CN-121972415-A - Automatic sorting system for nitrogen springs

CN121972415ACN 121972415 ACN121972415 ACN 121972415ACN-121972415-A

Abstract

The invention relates to the technical field of equipment control, in particular to an automatic nitrogen spring sorting system, which comprises a data acquisition module, a characteristic processing module and a high-precision pressure sensor, wherein the data acquisition module acquires geometric point cloud data of a nitrogen spring through three-dimensional scanning of structured light, acquires pressure time sequence data through the high-precision pressure sensor, and sends the pressure time sequence data to the characteristic processing module after preprocessing. The feature processing module generates a comprehensive quality vector and monitors data anomalies. The dynamic self-adaptive control module performs performance simulation, combines a fuzzy decision system to perform product classification, and dynamically adjusts sorting priority through a particle swarm optimization algorithm. The sorting execution module analyzes the control instruction, plans the collision-free motion track of the cooperative robot, realizes flexible grabbing through the six-dimensional force sensor and impedance control, and feeds back data. The system evolution module updates the identification model by adopting a federal learning framework, and achieves predictive maintenance through a long-term and short-term memory network. The system realizes the full-process automation of nitrogen spring sorting, and improves the production efficiency and the product consistency.

Inventors

ZHANG XIAODONG
CHENG YANG
GONG SHIHUA
ZHANG HAIMING
QI QINGLONG

Assignees

重庆特力普尔机械设备有限公司
安徽特力普尔智能科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251127

Claims (10)

1. The nitrogen spring automatic sorting system is characterized by comprising a data acquisition module, a characteristic processing module, a dynamic self-adaptive control module, a sorting execution module and a system evolution module; The data acquisition module is configured to acquire geometric point cloud data of the nitrogen spring through three-dimensional scanning of structured light and acquire pressure time sequence data through a high-precision pressure sensor, execute outlier rejection and downsampling processing on the geometric point cloud data, execute moving average filtering on the pressure time sequence data, and perform sensor drift compensation on the processed geometric point cloud data and the processed pressure time sequence data through a self-adaptive Kalman filtering algorithm; The characteristic processing module is configured to receive the standardized data packet, execute voxel processing on geometric point cloud data in the data packet, input the voxel processed point cloud data into a space encoder to extract geometric characteristics, execute normalization processing on pressure time sequence data in the data packet, input the normalized pressure time sequence data into the time sequence encoder to extract pressure dynamic characteristics, input the geometric characteristics and the pressure dynamic characteristics into a cross attention fusion layer to generate a comprehensive quality vector, and send the comprehensive quality vector to the dynamic self-adaptive control module, wherein the characteristic processing module also monitors data distribution of the standardized data packet by adopting a local outlier algorithm, and sends a re-detection instruction to the data acquisition module when data distribution abnormality is detected; The dynamic self-adaptive control module is configured to receive the comprehensive quality vector, input the comprehensive quality vector into a digital twin model based on finite element analysis to perform performance simulation to obtain a simulation result, input the simulation result and the comprehensive quality vector into a fuzzy decision system together to perform product classification to obtain product grades, receive real-time throughput data of a production line, dynamically adjust sorting priority by taking the product grades and the real-time throughput data of the production line as inputs by adopting a particle swarm optimization algorithm to generate a sorting control instruction with a time stamp, and send the sorting control instruction to the sorting execution module; The system comprises a sorting execution module, a dynamic self-adaptive control module, a six-dimensional force sensor, a force control module, a resistance control algorithm and a dynamic self-adaptive control module, wherein the sorting execution module is configured to receive the sorting control instruction, analyze the sorting control instruction to obtain a target sorting destination, and plan a collision-free motion track of a cooperative robot; The system evolution module is configured to acquire incremental data required by model training from the feature processing module and acquire a history operation log from the sorting execution module, to perform local model training on each sorting terminal by using the incremental data and the history operation log by adopting a federal learning framework, to upload encryption model parameters generated by training to a central server for aggregation to generate global model parameters, to distribute the global model parameters to the feature processing module to update the identification model, and to analyze equipment operation data in the history operation log through a long-short-term memory network, to predict maintenance period and to send early warning information to a maintenance terminal.
2. The nitrogen spring automated sorting system of claim 1, wherein the data acquisition module comprises a structured light three-dimensional scanning unit and a pressure sensing unit; The industrial camera acquires a deformed grating image modulated by the nitrogen spring surface, carries out phase resolving on the deformed grating image, calculates three-dimensional coordinates of points on the surface of a workpiece based on a triangulation principle, and generates three-dimensional point cloud data representing the outline of the workpiece; the pressure sensing unit is in butt joint with the nitrogen spring charging valve, pressure data sampling is carried out in the closed cavity, and the acquired analog voltage signals are converted into discrete values to form time sequence data of a pressure decay curve.
3. The nitrogen spring automated sorting system of claim 2, wherein the feature processing module comprises a spatial encoder, a temporal encoder, and a cross-attention fusion layer; The space encoder receives the voxelized point cloud data, extracts global geometric features and local shape features of the point cloud data through a self-attention mechanism, and outputs geometric feature vectors; the time sequence encoder receives normalized pressure sequence data, extracts dynamic response characteristics and variation trend of the pressure data through a cyclic neural network structure, and outputs a pressure characteristic vector; And the cross attention fusion layer receives the geometric feature vector and the pressure feature vector, calculates the association weight between the two feature modes, and performs weighted summation on the geometric feature vector and the pressure feature vector according to the association weight to generate a low-dimensional comprehensive quality vector.
4. The nitrogen spring automated sorting system of claim 3, wherein the dynamic adaptive control module comprises a digital twin model, a fuzzy decision system, and a particle swarm optimizer; The digital twin model takes the comprehensive mass vector as an initial condition, based on a gas state equation and a mechanical property equation, simulates the stress distribution and fatigue life of the nitrogen spring under various load conditions, and outputs a performance simulation result; The fuzzy decision system receives the performance simulation result and the comprehensive quality vector, adopts a multi-level if-then rule base to set membership functions, classifies the nitrogen springs, and outputs a product level classification result; the particle swarm optimizer dynamically calculates the optimal priority ordering of the sorting instructions by taking the throughput efficiency of the production line as an objective function and taking the product grade classification result and the real-time throughput data of the production line as inputs.
5. The automated nitrogen spring sorting system of claim 4, wherein the sorting execution module comprises a collaborative robotic motion planner and an impedance controller; The cooperative robot motion planner receives a sorting control instruction, analyzes the coordinates and sorting destinations of a target workpiece, and calculates a collision-free motion track from the current position to the target position of the mechanical arm by adopting a rapid random exploration tree algorithm; The impedance controller collects three-dimensional force and torque data in real time through the six-dimensional force sensor in the process that the mechanical arm moves along the movement track, simulates the tail end of the robot into a spring and damping system, and dynamically adjusts rigidity and damping parameters of the end effector according to the force and torque data collected in real time.
6. The automated nitrogen spring sorting system of claim 5, wherein the system evolution module comprises a federal learning framework and a long and short term memory network; The federal learning framework acquires model increment data from the feature processing module and acquires a history operation log from the sorting execution module, adopts a random gradient descent algorithm, uses the increment data and the history operation log to carry out local model training on each sorting terminal, and uploads encryption model parameters generated by training to a central server; The long-period and short-period memory network acquires current fluctuation data, vibration spectrum data of a speed reducer and zero drift data of a force sensor of a joint motor of the cooperative robot, performs time sequence mode identification on acquired equipment operation data, calculates the residual service life of each component when equipment degradation trend is identified, and sends early warning information to a maintenance terminal before the residual service life reaches a preset threshold.
7. The automated nitrogen spring sorting system of claim 6, wherein the feature processing module performs the following steps when detecting a data anomaly: The method comprises the steps of calculating local outlier factors of a newly input standardized data packet and a historical normal sample set in a multidimensional feature space, judging that current data are abnormal data points when the local outlier factors exceed a set threshold value, generating a rechecking instruction comprising workpiece marks corresponding to the abnormal data points, sending the rechecking instruction to a data acquisition module through reverse data flow, and triggering the data acquisition module to re-execute structured light three-dimensional scanning and pressure data acquisition on the current workpiece.
8. The nitrogen spring automated sorting system of claim 7, wherein the dynamic adaptive control module, when adjusting the sorting strategy, is configured to: The method comprises the steps of monitoring working states of sorting stations and lengths of a cache queue in real time, taking minimized idle time of a production line and maximized qualified product output in unit time as optimization targets, taking a product grade sorting result and real-time throughput data of the production line as inputs, dynamically calculating optimal sorting priority through a particle swarm optimization algorithm, and updating sorting threshold values and instruction sorting in a sorting decision tree according to calculation results.
9. The nitrogen spring automated sorting system of claim 8, wherein the sort execution module, upon completion of sorting, is configured to: The method comprises the steps of packaging a grabbing success state mark, an actual grabbing time-consuming and complete six-dimensional force control data sequence and an impedance control parameter adjustment record into a feedback data packet, and sending the feedback data packet to a dynamic self-adaptive control module; The dynamic self-adaptive control module receives the feedback data packet, updates the grabbing parameter knowledge base according to the force control data and grabbing parameter adjustment record in the feedback data packet, and optimizes the force control parameter setting in the follow-up sorting control instruction according to the updated grabbing parameter knowledge base.
10. The nitrogen spring automated sorting system of claim 9, wherein the system evolution module, when performing predictive maintenance, is configured to: The method comprises the steps of collecting current fluctuation data, vibration spectrum data of a speed reducer and zero drift data of a force sensor of a joint motor of a cooperative robot, inputting collected equipment operation data into a long-period and short-period memory network to conduct time sequence mode identification, calculating residual service life of each component when a time sequence mode identification result shows equipment degradation trend, and sending early warning information to a maintenance terminal when the residual service life reaches a preset maintenance threshold value.

Description

Automatic sorting system for nitrogen springs Technical Field The invention relates to the technical field of equipment control, in particular to an automatic sorting system for nitrogen springs. Background The existing nitrogen spring sorting technology has the technical pain that detection of parameters such as roundness, diameter and the like and subsequent sorting actions are highly finished manually in the nitrogen spring production process. Operators need to conduct naked eye observation and manual measurement by means of subjective experience, and individual judgment difference and fatigue error are easily introduced in the mode, so that the detection standard cannot be kept uniform. Meanwhile, the speed and the precision of manual sorting actions are limited, and the stability is difficult to maintain under continuous operation, so that the production beat is prolonged, and the product percent of pass is directly unstable due to the fluctuation of human factors. For example, when the roundness of the cylinder barrel is detected, operators in different shifts can understand the tolerance zone to have slight differences, so that products in a part of critical states are misjudged, and in a sorting link, the manual conveying rhythm is uneven, so that the whole speed of the production line is not only dragged, but also secondary quality problems can be caused by collision. The factors limit the improvement of the production efficiency and the guarantee of the consistency of the product batch. Disclosure of Invention Aiming at the defects of the prior art, the invention provides an automatic nitrogen spring sorting system, which solves the technical problems of low production efficiency and poor product consistency of nitrogen springs caused by lack of automatic control in manual detection and sorting. In order to solve the technical problems, the invention comprises the following specific contents: The invention provides an automatic nitrogen spring sorting system which comprises a data acquisition module, a characteristic processing module, a dynamic self-adaptive control module, a sorting execution module and a system evolution module, wherein the data acquisition module is used for acquiring data of a nitrogen spring; The data acquisition module is configured to acquire geometric point cloud data of the nitrogen spring through three-dimensional scanning of structured light and acquire pressure time sequence data through a high-precision pressure sensor, execute outlier rejection and downsampling processing on the geometric point cloud data, execute moving average filtering on the pressure time sequence data, and perform sensor drift compensation on the processed geometric point cloud data and the processed pressure time sequence data through a self-adaptive Kalman filtering algorithm; The characteristic processing module is configured to receive the standardized data packet, execute voxel processing on geometric point cloud data in the data packet, input the voxel processed point cloud data into a space encoder to extract geometric characteristics, execute normalization processing on pressure time sequence data in the data packet, input the normalized pressure time sequence data into the time sequence encoder to extract pressure dynamic characteristics, input the geometric characteristics and the pressure dynamic characteristics into a cross attention fusion layer to generate a comprehensive quality vector, and send the comprehensive quality vector to the dynamic self-adaptive control module, wherein the characteristic processing module also monitors data distribution of the standardized data packet by adopting a local outlier algorithm, and sends a re-detection instruction to the data acquisition module when data distribution abnormality is detected; The dynamic self-adaptive control module is configured to receive the comprehensive quality vector, input the comprehensive quality vector into a digital twin model based on finite element analysis to perform performance simulation to obtain a simulation result, input the simulation result and the comprehensive quality vector into a fuzzy decision system together to perform product classification to obtain product grades, receive real-time throughput data of a production line, dynamically adjust sorting priority by taking the product grades and the real-time throughput data of the production line as inputs by adopting a particle swarm optimization algorithm to generate a sorting control instruction with a time stamp, and send the sorting control instruction to the sorting execution module; The system comprises a sorting execution module, a dynamic self-adaptive control module, a six-dimensional force sensor, a force control module, a resistance control algorithm and a dynamic self-adaptive control module, wherein the sorting execution module is configured to receive the sorting control instruction, analyze the sorting control instruction to obtain a tar