Closed-loop control system for thermal management path planning based on laser fuse additive manufacturing

Abstract

The present invention relates to the field of path planning technology, and discloses a closed-loop control system for path planning based on thermal management of laser fuse additive manufacturing, including a data acquisition module, a data processing module, a temperature distribution prediction module and a path planning control module: the data acquisition module is used to divide the processing area into M sub-areas, and obtain temperature distribution characteristic data and first characteristic data; the data processing module is used to pre-process the temperature distribution characteristic data and the first characteristic data to obtain a regional characteristic sequence; the temperature distribution prediction module is used to construct a temperature distribution prediction model according to the regional characteristic sequence; the path planning control module is used to generate a control scheme in combination with an improved Parrot optimization algorithm and a temperature distribution prediction model. The present invention realizes global optimization and dynamic control of the processing process by combining temperature prediction, path optimization, multidimensional feature extraction and an improved optimization algorithm.

Description

Technical Field

[0001] The present invention belongs to the technical field of path planning, and in particular relates to a closed-loop control system for thermal management path planning based on laser fuse additive manufacturing. Background Art

[0002] Laser fuse additive manufacturing is a high-precision and high-efficiency processing technology that uses a laser beam to melt the wire material and deposit it layer by layer. However, this technology has significant thermal management challenges in practical applications, which are mainly closely related to the optimization of processing path planning.

[0003] During the laser fuse processing, the high energy density of the laser beam causes the temperature of the local area to rise rapidly, making it difficult for the heat to diffuse in time. This heat accumulation effect may cause a series of problems such as excessive residual stress, workpiece deformation, internal porosity, etc., which seriously affect the processing quality and workpiece performance. Traditional thermal management methods usually reduce heat accumulation by adjusting the laser power and scanning speed, but this type of method based on fixed parameters is difficult to adapt to the dynamic thermal changes in the processing process in real time, resulting in limited optimization effect.

[0004] Processing path planning is one of the key factors in thermal management. Traditional path planning mostly uses fixed geometric rules (such as parallel lines, spiral scanning) or simple heuristic algorithms to generate, lacking consideration of dynamic temperature distribution and heat diffusion characteristics, which can easily lead to path intersections or local repeated processing, exacerbating heat accumulation. At the same time, there is a lack of coordinated optimization of path planning and process parameters such as laser power and scanning speed, making it difficult to balance thermal management and path efficiency. Although some studies have introduced temperature field simulation and multi-objective optimization techniques, they still face problems of insufficient global optimization capabilities and weak dynamic adaptability under complex geometric shapes, material properties and real-time requirements. Summary of the invention

[0005] The present invention provides a closed-loop control system for thermal management path planning based on laser fuse additive manufacturing, which solves the technical problem of lack of dynamic adaptability in the coordinated optimization of processing path planning and thermal management in related technologies.

[0006] The present invention provides a closed-loop control system for thermal management path planning based on laser fuse additive manufacturing, including a data acquisition module, a data processing module, a temperature distribution prediction module and a path planning control module:

[0007] A data acquisition module is used to divide the processing area into M sub-areas, and obtain temperature distribution characteristic data and first characteristic data at a preset time interval d within a first preset time period T, wherein the temperature distribution characteristic data is composed of the temperature splicing of the M sub-areas, and the first characteristic data includes: laser power, scanning speed, two-dimensional coordinates of the current path point and path direction, and T, d and M are custom parameters;

[0008] A data processing module, used for preprocessing the temperature distribution characteristic data and the first characteristic data to obtain a regional characteristic sequence;

[0009] The temperature distribution prediction module is used to construct a temperature distribution prediction model according to the regional feature sequence. The input of the temperature distribution prediction model is the regional feature sequence, and the output is the temperature distribution feature data within the second preset time period H, where H is a custom parameter;

[0010] The path planning and control module is used to generate a control plan by combining the improved Parrot optimization algorithm and the temperature distribution prediction model. The control plan includes: the processing sequence of the laser head in the sub-area, the laser power and the scanning speed.

[0011] Further, the temperature distribution characteristic data and the first characteristic data are preprocessed to obtain a regional characteristic sequence, and the specific steps include:

[0012] Step S201, extracting the thermal response efficiency feature according to the temperature distribution characteristic data and the first characteristic data, the calculation formula of the thermal response efficiency feature is: ,in, represents the thermal response efficiency characteristics of the ith sub-region at the tth time point, and represents the temperature at the t-th time point and the t-1-th time point of the i-th sub-region, represents the laser power at the tth time point, represents the scanning speed at the tth time point;

[0013] Step S202, extracting the path thermal gradient effect feature according to the temperature distribution feature data and the first feature data, the calculation formula of the path thermal gradient effect feature is: ,in, represents the path thermal gradient effect characteristics of the i-th sub-region at the t-th time point, Indicates the path direction of the laser head at the tth time point, in radians, represents the partial derivative of temperature T in the x direction, represents the partial derivative of temperature in the y direction, and The cosine and sine values ​​representing the direction of the path;

[0014] Step S203, extracting the thermal field power efficiency feature according to the temperature distribution feature data and the first feature data, the calculation formula of the thermal field power efficiency feature is: ,in, represents the thermal field power efficiency characteristics of the ith sub-region at the tth time point, and represents the laser power at the t-th time point and the t-1-th time point, represents the set of sub-regions adjacent to the ith sub-region, represents the absolute value of the temperature difference between the ith sub-region and the adjacent jth sub-region, represents the distance weight coefficient, represents the Euclidean distance between the ith sub-region and the adjacent jth sub-region, represents the power change rate weight coefficient;

[0015] Step S204, concatenating the thermal response efficiency characteristics, the path thermal gradient effect characteristics, the thermal field power efficiency characteristics, the temperature distribution characteristic data and the first characteristic data, and normalizing them using a z-score normalization method to obtain a regional characteristic sequence.

[0016] Further, the temperature distribution prediction model is composed of a first hidden layer, a second hidden layer, a generator, and a first classifier;

[0017] The first hidden layer includes N first units, the nth first unit inputs the nth sequence unit of the regional feature sequence, and outputs the first hidden feature, and the N first hidden features are represented by the hidden feature sequence, where N=T / d, 1≤n≤N;

[0018] The second hidden layer includes N second units, the nth second unit inputs the nth first hidden feature of the hidden feature sequence, and outputs the nth weight feature;

[0019] The generator is used to concatenate the N weight features output by the N second units to obtain the first updated feature;

[0020] The first updated feature is input into the first classifier, and the classification space of the first classifier represents the temperature distribution feature data within the second preset time period H.

[0021] Furthermore, the first unit of the first hidden layer is constructed based on a gated recurrent unit.

[0022] Furthermore, the calculation formula of the second unit is:

[0023] in, represents the nth weight feature, represents the scaling factor, represents the first intermediate feature obtained by linearly transforming the nth first hidden feature, and represents the second intermediate feature and the third intermediate feature obtained by linearly transforming the jth and kth first hidden features, represents the distance weight coefficient, and represents the Euclidean distance between the first intermediate feature and the second intermediate feature and the third intermediate feature respectively, represents the fourth intermediate feature obtained by linearly transforming the jth first hidden feature, represents the adjustment coefficient of the residual connection, represents the linear transformation matrix, , and denote the nth, jth and kth first hidden features respectively, , and represent the first weight parameter, the second weight parameter and the third weight parameter respectively, , , and They represent the first bias parameter, the second bias parameter, the third bias parameter and the fourth bias parameter respectively, tanh represents the tanh activation function, and Indicates that the second intermediate feature and the third intermediate feature are transposed.

[0024] Furthermore, the improved mean square error function is used as the loss function of the temperature distribution prediction model. The improvement of the mean square error function is: adding spatial smoothing constraints on the basis of the standard mean square error function, and the calculation formula of the loss function is: , where G represents the number of time points in the second preset time period, represents the predicted temperature of the nth sub-region at the t+gth time point, represents the actual temperature of the nth sub-area at the t+gth time point, represents the predicted temperature of the mth sub-region at the t+gth time point, represents the set of adjacent sub-regions of the nth sub-region, represents the neighborhood consistency weight coefficient, and N represents the number of time points.

[0025] Furthermore, a control scheme is generated based on the improved Parrot optimization algorithm combined with the temperature distribution prediction model. The specific steps include:

[0026] Step S301, initializing a population satisfying constraint conditions based on chaotic mapping, the population consisting of P individuals;

[0027] Each individual is represented by a feature vector Indicates that the feature vector V represents the number of the sub-areas from the first to the Mth in the processing order, the laser power and the scanning speed, Indicates the processing order as the number, laser power and scanning speed of the Mth sub-area;

[0028] Step S302, constructing an objective function according to the temperature distribution prediction model;

[0029] Step S303, calculating the fitness value of each individual in the population according to the objective function;

[0030] Step S304, simulating the parrot behavior of the individual and updating the characteristic vector of the individual in the population;

[0031] Step S305, the current number of iterations is increased by 1. When it is determined that the current number of iterations reaches the maximum number of iterations, the current optimal individual code is output as the final control solution, otherwise, it returns to step S303, where the maximum number of iterations is a custom parameter.

[0032] Furthermore, a population satisfying the constraint condition is initialized based on the chaotic map, and the specific steps include:

[0033] Step S3011: set a chaotic sequence for each individual and generate a random initial value between 0 and 1 for the chaotic sequence. , the chaotic sequence is passed express;

[0034] Step S3012, iterating the chaotic sequence according to the Logistic chaotic mapping formula, the Logistic chaotic mapping formula is: ,in, and represents the i-th and i+1-th sequence values ​​in the chaotic sequence, represents the chaotic characteristic coefficient;

[0035] Step S3013, arranging the chaotic sequence in ascending order according to the sequence value, and determining the processing order of the sub-regions according to the index of the sorted sequence value.

[0036] Furthermore, the objective function in step S302 is:

[0037] ;

[0038] in, represents the fitness value of an individual, and denote the heat accumulation weight coefficient and the path length weight coefficient respectively, represents the temperature prediction value of the i-th sub-region obtained by the temperature distribution prediction model, Indicates the preset target temperature value. represents the Euclidean distance between the i-th subregion and the i+1-th subregion.

[0039] Furthermore, the specific steps of step S304 include:

[0040] Step S3041, simulating the random flight behavior of the parrot when searching for food, and combining the characteristics of the temperature distribution and the processing path, updating the feature vector of the individual, the calculation formula for the feature vector update is: ,in, and denote the eigenvectors of the i-th individual at the b-th iteration and the b+1-th iteration, respectively. and To independently draw two random variables from a standard normal distribution, represents the scaling factor, and Heat accumulation function The gradient and path length function The gradient of

[0041] Step S3042, simulating the behavior of a parrot stopping in a fixed area to observe the environment, and making targeted adjustments to the individual feature vectors. The calculation formula for the adjustment is: ,in, Indicates the optimal solution influence coefficient, ranging from 0 to 1. represents the optimal solution of the population at the bth iteration, Represents a random number between 0 and 1;

[0042] Step S3043, simulating the process of parrots learning by observing unfamiliar behaviors, introducing a diversity perturbation mechanism, and performing random mutation updates on the individual feature vectors. The calculation formula for random mutation updates is: ,in, represents the disturbance coefficient, represents a random vector from a normal distribution.

[0043] The beneficial effects of the present invention are as follows: the present invention combines the temperature distribution prediction model with the improved Parrot optimization algorithm to dynamically adjust the laser power, scanning speed and processing path to avoid the problem of heat accumulation, and considers the coupling relationship between the path and the temperature to make the heat diffusion more uniform, while reducing the path duplication and intersection, thereby improving the processing efficiency and workpiece quality;

[0044] The present invention extracts multi-dimensional features such as thermal response efficiency, path thermal gradient effect, and thermal field power efficiency to refine the temperature distribution in time and space, which helps to achieve more accurate thermal management and path planning in complex processing scenarios;

[0045] The present invention uses an initialization method based on chaos mapping and an improved parrot optimization algorithm to enhance global search capability and local optimization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1It is a module schematic diagram of a closed-loop control system for thermal management path planning based on laser fuse additive manufacturing of the present invention;

[0047] Figure 2 is a flow chart of preprocessing the temperature distribution characteristic data and the first characteristic data of the present invention;

[0048] Figure 3 It is a flow chart of the improved parrot optimization algorithm of the present invention.

[0049] In the figure: a data acquisition module 101, a data processing module 102, a temperature distribution prediction module 103 and a path planning and control module 104. DETAILED DESCRIPTION

[0050] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0051] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0052] like Figure 1-Figure 3 As shown, the closed-loop control system for thermal management path planning based on laser fuse additive manufacturing includes a data acquisition module 101, a data processing module 102, a temperature distribution prediction module 103 and a path planning control module 104:

[0053] The data acquisition module 101 is used to divide the processing area into M sub-areas, and obtain temperature distribution characteristic data and first characteristic data at a preset time interval d within a first preset time period T, wherein the temperature distribution characteristic data is composed of the temperature splicing of the M sub-areas, and the first characteristic data includes: laser power, scanning speed, two-dimensional coordinates of the current path point and path direction, and T, d and M are custom parameters;

[0054] The data processing module 102 is used to pre-process the temperature distribution characteristic data and the first characteristic data to obtain a regional characteristic sequence;

[0055] The temperature distribution prediction module 103 is used to construct a temperature distribution prediction model according to the regional feature sequence. The input of the temperature distribution prediction model is the regional feature sequence, and the output is the temperature distribution feature data within the second preset time period H, where H is a custom parameter;

[0056] The path planning and control module 104 is used to generate a control scheme by combining the improved Parrot optimization algorithm and the temperature distribution prediction model. The control scheme includes: the processing sequence of the laser head in the sub-area, the laser power and the scanning speed.

[0057] In one embodiment of the present invention, the processing area is divided according to its geometric characteristics; an infrared thermal imager is used to obtain a temperature map of the entire processing area, covering all sub-areas, and temperature data is distributed to each sub-area through pixel mapping; and the output laser power and the scanning speed of the laser head are obtained through a laser device.

[0058] In one embodiment of the present invention, the temperature distribution characteristic data and the first characteristic data are preprocessed to obtain a regional characteristic sequence, and the specific steps include:

[0059] Step S201, extracting the thermal response efficiency feature according to the temperature distribution characteristic data and the first characteristic data, the calculation formula of the thermal response efficiency feature is: ,in, represents the thermal response efficiency characteristics of the ith sub-region at the tth time point, and represents the temperature at the t-th time point and the t-1-th time point of the i-th sub-region, represents the laser power at the tth time point, represents the scanning speed at the tth time point;

[0060] Step S202, extracting the path thermal gradient effect feature according to the temperature distribution feature data and the first feature data, the calculation formula of the path thermal gradient effect feature is: ,in, represents the path thermal gradient effect characteristics of the i-th sub-region at the t-th time point, Indicates the path direction of the laser head at the tth time point, in radians, represents the partial derivative of temperature T in the x direction, that is, the rate of change of temperature T when the x coordinate changes slightly, represents the partial derivative of temperature in the y direction, that is, the rate of change of temperature T when the y coordinate changes slightly, and The cosine and sine values ​​represent the path direction, i.e., the temperature gradient is projected onto the direction of laser head movement;

[0061] Step S203, extracting the thermal field power efficiency feature according to the temperature distribution feature data and the first feature data, the calculation formula of the thermal field power efficiency feature is: ,in, represents the thermal field power efficiency characteristics of the ith sub-region at the tth time point, and represents the laser power at the t-th time point and the t-1-th time point, represents the set of sub-regions adjacent to the ith sub-region, represents the absolute value of the temperature difference between the ith sub-region and the adjacent jth sub-region, represents the distance weight coefficient, represents the Euclidean distance between the ith sub-region and the adjacent jth sub-region, represents the power change rate weight coefficient;

[0062] Step S204, concatenating the thermal response efficiency characteristics, the path thermal gradient effect characteristics, the thermal field power efficiency characteristics, the temperature distribution characteristic data and the first characteristic data, and normalizing them using a z-score normalization method to obtain a regional characteristic sequence.

[0063] By extracting the thermal response efficiency characteristics, path thermal gradient effect characteristics and thermal field power efficiency characteristics, the effect of temperature distribution prediction can be significantly improved. Specifically, it refines the prediction of temperature distribution in time and space, enhances the modeling capability of dynamic changes in complex machining processes, and provides a quantitative basis for path planning and thermal management.

[0064] In one embodiment of the present invention, the temperature distribution prediction model is composed of a first hidden layer, a second hidden layer, a generator and a first classifier;

[0065] The first hidden layer includes N first units, the nth first unit inputs the nth sequence unit of the regional feature sequence, and outputs the first hidden feature, and the N first hidden features are represented by the hidden feature sequence, where N=T / d, 1≤n≤N;

[0066] The second hidden layer includes N second units, the nth second unit inputs the nth first hidden feature of the hidden feature sequence, and outputs the nth weight feature;

[0067] The generator is used to concatenate the N weight features output by the N second units to obtain the first updated feature;

[0068] The first updated feature is input into the first classifier, and the classification space of the first classifier represents the temperature distribution feature data within the second preset time period H.

[0069] In one embodiment of the present invention, the first unit of the first hidden layer is constructed based on a gated recurrent unit.

[0070] In one embodiment of the present invention, the calculation formula of the second unit is:

[0071]

[0072] in, represents the nth weight feature, represents the scaling factor, represents the first intermediate feature obtained by linearly transforming the nth first hidden feature, and represents the second intermediate feature and the third intermediate feature obtained by linearly transforming the jth and kth first hidden features, represents the distance weight coefficient, and represents the Euclidean distance between the first intermediate feature and the second intermediate feature and the third intermediate feature respectively, represents the fourth intermediate feature obtained by linearly transforming the jth first hidden feature, represents the adjustment coefficient of the residual connection, represents the linear transformation matrix, , and denote the nth, jth and kth first hidden features respectively, , and represent the first weight parameter, the second weight parameter and the third weight parameter respectively, , , and They represent the first bias parameter, the second bias parameter, the third bias parameter and the fourth bias parameter respectively, tanh represents the tanh activation function, and Indicates that the second intermediate feature and the third intermediate feature are transposed;

[0073] By calculating the weight of the first hidden feature generated by the first hidden layer at each time point, the temperature distribution prediction model can dynamically adjust the degree of attention to different time points, thereby effectively filtering out noise information and focusing computing resources on key time steps. For example, when the laser power at the tth time point changes significantly and has a greater impact on the current temperature, the model can give this time point a higher weight through weight calculation, thereby enhancing its contribution to temperature prediction. This dynamic weight allocation mechanism can capture complex time dependencies and improve the prediction accuracy and robustness of the model.

[0074] In one embodiment of the present invention, an improved mean square error function is used as the loss function of the temperature distribution prediction model. The improvement of the mean square error function is: adding a spatial smoothing constraint on the basis of the standard mean square error function to avoid local outliers in the prediction. The calculation formula of the loss function is: , where G represents the number of time points in the second preset time period, represents the predicted temperature of the nth sub-region at the t+gth time point, represents the actual temperature of the nth sub-area at the t+gth time point, represents the predicted temperature of the mth sub-region at the t+gth time point, represents the set of adjacent sub-regions of the nth sub-region, represents the neighborhood consistency weight coefficient, and N represents the number of time points.

[0075] In one embodiment of the present invention, a control scheme is generated according to the improved Parrot optimization algorithm combined with the temperature distribution prediction model, and the specific steps include:

[0076] Step S301, initializing a population satisfying constraint conditions based on chaotic mapping, the population consisting of P individuals;

[0077] Each individual is represented by a feature vector Indicates that the feature vector V represents the number of the sub-areas from the first to the Mth in the processing order, the laser power and the scanning speed, Indicates the processing order as the number of the Mth sub-area, laser power and scanning speed, for example, , it means the processing order is that the number of the second sub-area is 3, the laser power is 200W, and the scanning speed is 400mm / s;

[0078] The constraints include: each sub-region is processed at least once;

[0079] The laser power cannot exceed the allowed range of the equipment;

[0080] The scanning speed cannot exceed the allowed range of the device;

[0081] During the processing, the temperature of each sub-area cannot exceed the maximum temperature allowed by the material;

[0082] For sub-areas with a fixed order in the machining path, the sub-areas need to be machined in a fixed order, such as machining along a certain axis in sequence;

[0083] Step S302, constructing an objective function according to the temperature distribution prediction model;

[0084] Step S303, calculating the fitness value of each individual in the population according to the objective function;

[0085] Step S304, simulating the parrot behavior of the individual and updating the characteristic vector of the individual in the population;

[0086] Step S305, the current number of iterations is increased by 1. When it is determined that the current number of iterations reaches the maximum number of iterations, the current optimal individual code is output as the final control solution, otherwise, the process returns to step S303, where the maximum number of iterations is a custom parameter. Preferably, the maximum number of iterations is set to 200.

[0087] In one embodiment of the present invention, a population satisfying constraint conditions is initialized based on a chaotic map, and the specific steps include:

[0088] Step S3011: set a chaotic sequence for each individual and generate a random initial value between 0 and 1 for the chaotic sequence. , the chaotic sequence is passed express;

[0089] Step S3012, iterating the chaotic sequence according to the Logistic chaotic mapping formula, the Logistic chaotic mapping formula is: ,in, and represents the i-th and i+1-th sequence values ​​in the chaotic sequence, represents the chaotic characteristic coefficient, preferably, Set to 4;

[0090] Step S3013, the chaotic sequence is sorted in ascending order according to the sequence value, and the processing order of the sub-regions is determined according to the index of the sorted sequence value. For example, If it is in the first place of the chaotic sequence, the third sub-region will be processed first in the processing order of the sub-regions.

[0091] In one embodiment of the present invention, the objective function in step S302 is:

[0092] ;

[0093] in, represents the fitness value of an individual, and denote the heat accumulation weight coefficient and the path length weight coefficient respectively, represents the temperature prediction value of the i-th sub-region obtained by the temperature distribution prediction model, Indicates the preset target temperature value. represents the Euclidean distance between the i-th subregion and the i+1-th subregion.

[0094] In one embodiment of the present invention, the specific steps of step S304 include:

[0095] Step S3041, simulating the random flight behavior of the parrot when searching for food, and combining the characteristics of the temperature distribution and the processing path, updating the feature vector of the individual, the calculation formula for the feature vector update is: ,in, and denote the eigenvectors of the i-th individual at the b-th iteration and the b+1-th iteration, respectively. and To independently draw two random variables from a standard normal distribution, represents the scaling factor, preferably, Set to 1.5, and Heat accumulation function The gradient and path length function The gradient of

[0096] Step S3042, simulating the behavior of a parrot stopping in a fixed area to observe the environment, and making targeted adjustments to the individual feature vectors. The calculation formula for the adjustment is: ,in, Indicates the optimal solution influence coefficient, ranging from 0 to 1. represents the optimal solution of the population at the bth iteration, Represents a random number between 0 and 1;

[0097] Step S3043, simulating the process of parrots learning by observing unfamiliar behaviors, introducing a diversity perturbation mechanism, and performing random mutation updates on the individual feature vectors. The calculation formula for random mutation updates is: ,in, represents the disturbance coefficient, represents a random vector from a normal distribution.

[0098] The improved Parrot optimization algorithm improves the adaptability of the Parrot optimization algorithm by combining the characteristics of temperature distribution and processing path, and by adding random disturbance terms

[0099] , maintaining a certain search flexibility, preventing individuals from completely gathering near the optimal solution, and introducing random mutations based on the optimal solution, giving the population a chance to jump out of the local optimal solution and explore a better solution.

[0100] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are protected by the present embodiment.

Claims

1. A closed-loop control system for thermal management path planning based on laser fuse additive manufacturing, characterized in that: Including data acquisition module, data processing module, temperature distribution prediction module and path planning and control module: A data acquisition module is used to divide the processing area into M sub-areas, and obtain temperature distribution characteristic data and first characteristic data at a preset time interval d within a first preset time period T, wherein the temperature distribution characteristic data is composed of the temperature splicing of the M sub-areas, and the first characteristic data includes: laser power, scanning speed, two-dimensional coordinates of the current path point and path direction, and T, d and M are custom parameters; A data processing module is used to pre-process the temperature distribution characteristic data and the first characteristic data to obtain a regional characteristic sequence, wherein a thermal response efficiency characteristic, a path thermal gradient effect characteristic and a thermal field power efficiency characteristic are extracted according to the temperature distribution characteristic data and the first characteristic data; The temperature distribution prediction module is used to construct a temperature distribution prediction model according to the regional feature sequence. The input of the temperature distribution prediction model is the regional feature sequence, and the output is the temperature distribution feature data within the second preset time period H, where H is a custom parameter; The path planning and control module is used to generate a control plan by combining the improved Parrot optimization algorithm and the temperature distribution prediction model. The specific steps include: Step S301, based on the chaotic map, a population satisfying the constraint conditions is initialized, the population is composed of P individuals, and the specific steps include: Step S3011, set a chaotic sequence for each individual, and generate a random initial value z0 between 0 and 1 for the chaotic sequence; Step S3012, iterating the chaotic sequence according to the Logistic chaotic mapping formula, the Logistic chaotic mapping formula is: i+1 =μ*z i *(1-z i ), where z i and z i+1 represents the i-th and i+1-th sequence values ​​in the chaotic sequence, μ represents the chaotic characteristic coefficient; Step S3013, arranging the chaotic sequence in ascending order according to the sequence value, and determining the processing order of the sub-regions according to the index of the sorted sequence value; Step S302: construct an objective function based on the temperature distribution prediction model. The objective function is: Among them, Fitness represents the fitness value of the individual, w1 and w2 represent the heat accumulation weight coefficient and path length weight coefficient respectively, T i represents the temperature prediction value of the ith sub-region obtained by the temperature distribution prediction model, T d Indicates the preset target temperature value, dist(A i , A i+1 ) represents the Euclidean distance between the i-th sub-region and the i+1-th sub-region; Step S303, calculating the fitness value of each individual in the population according to the objective function; Step S304, simulating the parrot behavior of the individual and updating the characteristic vector of the individual in the population; Step S305, the current number of iterations is increased by 1, and when it is determined that the current number of iterations reaches the maximum number of iterations, the current optimal individual code is output as the final control solution, otherwise, the process returns to step S303; The control scheme includes: the processing sequence of the laser head in the sub-areas, the laser power and the scanning speed.

2. The closed-loop control system for thermal management path planning based on laser fuse additive manufacturing according to claim 1 is characterized in that: The temperature distribution characteristic data and the first characteristic data are preprocessed to obtain a regional characteristic sequence. The specific steps include: Step S201, extracting the thermal response efficiency feature according to the temperature distribution characteristic data and the first characteristic data, the calculation formula of the thermal response efficiency feature is: in, represents the thermal response efficiency characteristics of the ith sub-region at the tth time point, and represents the temperature of the t-th time point and the t-1-th time point of the ith sub-region, P t represents the laser power at the tth time point, V t represents the scanning speed at the tth time point; Step S202, extracting the path thermal gradient effect feature according to the temperature distribution feature data and the first feature data, the calculation formula of the path thermal gradient effect feature is: in, represents the path thermal gradient effect characteristics of the i-th sub-region at the t-th time point, θ t Indicates the path direction of the laser head at the tth time point, in radians, represents the partial derivative of temperature T in the x direction, represents the partial derivative of temperature in the y direction, cos(θ t ) and sin(θ t ) represents the cosine and sine values ​​of the path direction; Step S203, extracting the thermal field power efficiency feature according to the temperature distribution feature data and the first feature data, the calculation formula of the thermal field power efficiency feature is: in, represents the thermal field power efficiency characteristics of the ith sub-region at the tth time point, P t and P t-1 represents the laser power at the t-th time point and the t-1-th time point, Ne(i) represents the set of sub-regions adjacent to the i-th sub-region, represents the absolute value of the temperature difference between the ith sub-region and the adjacent jth sub-region, β represents the distance weight coefficient, and d i，j represents the Euclidean distance between the ith sub-region and the adjacent jth sub-region, and α represents the power change rate weight coefficient; Step S204, concatenating the thermal response efficiency characteristics, the path thermal gradient effect characteristics, the thermal field power efficiency characteristics, the temperature distribution characteristic data and the first characteristic data, and normalizing them using a z-score normalization method to obtain a regional characteristic sequence.

3. The closed-loop control system for thermal management path planning based on laser fuse additive manufacturing according to claim 1 is characterized in that: The temperature distribution prediction model consists of a first hidden layer, a second hidden layer, a generator and a first classifier; The first hidden layer includes N first units, the nth first unit inputs the nth sequence unit of the regional feature sequence, and outputs the first hidden feature, and the N first hidden features are represented by the hidden feature sequence, where N=T / d, 1≤n≤N; The second hidden layer includes N second units, the nth second unit inputs the nth first hidden feature of the hidden feature sequence, and outputs the nth weight feature; The generator is used to concatenate the N weight features output by the N second units to obtain the first updated feature; The first updated feature is input into the first classifier, and the classification space of the first classifier represents the temperature distribution feature data within the second preset time period H.

4. The closed-loop control system for thermal management path planning based on laser fuse additive manufacturing according to claim 3 is characterized in that: The first unit of the first hidden layer is built based on a gated recurrent unit.

5. The closed-loop control system for thermal management path planning based on laser fuse additive manufacturing according to claim 3 is characterized in that: The calculation formula for the second unit is: Q n =W q *h n +b q ; K k =W k *h k +b k ; V j =W v *h j +b v ; Among them, A(h n ) represents the nth weight feature, γ represents the scaling factor, Q n represents the first intermediate feature obtained by linearly transforming the nth first hidden feature, K j and K k represents the second intermediate feature and the third intermediate feature obtained by linearly transforming the jth and kth first hidden features, λ represents the distance weight coefficient, ||Q n -K j ||2 and ||Q n -K k ||2 represents the Euclidean distance between the first intermediate feature and the second intermediate feature and the third intermediate feature, respectively. V j represents the fourth intermediate feature obtained by linearly transforming the jth first hidden feature, δ represents the adjustment coefficient of the residual connection, and W h represents the linear transformation matrix, h n 、h j and h k denote the nth, jth and kth first hidden features respectively, W q , W k and W v Respectively represent the first weight parameter, the second weight parameter and the third weight parameter, b q , b k and b v Respectively represent the first bias parameter, the second bias parameter, the third bias parameter and the fourth bias parameter, tanh represents the tanh activation function, K j T and K k T Indicates that the second intermediate feature and the third intermediate feature are transposed.

6. The closed-loop control system for thermal management path planning based on laser fuse additive manufacturing according to claim 1 is characterized in that: The improved mean square error function is used as the loss function of the temperature distribution prediction model. The improvement of the mean square error function is: adding spatial smoothing constraints on the basis of the standard mean square error function. The calculation formula of the loss function is: Wherein, G represents the number of time points in the second preset time period, represents the predicted temperature of the nth sub-region at the t+gth time point, represents the actual temperature of the nth sub-area at the t+gth time point, represents the predicted temperature of the mth sub-region at the t+gth time point, m∈Ne(n) represents the set of adjacent sub-regions of the nth sub-region, β represents the neighborhood consistency weight coefficient, and N represents the number of time points.

7. The closed-loop control system for thermal management path planning based on laser fuse additive manufacturing according to claim 1 is characterized in that: The specific steps of step S304 include: Step S3041, simulating the random flight behavior of the parrot when searching for food, and combining the characteristics of the temperature distribution and the processing path, updating the feature vector of the individual, the calculation formula for the feature vector update is: in, and They represent the eigenvectors of the i-th individual at the b-th iteration and the b+1-th iteration, μ and τ are two random variables independently drawn from the standard normal distribution, ρ represents the scaling factor, and Heat accumulation function The gradient and path length function The gradient of Step S3042, simulating the behavior of a parrot stopping in a fixed area to observe the environment, and making targeted adjustments to the individual feature vectors. The calculation formula for the adjustment is: in, Indicates the optimal solution influence coefficient, ranging from 0 to 1, X best represents the optimal solution of the population at the bth iteration, and rand(0,1) represents a random number between 0 and 1; Step S3043, simulating the process of parrots learning by observing unfamiliar behaviors, introducing a diversity perturbation mechanism, and performing random mutation updates on the individual feature vectors. The calculation formula for random mutation updates is: Among them, θ represents the perturbation coefficient, and randn(dim) represents a random vector that obeys the normal distribution.

Images