Search

CN-122021700-A - Quick printing method for backing-paper-free label

CN122021700ACN 122021700 ACN122021700 ACN 122021700ACN-122021700-A

Abstract

The invention belongs to the technical field of computer-controlled printing, and particularly relates to a rapid printing method of a bottomless paper label. The method comprises the steps of constructing a 5-dimensional state space containing position, speed, acceleration and vibration intensity, training a multi-objective optimization strategy based on a depth deterministic strategy gradient algorithm, comprehensively considering completion time, vibration energy and mechanical abrasion, deploying a convergence strategy in an embedded controller, realizing on-line dynamic path planning, carrying out closed-loop feedback correction by combining a vibration sensor and an encoder, and introducing a multi-label collaborative optimization and equipment health assessment mechanism. According to the technical scheme, the printing speed and the printing precision are improved, vibration and abrasion are effectively restrained, the service life of equipment is prolonged, and intelligent self-adaption and high-throughput flexible manufacturing are supported.

Inventors

  • ZHAI SUOQIANG

Assignees

  • 浙江美声智能系统有限公司

Dates

Publication Date
20260512
Application Date
20260123

Claims (10)

  1. 1. A method for rapid printing of a baseless label comprising the steps of: Constructing a state space model of the printing task, wherein the state space model forms a 5-dimensional state vector containing time, position, speed, acceleration and vibration intensity based on the current position of the printing head, the coordinates of a target printing point, the current motion speed, the acceleration and real-time vibration feedback data of a mechanical system; Defining a multi-objective optimized action space and a reward function, taking a discretized track point of a printing path as the action space, comprehensively considering the completion time, the vibration energy integral value and the mechanical abrasion accumulated quantity by the reward function, and setting fixed weight proportions for the three components respectively; Performing path exploration and training based on a depth deterministic strategy gradient algorithm, initializing a print head motion model in a simulation environment, and updating a strategy network and a value network by using an experience playback mechanism through interaction with the environment to acquire a state-action-rewarding sequence until the strategy converges; An online dynamic path planning module is deployed, a trained strategy network is embedded into a printing control system, and an optimal printing path is generated in real time according to current label typesetting information and a mechanical state before each printing task is started; And executing closed loop feedback correction, acquiring motion deviation and vibration response in real time through a vibration sensor and an encoder in the printing process, inputting a deviation signal into a strategy network for fine adjustment, outputting a corrected local path instruction, and ensuring printing precision and stability.
  2. 2. The method for rapid printing of a bottomless label according to claim 1, wherein the state space model for constructing the print job employs a sliding window mechanism, the window length is a predetermined period, the signals of the displacement, the speed, the acceleration and the triaxial vibration acceleration of the print head are synchronously collected at a preset sampling frequency, the high-frequency noise suppression is performed on the original signals by a low-pass filter, and the cut-off frequency is set to be smaller than a threshold value of the first-order natural frequency of the mechanical structure.
  3. 3. The rapid printing method of a bottomless label according to claim 2, characterized in that the track point density of the action space satisfies a predetermined number of points per millimeter or more, and the path resolution is not less than a preset resolution; the vibration energy integral value is obtained by squaring a triaxial vibration acceleration signal and then integrating in a time domain; The mechanical wear accumulation amount is calculated cumulatively based on the absolute value of the acceleration change rate.
  4. 4. The rapid printing method of the bottomless paper label according to claim 3, wherein the strategy network of the depth deterministic strategy gradient algorithm adopts a multi-layer fully-connected neural network structure, the number of input layer nodes is consistent with the state space dimension, the number of hidden layer nodes is preset, the number of output layer nodes is equal to the action space dimension, the activation function adopts a correction linear unit, the initial value of the learning rate is a preset initial value and adopts an exponential decay strategy, and the decay coefficient is a preset decay parameter.
  5. 5. The rapid printing method of the bottomless label according to claim 4, wherein the policy network is deployed in an embedded real-time controller, the inference delay is smaller than a preset delay threshold, the processing of more than a preset number of print job requests per second is supported, the path generation response time is not greater than a preset response time threshold, and the high-speed continuous printing beat requirement is met.
  6. 6. The method of claim 5, wherein the vibration sensor is a piezoelectric accelerometer having a predetermined sensitivity and frequency response range, the encoder has a predetermined resolution, the closed loop feedback correction period is a predetermined period, and the path correction amplitude is limited within a predetermined proportional range of the original path offset.
  7. 7. The rapid printing method of a bottomless label according to claim 6, further comprising establishing a mechanical system health status assessment module that builds a device fatigue index model based on vibration energy integral values and acceleration change rate data in historical print jobs, and when the fatigue index is greater than a preset threshold, automatically triggering maintenance pre-warning and adjusting the path aggressiveness of subsequent jobs.
  8. 8. The method of claim 7, further comprising a multi-label collaborative printing optimization mechanism, wherein when a single job includes multiple labels, the policy network optimizes the printing order and path engagement for all labels simultaneously, avoiding idle and scram operations, overall path length shortening, and vibration peak reduction.
  9. 9. The rapid printing method of a bottomless label according to claim 8, wherein the method introduces a course learning strategy in the training process, optimizes only a time target in an initial stage, then gradually adds vibration constraint, finally introduces a wear balance target, and completes training in a plurality of stages, wherein the number of training steps in each stage is not less than a preset number of steps, and ensures stable convergence of the strategy.
  10. 10. The rapid printing method of bottomless paper label according to claim 9, wherein the printing control system and the upper computer communicate through industrial ethernet to support OPCUA protocol, the path planning result is transmitted to the motion controller in NURBS curve form, the interpolation period is a predetermined period, and the track smoothness and execution accuracy are ensured.

Description

Quick printing method for backing-paper-free label Technical Field The invention belongs to the technical field of computer-controlled printing, and particularly relates to a rapid printing method of a bottomless paper label. Background Along with the deep fusion of industrial automation and intelligent label printing technology, the bottomless paper label is widely applied in the fields of logistics, retail, intelligent manufacturing and the like due to the advantages of environmental protection, material saving, high integration and the like. The traditional label printing equipment mostly adopts a fixed path planning strategy, and the motion control logic is executed based on a preset track, so that the coupling influence of dynamic load change and response characteristics of a mechanical system in the high-speed printing process is difficult to adapt, and the operation stability of the whole machine is limited. Particularly, under the working conditions of high-frequency start-stop and acceleration and deceleration, the inertia impact and structural resonance of the printing head are obviously aggravated, and the positioning precision and continuous operation efficiency of the label are directly affected. However, the existing path planning method generally depends on a static model or an empirical rule, lacks the capability of collaborative optimization of real-time vibration states and motion energy consumption, and is difficult to maximize throughput rate while guaranteeing printing quality. Meanwhile, the multi-target constraint in the high-speed printing scene has strong nonlinearity and coupling characteristics, and the traditional operation research method cannot solve on line due to high calculation complexity, so that the actual control strategy tends to be conservative. In addition, the mechanical system runs in a high-stress state for a long time, so that fatigue aging of key components is accelerated, and the reliability of the whole life cycle of the equipment is restricted. Therefore, there is a need for a rapid printing method for a bottomless label that combines reinforcement learning and dynamic motion control, and automatically explores and generates an optimal printing path that combines efficiency, accuracy and mechanical durability through an algorithm. Disclosure of Invention The invention aims to provide a rapid printing method of a bottomless paper label, which can effectively solve the problems in the background technology. In order to achieve the above object, the present invention provides a rapid printing method of a bottomless label, comprising the steps of: Constructing a state space model of the printing task, wherein the state space model forms a 5-dimensional state vector containing time, position, speed, acceleration and vibration intensity based on the current position of the printing head, the coordinates of a target printing point, the current motion speed, the acceleration and real-time vibration feedback data of a mechanical system; Defining a multi-objective optimized action space and a reward function, taking a discretized track point of a printing path as the action space, comprehensively considering the completion time, the vibration energy integral value and the mechanical abrasion accumulated quantity by the reward function, and setting fixed weight proportions for the three components respectively; Performing path exploration and training based on a depth deterministic strategy gradient algorithm, initializing a print head motion model in a simulation environment, and updating a strategy network and a value network by using an experience playback mechanism through interaction with the environment to acquire a state-action-rewarding sequence until the strategy converges; An online dynamic path planning module is deployed, a trained strategy network is embedded into a printing control system, and an optimal printing path is generated in real time according to current label typesetting information and a mechanical state before each printing task is started; And executing closed loop feedback correction, acquiring motion deviation and vibration response in real time through a vibration sensor and an encoder in the printing process, inputting a deviation signal into a strategy network for fine adjustment, outputting a corrected local path instruction, and ensuring printing precision and stability. Preferably, the state space model for constructing the print job adopts a sliding window mechanism, the window length is a preset time period, the displacement, the speed, the acceleration and the triaxial vibration acceleration signals of the print head are synchronously collected at a preset sampling frequency, the high-frequency noise suppression is carried out on the original signals through a low-pass filter, and the cutoff frequency is set to be smaller than the threshold value of the first-order natural frequency of the mechanical structure. Preferably, the track