An intelligent printing process multi-parameter monitoring and optimization system

Abstract

The present invention belongs to the field of printing technology and discloses a multi-parameter monitoring and optimization system for an intelligent printing process; the method comprises: obtaining multi-parameter real-time data in the printing process, constructing a parameter space based on the multi-parameter real-time data; constructing the multi-parameter real-time data into a weighted directed graph, and obtaining a parameter adjustment reference route in the constructed weighted directed graph; generating an initial solution set in the parameter space, and using an optimization algorithm to solve the initial solution set to obtain an optimal parameter combination; combining the parameter adjustment reference route with the optimal parameter combination to form a dynamic adjustment strategy in the printing process and sending the strategy to a monitoring and optimization terminal; the system can reduce waste and extend the service life of equipment; at the same time, it can reduce the waste of raw materials and reduce the generation of waste.

Description

Technical Field

[0001] The present invention relates to the field of printing technology, and more particularly to an intelligent printing process multi-parameter monitoring and optimization system. Background Art

[0002] The patent application with application publication number CN117575098A discloses a method for optimizing printing parameters on gravure printing equipment. An initial parameter data set is randomly generated according to the set range of the printing parameters; the initial parameter data set is encoded using binary coding to initialize the population; the fitness corresponding to each chromosome in the initialized population is calculated, and the excellent chromosomes in the population are selected as parents, and crossover and mutation operations are performed to obtain a new population; the selection, crossover, and mutation operations are repeated on the new population until the stopping condition is met, and the individual with the smallest fitness value is output as the optimal solution. This printing parameter optimization method has automatic search and global search capabilities, automatically optimizes printing parameters, and optimizes and adjusts parameters in the printing process by simulating the evolutionary process in nature, thereby achieving more precise, stable and efficient printing control, improving printing quality, and having strong universality and flexibility.

[0003] However, in the actual printing production process, print quality is still affected by many parameters, which have complex interactions and require comprehensive trade-offs and adjustments. However, current parameter adjustments in the printing process mainly rely on the operator's experience and lack scientific optimization methods, making it difficult to ensure the stability and consistency of printing quality. At the same time, the printing process is a dynamic process, and various parameters may fluctuate at any time. If the parameters cannot be monitored and adjusted in real time, it will lead to fluctuations in print quality, resulting in a large number of defective and waste products, reducing production efficiency, and increasing energy and raw material waste. However, existing technologies lack real-time monitoring and dynamic adjustment mechanisms, and cannot respond to changes in printing status in a timely manner. In addition, existing technologies for evaluating print quality mainly rely on manual visual inspection, lack scientific quantitative indicators, and it is difficult to accurately optimize printing parameters. At the same time, manual inspection is inefficient, subject to subjectivity and uncertainty, and cannot guarantee the accuracy and consistency of the evaluation.

[0004] In view of this, the present invention proposes an intelligent printing process multi-parameter monitoring and optimization system to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned objectives, the present invention provides the following technical solutions: an intelligent printing process multi-parameter monitoring and optimization system, comprising: a space construction module for acquiring multi-parameter real-time data during the printing process and constructing a parameter space based on the multi-parameter real-time data;

[0006] A reference route fitting module is used to construct multi-parameter real-time data into a weighted directed graph, and obtain a parameter adjustment reference route in the constructed weighted directed graph;

[0007] The optimal solution module is used to generate an initial solution set in the parameter space and use the optimization algorithm to solve the optimal parameter combination based on the initial solution set;

[0008] The strategy adjustment module is used to combine the parameter adjustment baseline route with the optimal parameter combination to form a dynamic adjustment strategy during the printing process and send it to the monitoring optimization terminal; each module is connected by wired and / or wireless means.

[0009] Furthermore, the multi-parameter real-time data includes printing machine parameters, printing material parameters, environmental parameters and production efficiency parameters;

[0010] Printing machinery parameters include printing speed, printing pressure, ink supply, drying temperature and positioning accuracy; printing material parameters include paper weight, ink concentration, ink viscosity and the amount of auxiliary materials used; environmental parameters include ambient temperature, ambient humidity and ambient ventilation volume per unit time; production efficiency parameters include production quantity per unit time, raw material consumption, scrap rate and downtime.

[0011] Furthermore, the method of constructing the parameter space includes:

[0012] Based on the parameters of the multi-parameter real-time data, the number of dimensions required to construct the parameter space is determined; for each parameter of the multi-parameter real-time data, its value range is determined; for continuous parameters, they are discretized, that is, the continuous value range is divided into several discrete values;

[0013] The value ranges of all parameters are combined in sequence to construct a preliminary parameter space; each combination obtained corresponds to a spatial point in the preliminary parameter space; the feasible solution area of ​​the preliminary parameter space is marked; and the final parameter space is obtained.

[0014] Furthermore, the method of marking the feasible solution region in the preliminary parameter space includes:

[0015] From all the dimensions of the preliminary parameter space, select one as a partition dimension, and determine several partition points on the partition dimension; divide the preliminary parameter space into several intervals on the partition dimension using the partition points, and perform Cartesian products on the intervals on dimensions other than the partition dimension with the intervals obtained by the partition to obtain several subspaces;

[0016] Preset constraints and treat the subspace as a two-dimensional plane; express all constraints as geometric objects in the subspace; for each geometric object, calculate its boundary equation; the boundary equation is the equation of a line passing through a fixed point or the parametric equation of the geometric object;

[0017] From all dimensions of the subspace, select one as the scanning dimension; traverse all geometric objects, and for each geometric object, extract all points on its boundary equation, use these points as event points, and record the coordinates of the event points and the geometric objects they belong to; for each event point, if the event point is the starting point of the boundary equation of a geometric object, mark it as a starting event; if the event point is the end point of the boundary equation of a geometric object, mark it as an end point event;

[0018] All event points are sorted from smallest to largest according to their coordinate values ​​along the scanning dimension. If more than one event point has the same coordinate value along the scanning dimension, they are sorted according to the coordinate values ​​along the other dimensions. A sorted sequence of event points is obtained. An event queue is constructed based on the sorted sequence of event points. Each element in the event queue contains the coordinates of the event point, the event type, and the geometric object to which it belongs. Event types include start events and end events.

[0019] Define a data structure as the active state structure and initialize the data structure to be empty; take the first event point from the event queue. If the event point is a starting event, insert the corresponding geometric object into the active state structure; if the event point is an end event, delete the corresponding geometric object from the active state structure, and repeat until the event queue is empty; define a scan line and start scanning from the scan dimension. During the scanning process, store all geometric objects spanned by the current scan line in the active state structure. For each point in the subspace, query whether the coordinates of the point in dimensions other than the scan dimension fall within the interval in the active state structure. If the coordinates in all dimensions fall within the corresponding interval, mark the point in the feasible solution area; otherwise, mark the point in the infeasible solution area.

[0020] Furthermore, the weighted directed graph is constructed by:

[0021] Initialize a directed graph; treat each parameter in the multi-parameter real-time data as a state variable. The value range of each state variable constitutes the state space of the corresponding state variable. The Cartesian product of the state spaces of all state variables is used as the state space of the entire directed graph. Define the state of each time point in the state space of the entire directed graph as a multidimensional vector, where each component in the multidimensional vector corresponds to the value of a parameter at the corresponding time point.

[0022] For each time step t, record the current state and the next state ; Use a counter To track the status Transition to the next state The number of times, each time it is observed from the state To the next state If the transfer is successful, the counter will be increased by 1;

[0023] Set a Dirichlet prior distribution , is a prior positive vector; for each time step t, from the Dirichlet prior distribution Get the current status And the corresponding prior positive vector, denoted as ;

[0024] Defined at time step t, from the current state Transition to the next state The prior probability of ;in, is the preset attenuation factor, and ;

[0025] Based on the prior probability, for the current state and the next state , define the posterior distribution ;in, That is to be estimated from the current state Transition to the next state The transition probability of ; take the mean of the posterior distribution as the current state Transfer to the next state The transition probability of represents Dirichlet distribution;

[0026] Create a state transfer matrix. The rows of the state transfer matrix represent the current state, the columns of the state transfer matrix represent the next state, and the values ​​of the elements of the state transfer matrix are the corresponding transition probabilities.

[0027] Define the state cost function ; Calculate the function value of the state cost function of each state, recorded as the cost value; In the directed graph, define the node to represent the state. If there is a non-zero transition probability between two states, then the two corresponding nodes and A directed edge between ; Define the corresponding directed edge Weight ;in, For two nodes and The transition probability between the two corresponding states; For nodes The cost value of the corresponding state, For nodes The cost value of the corresponding state; is the balance coefficient, and ; Then the weighted directed graph G is obtained.

[0028] Furthermore, the state cost function The formula is:

[0029] ;in, 、 、 、 and is the corresponding weight coefficient; is an indicator of production efficiency; For energy consumption; For printing contrast, It is the printing color deviation; is the raw material utilization rate;

[0030] ;in, is the number of color channels to be considered; For the The weights of the color channels; In state Next, The actual value of each color channel, In state Next, target value for each color channel;

[0031] ;in, is the benchmark target value; Status The ink density or color concentration under Status Next, The standard deviation of each color channel; 、 and is the correction parameter.

[0032] Furthermore, the method for obtaining the parameter adjustment reference route includes:

[0033] Based on the original weighted directed graph G, create a new extended graph G'. Define a source point s and a sink point e on the extended graph G', which are connected to all nodes with in-degree 0 and out-degree 0 in the weighted directed graph G respectively; for each node u with in-degree 0 in G, connect a directed edge with weight 0 from the source point s to u in G'; for each node v with out-degree 0 in G, connect a directed edge with weight 0 from v to the sink point e in G';

[0034] For each node u in G', define its node label q(u) as the longest path length from u to the sink e; define a queue, add all nodes with in-degree 0 to the queue, initialize their node labels to 0, and then continuously remove nodes v with in-degree 0 from the queue. For each predecessor node v' of v, calculate its node label q(v') = max(q(v'), q(v) + w(v', v)); where w(v', v) is the weight of the directed edge from v' to v, until the node labels of all nodes are calculated; based on the calculated node labels, calculate the net weight of each edge in G';

[0035] The net weight is calculated as:

[0036] ;in, is the net weight of the directed edge between node u and node v, is the weight of the directed edge between node u and node v; For nodes The longest path length to the sink e;

[0037] Use Dijkstra's algorithm to find an edge from source point s to sink point e in G', where the sum of the net weights of the edges is equal to The path P consists of a series of nodes in G', which is recorded as a node sequence. This node sequence is projected back to the original weighted directed graph G, and a path in G is obtained. The series of states corresponding to this path are the parameter adjustment reference route.

[0038] Furthermore, the step of generating an initial solution set in the parameter space is:

[0039] Step 1: Discretize the parameter space and divide the value range of each parameter into discrete values, that is, the parameter space is divided into A small grid; is the number of dimensions of the parameter space; is an integer greater than 1;

[0040] Step 2: Initialize a dimensional Boolean matrix A, set the initial value of all elements in the Boolean matrix A to 0;

[0041] Step 3: Randomly select a small grid, set the element corresponding to it in the Boolean matrix A to 1, and use it as the initial solidification area; randomly select a dimension from all dimensions in the parameter space as the expansion direction; in the expansion direction, starting from the boundary of the solidification area, randomly select an adjacent small grid; if the element corresponding to the small grid in the Boolean matrix A is 0, it means that the small grid is not occupied, and it is occupied and added to the solidification area; repeat until the solidification area reaches the preset area size threshold M1, and obtain the comprehensive solidification area;

[0042] Step 4: Randomly select a small grid in the integrated solidification area; set the corresponding element in the Boolean matrix A to 0; repeat until the number of melts reaches the preset melting threshold M2, and M2 is less than M1; obtain the initial solution area; the combination of parameters corresponding to the small grids in the initial solution area is an initial solution;

[0043] Step 5: Repeat steps 3-4 until a preset number of initial solutions are generated to form an initial solution set.

[0044] Furthermore, the solution to the optimal parameter combination includes:

[0045] Defining the optimization function ;in, Score the registration accuracy, Score the glossiness of the imprint surface. and is the weight parameter of the corresponding item, and its value is greater than 0 and less than 1;

[0046] Randomly select N2 initial solutions from the initial solution set to form the initial hunter group, where N2 is the preset number of hunters; perform iterative optimization based on the initial hunter group, and for each iterative step of the iterative optimization, calculate the value of the optimization function corresponding to each hunter; record it as the hunting value of the corresponding hunter;

[0047] Sort the hunters by their hunting value from large to small, record the first n3 hunters as candidate prey, and the rest as irrelevant hunters; n3 is an integer greater than 1;

[0048] Calculate the average distance from each candidate prey to all other candidate prey; calculate the difference between the distance from each candidate prey to the hunter and its average distance, recorded as hunting difference df; calculate the weight of each candidate prey based on hunting difference df, this weight is the hunting difference df divided by the sum of all differences;

[0049] The candidate prey with the largest weight is used as the tracking prey Y, and the displacement vector V = ra × (YX) of hunter X in the direction of the tracking prey is calculated; where ra is a random number between (0, 1); based on the displacement vector V, the new position X_new = X + V of hunter X is updated;

[0050] Calculate the hunting value corresponding to the new position X_new. If this hunting value is greater than the hunting value of the original hunter X, update the position of hunter X to X_new. Obtain the optimal hunter group.

[0051] Preset the exchange probability PR, randomly select one of the irrelevant hunters A from all the irrelevant hunters, randomly select a hunter B from the hunters whose positions have been updated, and use the combination of A and B with probability PR to form a new hunter C to replace the irrelevant hunter A; repeat until all the remaining irrelevant hunters are replaced to obtain a new generation of hunters; merge the optimal hunter group with the new generation of hunters to form a new hunter group; repeat the formation of the new hunter group until the preset maximum number of iterations is reached; from the final new hunter group, select the hunter with the highest hunting value, and the corresponding parameter combination is the optimal parameter combination obtained by solution.

[0052] Furthermore, the registration accuracy score ;in, For the The registration error of each measuring point is is the preset maximum registration error threshold, and All are adjustment parameters greater than 0; For the The distance from each measuring point to the edge of the printed part; dmax is the maximum value of the distances from all measuring points to the edge of the printed part; is a penalty coefficient greater than 0;

[0053] Imprint surface gloss score ;in, For the The weight of each measurement point, For the Gloss value of each measuring point, For the Gloss gradient value of each measuring point; is the gradient adjustment parameter, and its value is greater than or equal to 0; is the distribution penalty coefficient, and ; is the standard deviation of the gloss values ​​of all measurement points; is the preset reference gloss standard deviation.

[0054] The technical effects and advantages of the intelligent printing process multi-parameter monitoring and optimization system of the present invention are as follows:

[0055] The present invention realizes the intelligence and automation of the printing process, greatly reduces the uncertainty and subjectivity of human operation, and ensures the stability and consistency of printing quality; through real-time monitoring and dynamic adjustment, it can adaptively respond to changes in printing status, promptly correct deviations, eliminate fluctuations, greatly reduce the generation of defective and waste products, and improve the qualified product rate; secondly, by adopting scientific optimization algorithms and quantitative evaluation indicators, it can search for the global optimal solution in the high-dimensional solution space of multiple parameters, realize the precise optimization of printing parameters, fully tap the performance potential of printing equipment, maximize the printing quality, and make the clarity, contrast, color reproduction and other indicators of printed parts reach the ideal level; in addition, through parameter optimization, it can effectively reduce the energy consumption and raw material consumption of the printing process, and improve resource utilization efficiency; the optimized parameter combination can reduce waste and extend the service life of equipment; at the same time, it reduces the waste of raw materials and the generation of waste, thereby reducing production costs and achieving a win-win situation of economic and environmental benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Schematic diagram of a multi-parameter monitoring and optimization system for an intelligent printing process according to the present invention;

[0057] Figure 2 Schematic diagram of a multi-parameter monitoring and optimization method for an intelligent printing process according to the present invention. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0059] Example 1

[0060] See also Figure 1 As shown, the intelligent printing process multi-parameter monitoring and optimization system described in this embodiment includes:

[0061] A space construction module is used to obtain multi-parameter real-time data during the printing process and construct a parameter space based on the multi-parameter real-time data;

[0062] A reference route fitting module is used to construct multi-parameter real-time data into a weighted directed graph, and obtain a parameter adjustment reference route in the constructed weighted directed graph;

[0063] The optimal solution module is used to generate an initial solution set in the parameter space and use the optimization algorithm to solve the optimal parameter combination based on the initial solution set;

[0064] The strategy adjustment module is used to combine the parameter adjustment baseline route with the optimal parameter combination to form a dynamic adjustment strategy during the printing process and send it to the monitoring optimization terminal; each module is connected through wired and / or wireless means to realize data transmission between modules.

[0065] Multi-parameter real-time data includes printing machinery parameters, printing material parameters, environmental parameters and production efficiency parameters; printing machinery parameters include printing speed, printing pressure, ink supply, drying temperature and positioning accuracy (the degree of deviation of the printed image relative to the edge of the paper, which affects the appearance of the printed part); it is obtained through real-time monitoring by various sensors installed on the printing press, such as encoders, pressure sensors, flow meters, etc.

[0066] Printing material parameters include paper weight, ink concentration, ink viscosity, and the amount of auxiliary materials (such as curing agent, thickener, etc.); ink concentration and viscosity can be obtained by real-time measurement using an online color densitometer and viscometer.

[0067] Environmental parameters include ambient temperature, ambient humidity, and ambient ventilation volume per unit time; production efficiency parameters include production quantity per unit time, raw material consumption, scrap rate, and downtime; production quantity per unit time is obtained through real-time statistics using a counter or encoder; raw material consumption is obtained through real-time monitoring using a flow meter or weighing sensor; scrap rate is obtained through real-time statistics using manual inspection; and downtime is obtained through real-time recording using a timer or encoder.

[0068] Ways to construct parameter space include:

[0069] Based on the parameters of the multi-parameter real-time data (such as printing speed, printing pressure, ink supply, etc.), determine the number of dimensions required to construct the parameter space; for each parameter (data type) of the multi-parameter real-time data, determine its value range; for example, the printing speed may be 1000-5000 sheets / minute, the printing pressure may be 50-200N, etc.; for continuous parameters, they are discretized, that is, the continuous value range is divided into several discrete values; for example, the printing speed may be a continuous value.

[0070] The value ranges of all parameters are combined in sequence to construct a preliminary parameter space; each combination obtained corresponds to a spatial point in the preliminary parameter space; since there may be mutual constraints between different parameters, some parameter combinations may not be feasible. According to the actual process limitations, the feasible solution area of ​​the preliminary parameter space is marked to obtain the final parameter space.

[0071] Specifically, one dimension is selected from all the dimensions of the preliminary parameter space as a partitioning dimension, and several partitioning points are determined on the partitioning dimension, which can be selected according to heuristic strategies, such as equal division, density-based, etc. The preliminary parameter space is divided into several intervals on the partitioning dimension by the partitioning points, and the intervals on other dimensions except the partitioning dimension are subjected to Cartesian products with the intervals obtained by these partitions to obtain several subspaces.

[0072] Preset constraints. Specifically, the constraints are parameterized according to the specific printing process, equipment, materials and quality requirements. The constraints are equations, inequalities, logical expressions, etc. between parameters. The subspace is regarded as a two-dimensional plane (coordinate system). All constraints are expressed as geometric objects in the subspace (such as line segments, polygons, etc.). For each geometric object, its boundary equation is calculated. The boundary equation is the equation of a straight line passing through a fixed point or the parametric equation of the geometric object.

[0073] From all dimensions of the subspace, select one as the scanning dimension; traverse all geometric objects, and for each geometric object, extract all points on its boundary equation, use these points as event points, and record the coordinates of the event points and the geometric objects they belong to.

[0074] For each event point, if the event point is the starting point of the boundary equation of a geometric object, it is marked as a starting point event; if the event point is the end point of the boundary equation of a geometric object, it is marked as an end point event; all event points are sorted in ascending order according to the coordinate values ​​on the scanning dimension; if more than one event point has the same coordinate value on the scanning dimension, secondary sorting is performed according to the coordinate values ​​of other dimensions; a sorted sequence of event points is obtained; an event queue is constructed based on the sorted sequence of event points; each element in the event queue contains the coordinates of the event point, the event type, and the geometric object to which it belongs; event types include starting events and end events.

[0075] Define a data structure as the active state structure, such as a red-black tree or segment tree, and initialize the data structure to be empty; take the first event point from the event queue. If the event point is a starting event, insert the corresponding geometric object into the active state structure; if the event point is an ending event, delete the corresponding geometric object from the active state structure. Repeat until the event queue is empty.

[0076] A scan line is defined, starting from the scan dimension. During the scan, all geometric objects spanned by the current scan line are stored in the active state structure. For each point in the subspace, the coordinates of the point in dimensions other than the scan dimension are checked to see if they fall within the interval in the active state structure. If the coordinates in all dimensions fall within the corresponding interval, the point is in the feasible solution region and marked as such. Otherwise, the point is in the infeasible solution region and marked as such. The corresponding parameter space is constructed based on multi-parameter real-time data to prepare for subsequent optimization calculations. The finer the parameter space, the better the optimization effect may be, but it also requires more computing resources.

[0077] The construction methods of weighted directed graphs include:

[0078] Initialize a directed graph; regard each parameter in the multi-parameter real-time data as a state variable, the value range of each state variable constitutes the state space of the corresponding state variable, and the Cartesian product of the state spaces of all state variables is used as the state space of the entire directed graph; define the state of each time point in the state space of the entire directed graph as a multidimensional vector, and each component in the multidimensional vector corresponds to the value of a parameter at the corresponding time point.

[0079] For each time step t, record the current state and the next state ; Use a counter To track the status Transition to the next state The number of times, each time it is observed from the state To the next state If a transfer occurs, the counter is increased by 1.

[0080] Set a Dirichlet prior distribution , It is a priori positive vector that reflects the prior knowledge of the state transition probability. Specifically, the multidimensional parameter data of the current time step are clustered to obtain several clusters, each cluster represents a parameter state, and the transition frequency of each cluster to other clusters is counted. These transition frequencies are normalized and used as the prior positive vector of the corresponding state. The Dirichlet prior distribution is a matrix distributed according to the time step, which contains each state and its corresponding positive vector.

[0081] For each time step t, from the Dirichlet prior distribution Get the current status And the corresponding prior positive vector, denoted as ; Defined at time step t, from the current state Transition to the next state The prior probability of ;in, is the preset attenuation factor, and ; Newly observed transitions will have a greater impact on prior knowledge.

[0082] Based on the prior probability, for the current state and the next state , define the posterior distribution ;in, That is to estimate the current state Transition to the next state The transition probability of ; take the mean of the posterior distribution as the current state Transfer to the next state The transition probability of represents Dirichlet distribution.

[0083] Create a state transfer matrix. The rows of the state transfer matrix represent the current state, the columns of the state transfer matrix represent the next state, and the values ​​of the elements of the state transfer matrix are the corresponding transition probabilities.

[0084] Define the state cost function ;

[0085] ;in, 、 、 、 and is the corresponding weight coefficient, reflecting the importance of each cost function; It is a production efficiency indicator, specifically the number of products produced per unit time, obtained by multiplying the printing speed by the actual operating time ratio; Energy consumption is the sum of electrical energy consumption and other energy consumption. Electrical energy consumption is power multiplied by operating time. Power and other energy consumption can be modeled and measured based on parameter settings under corresponding states, such as printing speed and temperature. is the printing contrast (the ratio between the maximum brightness value and the minimum brightness value), It is the printing color deviation; It is the raw material utilization rate. Specifically, the raw material utilization rate is the actual output quantity divided by the sum of the theoretical output quantity and the waste quantity. The theoretical output quantity is the quantity that should be output under the corresponding conditions under ideal circumstances; the waste quantity is the quantity of waste generated in the actual production process.

[0086] ;in, The number of color channels to consider, usually 3 (RGB) or 4 (CMYK); For the The weight of each color channel reflects the importance of the channel to the overall color; In state Next, The actual value of each color channel, In state Next, Target value for each color channel;

[0087] ;in, The reference target value can be a standard color value, such as RGB (255, 255, 255) representing pure white, or a theoretical target value set according to the type and concentration of ink used; Status The ink density or color concentration under Status Next, The standard deviation of each color channel; 、 and To correct the parameters, fitting is performed based on the measured data; this can effectively eliminate color deviations caused by changes in environmental and material conditions.

[0088] Calculate the function value of the state cost function of each state, recorded as the cost value, which represents the quality of the state; in the directed graph, define the node to represent the state. If there is a non-zero transition probability between two states, then the two corresponding nodes and A directed edge between ; Define the corresponding directed edge Weight ;in, For two nodes and The transition probability between the two corresponding states; For nodes The cost value of the corresponding state, For nodes The cost value of the corresponding state; is the balance coefficient, and , which is used to balance the influence of transfer probability and cost value; then a weighted directed graph G is obtained.

[0089] Methods for obtaining parameter adjustment benchmark routes include:

[0090] Based on the original weighted directed graph G, a new extended graph G' is created, and a source point s and a sink point e are defined on the extended graph G', which are respectively connected to all nodes with in-degree 0 and out-degree 0 in the weighted directed graph G; for each node u with in-degree 0 in G, a directed edge with weight 0 is connected from the source point s to u in G', and for each node v with out-degree 0 in G, a directed edge with weight 0 is connected from v to the sink point e in G'; for a node u, the number of directed edges starting from the node u is called the out-degree of the node, and the number of directed edges pointing to the node u is called the in-degree of the node.

[0091] For each node u in G', define its node label q(u) as the longest path length from u to the sink e; define a queue, add all nodes with in-degree 0 (including e) to the queue, initialize their node labels to 0, and then continuously take out nodes v with in-degree 0 from the queue, and for each predecessor node v' of v, calculate its node label q(v') = max(q(v'), q(v) + w(v', v)); more specifically, for each node u, maintain a predecessor node, which records the previous node of u on the shortest path from the starting point to u; by continuously tracking the predecessor nodes, the shortest path from the starting point to any node can be reconstructed; where w(v', v) is the weight of the directed edge from v' to v (inherited from the weighted directed graph G), until the values ​​of the node labels of all nodes are calculated; based on the calculated node labels, calculate the net weight of each edge in G'.

[0092] It should be explained that G' contains not only the source node s and the sink node e and the nodes connected to them, but also the topological structure of the original weighted directed graph G.

[0093] The net weight is calculated as:

[0094] ;in, is the net weight of the directed edge between node u and node v, is the weight of the directed edge between node u and node v (inherited from the weighted directed graph G); For nodes The longest path length to the sink e.

[0095] Use Dijkstra's algorithm to find an edge from source point s to sink point e in G', where the sum of the net weights of the edges is equal to The path P consists of a series of nodes in G', which is recorded as a node sequence. This node sequence is projected back to the original weighted directed graph G, and a path in G is obtained. The series of states corresponding to this path are the parameter adjustment reference route.

[0096] The steps to generate an initial solution set in parameter space are:

[0097] Step 1: Discretize the parameter space and divide the value range of each parameter into discrete values, that is, the parameter space is divided into A small grid (hypercube); is the number of dimensions of the parameter space; is an integer greater than 1.

[0098] Step 2: Initialize a -dimensional Boolean matrix A, and set the initial value of all elements in Boolean matrix A to 0.

[0099] Step 3: Randomly select a small grid, set the element in the corresponding Boolean matrix A to 1, and use it as the initial solidification area; randomly select a dimension from all dimensions in the parameter space as the expansion direction ( expansion direction); in the expansion direction, starting from the boundary of the solidification area, randomly select an adjacent small grid; if the element corresponding to the small grid in the Boolean matrix A is 0, it means that the small grid is not occupied, and it is occupied (the element is set to 1) and added to the solidification area; repeat until the solidification area reaches the preset area size threshold M1; and obtain the comprehensive solidification area.

[0100] Step 4: Randomly select a small grid in the integrated solidification area; melt it, that is, set the corresponding element in the Boolean matrix A to 0; repeat until the number of melts reaches the preset melting threshold M2, and M2 is less than M1; obtain the initial solution area; the combination of parameters corresponding to the small grids in the initial solution area is an initial solution.

[0101] Step 5: Repeat steps 3-4 until a preset number of initial solutions are generated to form an initial solution set.

[0102] The methods for finding the optimal parameter combination include:

[0103] Defining the optimization function ;in, Score the registration accuracy, Score the glossiness of the imprint surface. and is the weight parameter of the corresponding item, and its value is greater than 0 and less than 1;

[0104] Registration accuracy score ;in, For the The registration error of each measuring point is is the preset maximum registration error threshold, and All are adjustment parameters greater than 0; For the The distance from each measuring point to the edge of the printed part; dmax is the maximum value of the distances from all measuring points to the edge of the printed part; It is a penalty coefficient greater than 0 and is used to adjust the degree of penalty.

[0105] Imprint surface gloss score ;in, For the The weight of each measurement point (preset) reflects the importance of the point to the overall glossiness. For the Gloss value of each measuring point, For the The gloss gradient value of each measuring point reflects the rate of change of gloss near the point; is a gradient adjustment parameter, and its value is greater than or equal to 0. When the gloss gradient value is large, increasing the weight of the area with drastic gloss changes can increase the weight of the gloss transition area. is the distribution penalty coefficient, and ; is the standard deviation of the gloss values ​​of all measurement points, reflecting the degree of dispersion of gloss distribution; is the preset reference gloss standard deviation.

[0106] Randomly select N2 initial solutions from the initial solution set to form the initial hunter group, where N2 is the preset number of hunters; perform iterative optimization based on the initial hunter group, and for each iterative step of the iterative optimization, calculate the value of the optimization function corresponding to each hunter (combination of parameters); record it as the hunting value of the corresponding hunter.

[0107] Sort the hunters according to their hunting values ​​from large to small, record the first n3 hunters as candidate prey, and the rest as irrelevant hunters; n3 is an integer greater than 1; calculate the average distance from each candidate prey to all other candidate prey, and the average distance is the average of the differences in hunting values; calculate the difference between the distance from each candidate prey to the hunter (the difference in hunting values) and its average distance, and record it as the hunting difference df; calculate the weight of each candidate prey based on the hunting difference df, and this weight is the hunting difference df divided by the sum of all differences.

[0108] The candidate prey with the largest weight is used as the tracking prey Y, and the displacement vector V = ra × (YX) of hunter X in the direction of the tracking prey is calculated; where ra is a random number between (0, 1); based on the displacement vector V, the new position of hunter X is updated to X_new = X + V; the hunting value corresponding to the new position X_new is calculated. If this hunting value is greater than the hunting value of the original hunter X, the position of hunter X is updated to X_new. The optimal hunter group is obtained.

[0109] Preset the exchange probability PR, randomly select one of the irrelevant hunters A from all the irrelevant hunters, randomly select a hunter B from the hunters whose positions have been updated, and use the combination of A and B with probability PR to form a new hunter C to replace the irrelevant hunter A; repeat until all the remaining irrelevant hunters are replaced, and a new generation of hunters is obtained.

[0110] The optimal hunter group is merged with the new generation hunter group to form a new hunter group; the new hunter group is formed repeatedly until the preset maximum number of iterations is reached; from the final new hunter group, the hunter with the highest hunting value is selected, and the corresponding parameter combination is the optimal parameter combination obtained by solution.

[0111] In the above process, it is necessary to encode the combination of parameters. In addition, it is necessary to set some hyperparameters and tune them through cross-validation; perform efficient global search in the solution space to avoid falling into local optimality and thus find the global optimal solution.

[0112] The optimal parameter combination corresponds to a point (multi-dimensional vector) in the parameter space. On the parameter adjustment benchmark route, find the node (state) closest to the optimal parameter combination, recorded as node S*, and calculate the Euclidean distance dist (between vectors) between the optimal parameter combination and the state corresponding to node S*. If dist is less than the preset distance threshold, S* is used as the starting point; otherwise, the optimal parameter combination is added to the parameter adjustment benchmark route as a new starting point; starting from the starting point, along the parameter adjustment benchmark route, determine the combination of parameters for each time step to form a state sequence; for the state of each time step in the state sequence, according to the established state cost function, Calculate the cost of the state; set two cost thresholds, namely the ideal cost threshold and the warning cost threshold, and the ideal cost threshold is less than the warning cost threshold. If the cost of the current state is lower than the ideal cost threshold, maintain the current state and enter the next time step; if the cost of the current state is higher than the warning cost threshold, find a new optimal path to transfer to the new state according to the previously established weighted directed graph, and use it as the state of the next time step; if the cost of the current state is between the ideal cost threshold and the warning cost threshold, there is a certain probability (preset) to transfer to the new state, and the new state is determined by the previous weighted directed graph, which is a dynamic adjustment strategy.

[0113] During the printing process, the current printing parameter data is monitored in real time and mapped to the corresponding state. If the current state deviates from the expected state of the dynamic adjustment strategy, the corresponding parameters are automatically adjusted according to the rules of the dynamic adjustment strategy to gradually return to the trajectory of the expected state sequence.

[0114] While maintaining the overall adjustment direction, it also has certain adaptability and robustness, and can make dynamic adjustments based on real-time status, thereby improving the stability of the printing process and product quality; at the same time, the ideal and warning thresholds are set to maintain a certain inertia when the cost value is in a reasonable range, avoiding too frequent parameter adjustments.

[0115] This embodiment realizes the intelligence and automation of the printing process, greatly reduces the uncertainty and subjectivity of human operation, and ensures the stability and consistency of printing quality; through real-time monitoring and dynamic adjustment, it can adaptively respond to changes in printing status, promptly correct deviations, eliminate fluctuations, significantly reduce the generation of defective and waste products, and improve the qualified product rate; secondly, by adopting scientific optimization algorithms and quantitative evaluation indicators, it can search for the global optimal solution in the high-dimensional solution space of multiple parameters, realize the precise optimization of printing parameters, fully tap the performance potential of printing equipment, maximize the printing quality, and make the clarity, contrast, color reproduction and other indicators of printed parts reach the ideal level; in addition, through parameter optimization, it can effectively reduce the energy consumption and raw material consumption of the printing process, and improve resource utilization efficiency; the optimized parameter combination can reduce waste and extend the service life of equipment; at the same time, it reduces the waste of raw materials and the generation of waste, thereby reducing production costs and achieving a win-win situation of economic and environmental benefits.

[0116] Example 2

[0117] See also Figure 2 As shown, for the parts not described in detail in this embodiment, please refer to the description of embodiment 1. A method for multi-parameter monitoring and optimization of an intelligent printing process is provided, comprising:

[0118] S1. Acquire multi-parameter real-time data during the printing process and construct a parameter space based on the multi-parameter real-time data;

[0119] S2. Constructing the multi-parameter real-time data into a weighted directed graph, and obtaining a parameter adjustment reference route in the constructed weighted directed graph;

[0120] S3. Generate an initial solution set in the parameter space, and use an optimization algorithm to solve the optimal parameter combination based on the initial solution set;

[0121] S4. Combining the parameter adjustment baseline route with the optimal parameter combination to form a dynamic adjustment strategy during the printing process and sending it to the monitoring optimization terminal.

[0122] Example 3

[0123] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the multi-parameter monitoring and optimization method of the intelligent printing process provided above is implemented.

[0124] Since the electronic device introduced in this embodiment is an electronic device used to implement a method for multi-parameter monitoring and optimization of an intelligent printing process in the embodiment of this application, based on the method for multi-parameter monitoring and optimization of an intelligent printing process introduced in the embodiment of this application, those skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application will not be described in detail here. As long as those skilled in the art implement the electronic device used in the method for multi-parameter monitoring and optimization of an intelligent printing process in the embodiment of this application, it falls within the scope of protection of this application.

[0125] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0126] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for users of ordinary skill in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. An intelligent printing process multi-parameter monitoring and optimization system, characterized in that: include: A space construction module is used to obtain multi-parameter real-time data during the printing process and construct a parameter space based on the multi-parameter real-time data; The multi-parameter real-time data includes printing machine parameters, printing material parameters, environmental parameters and production efficiency parameters; Printing machinery parameters include printing speed, printing pressure, ink supply, drying temperature, and positioning accuracy; printing material parameters include paper weight, ink concentration, ink viscosity, and the amount of auxiliary materials used; environmental parameters include ambient temperature, ambient humidity, and ambient ventilation volume per unit time; production efficiency parameters include production quantity per unit time, raw material consumption, scrap rate, and downtime; A reference route fitting module is used to construct multi-parameter real-time data into a weighted directed graph, and obtain a parameter adjustment reference route in the constructed weighted directed graph; The weighted directed graph is constructed in the following manner: Initialize a directed graph; treat each parameter in the multi-parameter real-time data as a state variable. The value range of each state variable constitutes the state space of the corresponding state variable. The Cartesian product of the state spaces of all state variables is used as the state space of the entire directed graph. Define the state of each time point in the state space of the entire directed graph as a multidimensional vector, where each component in the multidimensional vector corresponds to the value of a parameter at the corresponding time point. For each time step t, record the current state and the next state ; Use a counter To track the status Transition to the next state The number of times, each time it is observed from the state To the next state If the transfer is successful, the counter will be increased by 1; Set a Dirichlet prior distribution , is a prior positive vector; for each time step t, from the Dirichlet prior distribution Get the current status And the corresponding prior positive vector, denoted as ; Defined at time step t, from the current state Transition to the next state The prior probability of ;in, is the preset attenuation factor, and ; Based on the prior probability, for the current state and the next state , define the posterior distribution ;in, That is to be estimated from the current state Transition to the next state The transition probability of ; take the mean of the posterior distribution as the current state Transfer to the next state The transition probability of Indicates that the parameter is Dirichlet distribution of ; Create a state transfer matrix. The rows of the state transfer matrix represent the current state, the columns of the state transfer matrix represent the next state, and the values ​​of the elements of the state transfer matrix are the corresponding transition probabilities. Define the state cost function ; Calculate the function value of the state cost function of each state, recorded as the cost value; In the directed graph, define the node to represent the state. If there is a non-zero transition probability between two states, then the two corresponding nodes and A directed edge between ; Define the corresponding directed edge Weight ;in, For two nodes and The transition probability between the two corresponding states; For nodes The cost value of the corresponding state, For nodes The cost value of the corresponding state; is the balance coefficient, and ; Then we get the weighted directed graph G; The optimal solution module is used to generate an initial solution set in the parameter space and use the optimization algorithm to solve the optimal parameter combination based on the initial solution set; The strategy adjustment module is used to combine the parameter adjustment baseline route with the optimal parameter combination to form a dynamic adjustment strategy during the printing process and send it to the monitoring optimization terminal; each module is connected by wired and / or wireless means.

2. The intelligent printing process multi-parameter monitoring and optimization system according to claim 1, characterized in that: The method of constructing the parameter space includes: Based on the parameters of the multi-parameter real-time data, the number of dimensions required to construct the parameter space is determined; for each parameter of the multi-parameter real-time data, its value range is determined; for continuous parameters, they are discretized, that is, the continuous value range is divided into several discrete values; The value ranges of all parameters are combined in sequence to construct a preliminary parameter space; each combination obtained corresponds to a spatial point in the preliminary parameter space; the feasible solution area of ​​the preliminary parameter space is marked; and the final parameter space is obtained.

3. The intelligent printing process multi-parameter monitoring and optimization system according to claim 2, characterized in that: The method of marking the feasible solution region in the preliminary parameter space includes: From all the dimensions of the preliminary parameter space, select one as a partition dimension, and determine several partition points on the partition dimension; divide the preliminary parameter space into several intervals on the partition dimension using the partition points, and perform Cartesian products on the intervals on dimensions other than the partition dimension with the intervals obtained by the partition to obtain several subspaces; Preset constraints and treat the subspace as a two-dimensional plane; express all constraints as geometric objects in the subspace; for each geometric object, calculate its boundary equation; the boundary equation is the equation of a line passing through a fixed point or the parametric equation of the geometric object; From all dimensions of the subspace, select one as the scanning dimension; traverse all geometric objects, and for each geometric object, extract all points on its boundary equation, use these points as event points, and record the coordinates of the event points and the geometric objects they belong to; for each event point, if the event point is the starting point of the boundary equation of a geometric object, mark it as a starting event; if the event point is the end point of the boundary equation of a geometric object, mark it as an end point event; All event points are sorted from smallest to largest according to their coordinate values ​​along the scanning dimension. If more than one event point has the same coordinate value along the scanning dimension, they are sorted according to the coordinate values ​​along the other dimensions. A sorted sequence of event points is obtained. An event queue is constructed based on the sorted sequence of event points. Each element in the event queue contains the coordinates of the event point, the event type, and the geometric object to which it belongs. Event types include start events and end events. Define a data structure as the active state structure and initialize the data structure to be empty; take the first event point from the event queue. If the event point is a starting event, insert the corresponding geometric object into the active state structure; if the event point is an end event, delete the corresponding geometric object from the active state structure, and repeat until the event queue is empty; define a scan line and start scanning from the scan dimension. During the scanning process, store all geometric objects spanned by the current scan line in the active state structure. For each point in the subspace, query whether the coordinates of the point in dimensions other than the scan dimension fall within the interval in the active state structure. If the coordinates in all dimensions fall within the corresponding interval, mark the point in the feasible solution area; otherwise, mark the point in the infeasible solution area.

4. The intelligent printing process multi-parameter monitoring and optimization system according to claim 3, characterized in that: The state cost function The formula is: ;in, 、 、 、 and is the corresponding weight coefficient; is an indicator of production efficiency; For energy consumption; For printing contrast, It is the printing color deviation; is the raw material utilization rate; ;in, is the number of color channels to be considered; For the The weights of the color channels; In state Next, The actual value of each color channel, In state Next, Target value for each color channel; ;in, is the benchmark target value; Status The ink density or color concentration under Status Next, The standard deviation of each color channel; 、 and is the correction parameter.

5. The intelligent printing process multi-parameter monitoring and optimization system according to claim 4, characterized in that: The method for obtaining the parameter adjustment reference route includes: Based on the original weighted directed graph G, create a new extended graph G'. Define a source point s and a sink point e on the extended graph G', which are connected to all nodes with in-degree 0 and out-degree 0 in the weighted directed graph G respectively; for each node u with in-degree 0 in G, connect a directed edge with weight 0 from the source point s to u in G'; for each node v with out-degree 0 in G, connect a directed edge with weight 0 from v to the sink point e in G'; For each node u in G', define its node label q(u) as the longest path length from u to the sink e; define a queue, add all nodes with in-degree 0 to the queue, initialize their node labels to 0, and then continuously remove nodes v with in-degree 0 from the queue. For each predecessor node v' of v, calculate its node label q(v') = max(q(v'), q(v) + w(v', v)); where w(v', v) is the weight of the directed edge from v' to v, until the node labels of all nodes are calculated; based on the calculated node labels, calculate the net weight of each edge in G'; The net weight is calculated as: ;in, is the net weight of the directed edge between node u and node v, is the weight of the directed edge between node u and node v; For nodes The longest path length to the sink e; Use Dijkstra's algorithm to find an edge from source point s to sink point e in G', where the sum of the net weights of the edges is equal to The path P consists of a series of nodes in G', which is recorded as a node sequence. This node sequence is projected back to the original weighted directed graph G, and a path in G is obtained. The series of states corresponding to this path are the parameter adjustment reference route.

6. The intelligent printing process multi-parameter monitoring and optimization system according to claim 5, characterized in that: The steps of generating an initial solution set in the parameter space are: Step 1: Discretize the parameter space and divide the value range of each parameter into discrete values, that is, the parameter space is divided into A small grid; is the number of dimensions of the parameter space; is an integer greater than 1; Step 2: Initialize a dimensional Boolean matrix A, set the initial value of all elements in the Boolean matrix A to 0; Step 3: Randomly select a small grid, set the element corresponding to it in the Boolean matrix A to 1, and use it as the initial solidification area; randomly select a dimension from all dimensions in the parameter space as the expansion direction; in the expansion direction, starting from the boundary of the solidification area, randomly select an adjacent small grid; if the element corresponding to the small grid in the Boolean matrix A is 0, it means that the small grid is not occupied, and it is occupied and added to the solidification area; repeat until the solidification area reaches the preset area size threshold M1, and obtain the comprehensive solidification area; Step 4: Randomly select a small grid in the integrated solidification area; set the corresponding element in the Boolean matrix A to 0; repeat until the number of melts reaches the preset melting threshold M2, and M2 is less than M1; obtain the initial solution area; the combination of parameters corresponding to the small grids in the initial solution area is an initial solution; Step 5: Repeat steps 3-4 until a preset number of initial solutions are generated to form an initial solution set.

7. The intelligent printing process multi-parameter monitoring and optimization system according to claim 6, characterized in that: The solution method for the optimal parameter combination includes: Defining the optimization function ;in, Score the registration accuracy, Score the glossiness of the imprint surface. and is the weight parameter of the corresponding item, and its value is greater than 0 and less than 1; Randomly select N2 initial solutions from the initial solution set to form the initial hunter group, where N2 is the preset number of hunters; perform iterative optimization based on the initial hunter group, and for each iterative step of the iterative optimization, calculate the value of the optimization function corresponding to each hunter; record it as the hunting value of the corresponding hunter; Sort the hunters by their hunting value from large to small, record the first n3 hunters as candidate prey, and the rest as irrelevant hunters; n3 is an integer greater than 1; Calculate the average distance from each candidate prey to all other candidate prey; calculate the difference between the distance from each candidate prey to the hunter and its average distance, recorded as hunting difference df; calculate the weight of each candidate prey based on hunting difference df, this weight is the hunting difference df divided by the sum of all differences; The candidate prey with the largest weight is used as the tracking prey Y, and the displacement vector V = ra × (YX) of hunter X in the direction of the tracking prey is calculated; where ra is a random number between (0, 1); based on the displacement vector V, the new position X_new = X + V of hunter X is updated; Calculate the hunting value corresponding to the new position X_new. If this hunting value is greater than the hunting value of the original hunter X, update the position of hunter X to X_new. Obtain the optimal hunter group. Preset the exchange probability PR, randomly select one of the irrelevant hunters A from all the irrelevant hunters, randomly select a hunter B from the hunters whose positions have been updated, and use the combination of A and B with probability PR to form a new hunter C to replace the irrelevant hunter A; repeat until all the remaining irrelevant hunters are replaced to obtain a new generation of hunters; merge the optimal hunter group with the new generation of hunters to form a new hunter group; repeat the formation of the new hunter group until the preset maximum number of iterations is reached; from the final new hunter group, select the hunter with the highest hunting value, and the corresponding parameter combination is the optimal parameter combination obtained by solution.

8. The intelligent printing process multi-parameter monitoring and optimization system according to claim 7, characterized in that: The registration accuracy score ;in, For the The registration error of each measuring point is is the preset maximum registration error threshold, and All are adjustment parameters greater than 0; For the The distance from each measuring point to the edge of the printed part; dmax is the maximum value of the distances from all measuring points to the edge of the printed part; is a penalty coefficient greater than 0; Imprint surface gloss score ;in, For the The weight of each measurement point, For the Gloss value of each measuring point, For the Gloss gradient value of each measuring point; is the gradient adjustment parameter, and its value is greater than or equal to 0; is the distribution penalty coefficient, and ; is the standard deviation of the gloss values ​​of all measurement points; is the preset reference gloss standard deviation.

Images