and influence weights. By summing the direct relative deviation coefficient of each dimension and the indirect influence values ​​passed from all other dimensions, the overall influence strength of each dimension is obtained. All performance indicators are sorted in descending order of their overall impact intensity to generate a sorted list of key deviation dimensions. The key deviation dimension sorting list is integrated with the multi-dimensional performance deviation measurement set, key performance indicator prediction value set, and performance target parameter set to generate a final evaluation result containing in-depth analysis information.

Images