CN-121980198-A - Filling machine material error prediction system based on multisensor fuses

CN121980198ACN 121980198 ACN121980198 ACN 121980198ACN-121980198-A

Abstract

The invention discloses a filling machine material error prediction system based on multi-sensor fusion, which comprises a multi-sensor data acquisition and fusion module, a deterministic filling quantity prediction module, a material filling error state construction module, an improved SDE-Net modeling module, a componentization diffusion modeling module, a working condition collaborative modulation module and a material filling error prediction module, wherein the multi-sensor data acquisition and fusion module is used for acquiring a multi-sensor data set and a working condition data set and constructing a multi-sensor fusion input data sequence, the deterministic filling quantity prediction module is used for generating a reference filling quantity prediction sequence, the material filling error state construction module is used for constructing a material filling error state sequence, the improved SDE-Net modeling module is used for constructing an improved SDE-Net model and configuring model parameters, the componentization diffusion modeling module is used for constructing error random disturbance structural parameters, the working condition collaborative modulation module is used for generating collaborative modulation parameters and executing synchronous modulation update, and the material filling machine material error prediction result is executed by continuous time evolution prediction processing. The method improves the accuracy and stability of material error prediction of the filling machine.

Inventors

LU YADONG
LING QING
LU YONGJUN

Assignees

合肥浩普智能装备科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260403

Claims (9)

1. A filling machine material error prediction system based on multi-sensor fusion is characterized by comprising the following steps: The multi-sensor data acquisition and fusion module is used for acquiring a multi-sensor data set and a working condition data set, executing time synchronization processing and data alignment processing, and constructing a multi-sensor fusion input data sequence; The deterministic filling quantity prediction module is used for receiving the multi-sensor fusion input data sequence, executing deterministic modeling processing through a deterministic filling quantity prediction substructure and generating a reference filling quantity prediction sequence; The material filling error state construction module is used for generating an actual filling quantity sequence based on the metering data, and executing difference value calculation processing with the reference filling quantity prediction sequence to construct a material filling error state sequence; The improved SDE-Net modeling module is used for constructing an improved SDE-Net model based on a material filling error state sequence, setting a drift sub-network and a diffusion sub-network in the model, and generating an error deterministic evolution parameter and channel disturbance contribution parameter set; The fractional quantization diffusion modeling module is used for constructing a fractional quantization diffusion structure in the diffusion sub-network and constructing error random disturbance structure parameters; The working condition collaborative modulation module is used for constructing a working condition representation vector based on the working condition data set, generating collaborative modulation parameters, and executing synchronous modulation updating to obtain modulated error deterministic evolution parameters and modulated error random disturbance structure parameters; And the material filling error prediction module is used for executing continuous time evolution prediction processing in the improved SDE-Net model to generate a filling material error prediction result.
2. The filling machine material error prediction system based on multi-sensor fusion according to claim 1, wherein the modules are realized by the following method: Acquiring a multi-sensor data set generated in the operation process of the filling machine, constructing a multi-sensor fusion input data sequence, and acquiring a working condition data set; inputting the multi-sensor fusion input data sequence into a deterministic filling quantity prediction substructure, and executing deterministic modeling processing on the filling process to generate a reference filling quantity prediction sequence; obtaining an actual filling quantity sequence based on the multi-sensor data set, and performing difference value calculation processing on the actual filling quantity sequence and the reference filling quantity prediction sequence to generate a material filling error state sequence; constructing an improved SDE-Net model based on a material filling error state sequence, wherein the model comprises a drift sub-network and a diffusion sub-network, and receiving the material filling error state sequence and generating an error deterministic evolution parameter and a channel disturbance contribution parameter set; in a diffusion sub-network, constructing a fractional diffusion structure based on a channel disturbance contribution parameter set, and receiving a multi-sensor data set to construct error random disturbance structure parameters; Constructing a working condition representation vector based on the working condition data set, inputting the drifting subnetwork and the diffusion subnetwork, executing parameter cooperative modulation processing, generating cooperative modulation parameters, and executing parameter modulation updating; in the improved SDE-Net model, material filling error state prediction calculation is executed based on the parameter modulation updating result, and a filling machine material error prediction result is generated.
3. The filling machine material error prediction system based on multi-sensor fusion according to claim 2, wherein the generation of the multi-sensor fusion input data sequence and the working condition data set comprises: The method comprises the steps of performing access processing on raw acquired data of each sensor in the operation process of the filling machine to generate a multi-sensor data set, wherein pressure data, flow data, vibration data, temperature data and metering data in the multi-sensor data set are recorded according to corresponding acquisition time stamps, and the metering data represent actual filling quantity in a corresponding filling period; performing time synchronization processing and data alignment processing on the multi-sensor data set and performing joint organization to construct a multi-sensor fusion input data sequence; And executing acquisition and finishing processing on valve opening data, a target filling quantity set value, a filling beat identifier and a material batch identifier in the running process of the filling machine, and generating a working condition data set.
4. The filling material error prediction system based on multi-sensor fusion according to claim 2, wherein the generating of the reference filling quantity prediction sequence comprises: inputting the multi-sensor fusion input data sequence into a deterministic filling quantity prediction substructure, setting an input organization unit in the deterministic filling quantity prediction substructure, and executing joint organization processing to construct a deterministic input characteristic sequence; Setting a deterministic feature mapping unit, receiving a deterministic input feature sequence, executing deterministic calculation processing, and generating a filling quantity prediction intermediate representation; and setting a filling quantity output unit, receiving a filling quantity prediction intermediate representation, performing numerical mapping processing, and organizing to form a reference filling quantity prediction sequence.
5. The multi-sensor fusion-based filling machine material error prediction system of claim 2, wherein the generation of the material filling error state sequence comprises: based on metering data in the multi-sensor data set, organizing and processing the metering data according to the filling beat identifier to obtain an actual filling quantity sequence; Calculating filling quantity error values corresponding to each filling period based on the reference filling quantity prediction sequence and the actual filling quantity sequence, and arranging to generate a filling quantity error time sequence; performing time-continuous organization processing on the filling quantity error values of adjacent filling periods based on the filling quantity error time sequence, and constructing an error state representation; Constructing an error state vector comprising a filling quantity error value corresponding to the current filling period, an error variation between adjacent filling periods and an accumulated variation of filling quantity errors based on the error state representation; and (3) carrying out joint organization on the error state vector and the filling beat mark and the material batch mark to construct a material filling error state sequence.
6. The filling machine material error prediction system based on multi-sensor fusion according to claim 2, wherein the generating of the error deterministic evolution parameter and the channel disturbance contribution parameter set comprises: Constructing a model structure of an improved SDE-Net model based on a material filling error state sequence, and constructing a drift sub-network and a diffusion sub-network in a state space of the model; Inputting the material filling error state sequence into a drift sub-network, and performing state mapping processing based on the current state of the error state sequence to generate an error deterministic evolution parameter; And (3) inputting the material filling error state sequence and the multi-sensor fusion input data sequence into a diffusion sub-network together, and executing disturbance modeling processing based on the combined characteristics of the error state and the multi-sensor fusion input data to generate a channel disturbance contribution parameter set.
7. The multi-sensor fusion-based filling machine material error prediction system according to claim 2, wherein the construction and use of the componentized diffusion structure comprises: Constructing a sub-quantitative diffusion structure based on a channel disturbance contribution parameter set in a diffusion sub-network, and setting a pressure diffusion channel, a flow diffusion channel, a vibration diffusion channel and a temperature diffusion channel; Inputting pressure data in a multi-sensor fusion input data sequence into a pressure diffusion channel, inputting flow data into a flow diffusion channel, inputting vibration data into a vibration diffusion channel and inputting temperature data into a temperature diffusion channel, and executing disturbance modeling processing in each diffusion channel based on the change characteristics of corresponding physical quantities to respectively generate a pressure disturbance contribution parameter, a flow disturbance contribution parameter, a vibration disturbance contribution parameter and a temperature disturbance contribution parameter; Performing unified scale mapping processing on the pressure disturbance contribution parameter, the flow disturbance contribution parameter, the vibration disturbance contribution parameter and the temperature disturbance contribution parameter; On the basis of completing the unified scale mapping process, performing weighted integration and dimension alignment process according to the corresponding relation between the error state dimension and each disturbance contribution parameter, mapping each disturbance contribution parameter into the same random disturbance structure representation, and constructing error random disturbance structure parameters; And taking the error random disturbance structure parameter as a disturbance input item of random evolution of an error state in the improved SDE-Net model, so that the random disturbance generated by the componentized diffusion structure directly participates in the random evolution modeling of the material filling error state.
8. A filling machine material error prediction system based on multi-sensor fusion according to claim 2, wherein the parameter modulation update comprises: Carrying out joint organization and numerical coding processing on all working condition elements according to a filling period based on the working condition data set to construct a working condition representation vector; carrying out joint mapping treatment on the working condition expression vector and the material filling error state sequence, and generating a collaborative modulation parameter based on a joint mapping result; respectively inputting the cooperative modulation parameters into an output modulation position of the drift sub-network and a disturbance aggregation modulation position of the diffusion sub-network, and respectively acting on a generation process of the error deterministic evolution parameters and an aggregation process of the error random disturbance structure parameters; and synchronous modulation updating is carried out on the error deterministic evolution parameter and the error random disturbance structure parameter in the same filling period, so that the modulated error deterministic evolution parameter and the modulated error random disturbance structure parameter are obtained.
9. The filling material error prediction system based on multi-sensor fusion according to claim 2, wherein the generating of the filling material error prediction result comprises: Inputting the multi-sensor fusion input data sequence, the working condition expression vector and the material filling error state sequence into an improved SDE-Net model together, and determining a continuous time propulsion starting state of error state prediction based on a constructed drift sub-network and a diffusion sub-network; under the constraint of the modulated error deterministic evolution parameters, performing continuous time evolution propulsion processing on the material filling error state sequence to obtain a deterministic evolution result of the error state; under the constraint of the modulated error random disturbance structure parameters, superimposing random disturbance propulsion processing on the continuous time evolution process of the error state to generate an error state evolution result; And taking the error state evolution result as the material filling error state input of the next time step, executing multi-time-step continuous prediction processing according to a preset time advancing rule, generating a material filling error prediction sequence which comprises error state predicted values corresponding to each prediction time step, and determining a filling material error predicted result based on the error state predicted values corresponding to the target prediction time.

Description

Filling machine material error prediction system based on multisensor fuses Technical Field The invention relates to the technical field of packaging machinery and filling equipment, in particular to a filling machine material error prediction system based on multi-sensor fusion. Background The filling machine is used as key equipment in a packaging production line, and the filling quantity precision directly influences the product quality and the production stability. In the actual production process, the actual filling quantity is inconsistent with the target filling quantity in the filling process often under the influence of material physical property change, equipment running state fluctuation and external environmental factors, so that a filling error is formed. How to effectively model and predict filling errors in advance is always a technical problem of concern in the field of filling equipment. In the prior art, the processing mode aiming at the filling error is mainly focused on post detection or simple statistical analysis, and is generally based on single sensor data or a small amount of working condition parameters for experience judgment, so that the error evolution rule under the coupling action of multiple physical quantities in the filling process is difficult to comprehensively reflect. Although the partial scheme introduces multi-sensor data, the partial scheme stays at the characteristic splicing or static modeling level, lacks the characterization of the continuous change characteristic of errors along with time, and is difficult to adapt to dynamic fluctuation under complex working conditions. In addition, the existing filling error prediction method often cannot distinguish the deterministic change in error evolution from random disturbance sources, and a structural modeling mechanism is lacked on disturbance influences introduced by different sensors, so that the stability of a prediction result is insufficient. Meanwhile, the influence of the working condition on the error evolution path and the disturbance intensity is usually embodied in a simple condition input mode, and the collaborative depiction of various running states is difficult to realize. Therefore, how to provide a filling machine material error prediction system based on multi-sensor fusion is a problem that needs to be solved by those skilled in the art. Disclosure of Invention The invention provides a filling machine material error prediction system based on multi-sensor fusion, which constructs a multi-sensor fusion input data sequence by acquiring a multi-sensor data set and a working condition data set, generates a reference filling quantity prediction sequence and a material filling error state sequence, constructs an improved SDE-Net model on the basis, models error deterministic evolution parameters and error random disturbance structure parameters, introduces a quantitative diffusion structure and a working condition cooperative modulation mechanism, and realizes continuous time prediction of a material filling error state. According to the invention, the error is modeled as an evolvable state, so that the comprehensive influence of multi-source disturbance and working condition change on the evolution of the filling error can be described, and the accuracy and stability of the material error prediction of the filling machine are improved. According to an embodiment of the invention, a filling machine material error prediction system based on multi-sensor fusion comprises: The multi-sensor data acquisition and fusion module is used for acquiring a multi-sensor data set and a working condition data set, executing time synchronization processing and data alignment processing, and constructing a multi-sensor fusion input data sequence; The deterministic filling quantity prediction module is used for receiving the multi-sensor fusion input data sequence, executing deterministic modeling processing through a deterministic filling quantity prediction substructure and generating a reference filling quantity prediction sequence; The material filling error state construction module is used for generating an actual filling quantity sequence based on the metering data, and executing difference value calculation processing with the reference filling quantity prediction sequence to construct a material filling error state sequence; The improved SDE-Net modeling module is used for constructing an improved SDE-Net model based on a material filling error state sequence, setting a drift sub-network and a diffusion sub-network in the model, and generating an error deterministic evolution parameter and channel disturbance contribution parameter set; The fractional quantization diffusion modeling module is used for constructing a fractional quantization diffusion structure in the diffusion sub-network and constructing error random disturbance structure parameters; The working condition collaborative modulation module is used fo