Self-adaptive correction method and system for concrete 3D printing

Abstract

The invention provides a concrete 3D printing self-adaptive deviation correction method and system, and relates to the technical field of intelligent construction and additive manufacturing, the method comprises the following steps: determining a deviation field matrix based on three-dimensional point cloud data, a two-dimensional texture image and printing parameters of a printed concrete layer; and extracting a two-dimensional texture image based on a convolutional neural network model, determining a defect feature data set, inputting the deviation field matrix, the defect feature data set and the printing parameters into a deformation trend prediction model to obtain a deformation trend set, adjusting the printing parameters of the to-be-printed concrete layer, and generating a target concrete member. The technical problems that traditional concrete 3D printing depends on a preset program, dynamic changes of material rheological characteristics cannot be sensed in real time, the extrusion shape deviates from a design target, interlayer dislocation and deformation can be accumulated layer by layer, and structural instability is caused are solved, and the effect of improving the stability of a printed and formed concrete member is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent construction and additive manufacturing, and particularly relates to a concrete 3D printing self-adaptive deviation correction method and system. BACKGROUND

[0002] With the accelerated development of intelligent construction and building industrialization, the traditional construction of concrete structures mainly relies on mold pouring. This method needs to pre-produce fixed molds, and has problems of complicated process, long cycle and high cost, especially it is difficult to adapt to the construction demand of complex special-shaped structures, and cannot meet the requirements of flexibility, adaptability and construction efficiency of modern buildings.

[0003] To solve the limitations of traditional pouring, concrete 3D printing technology has gradually emerged. This printing technology has advantages of no need for molds, high material utilization rate, and customized molding, and realizes structure forming through a material deposition process of layer-by-layer accumulation, thereby providing a new path for rapid construction of complex structures. However, the 3D printing technology currently mainly relies on fixed parameters such as pre-set printing path and extrusion speed to complete printing, and lacks real-time monitoring and dynamic adjustment capability of the printing process. This means that in actual application, it cannot perceive the dynamic changes of the rheological properties of the concrete mixture caused by factors such as raw material batch and temperature, thereby causing deviations between the extrusion shape and the design target, and the small misalignment or deformation of each layer in the printing process will be accumulated through the interlayer stacking effect, eventually causing the upper layer to lose stability and leading to the failure of the overall structure. SUMMARY

[0004] In view of the above deficiencies in the prior art, the present application aims to provide a concrete 3D printing self-adaptive deviation correction method, which has the characteristics of improving the stability of the printed concrete member.

[0005] The above-mentioned first application of the present application is realized by the following technical scheme: A concrete 3D printing self-adaptive deviation correction method, comprising: real-time acquisition of three-dimensional point cloud data, two-dimensional texture images and printing parameters of the printed concrete layer; determination of a deviation field matrix based on the three-dimensional point cloud data and the theoretical digital model; extraction of surface defect features of the two-dimensional texture images based on a convolutional neural network model to determine a defect feature dataset; input of the deviation field matrix, the defect feature dataset and the printing parameters into a deformation trend prediction model to determine a deformation trend set of the to-be-printed concrete layer; adjustment of the printing parameters of the to-be-printed concrete layer based on the deformation trend set to generate a target concrete member.

[0006] By adopting the technical scheme, based on the real-time collected three-dimensional morphology of the printed layer of concrete, surface texture and printing parameters, combined with the design reference provided by the theoretical digital model, a deviation field and defect feature dataset integrating multi-source perception information are constructed and input into a deformation trend prediction model to generate a deformation trend set for guiding the printing parameters of the next layer, and finally closed-loop adaptive correction is realized. The problems of inconsistent extrusion shape, cumulative layer misplacement caused by time-varying rheological properties of concrete and environmental disturbance, and printing body deformation, size out-of-tolerance and structure instability caused by lack of real-time online control means in traditional open-loop printing are effectively solved, the forming precision and stability of concrete 3D printing are improved, and when applied to the building scene, the labor cost is reduced while the work efficiency is effectively improved.

[0007] Preferably, based on the three-dimensional point cloud data and the theoretical digital model, a deviation field matrix is determined, including: The three-dimensional surfaces of the three-dimensional point cloud data and the theoretical digital model are registered by an iterative closest point algorithm to obtain a spatial transformation matrix; The three-dimensional point cloud data is subjected to coordinate transformation based on the spatial transformation matrix to obtain a target point cloud dataset; The deviations of each data point in the target point cloud dataset to the nearest point of the three-dimensional surface are calculated and decomposed to obtain Z-axis deviation values, X-axis deviation values and Y-axis deviation values; The deviation field matrix is determined based on the Z-axis deviation values, the X-axis deviation values and the Y-axis deviation values.

[0008] By adopting the technical scheme, the iterative closest point algorithm is used to realize high-precision automatic registration, and the measured point cloud data and the theoretical model are unified in the same coordinate system. By calculating the nearest point deviation of each data point to the model surface and decomposing it, the deviation components of the printed structure in each direction in the three-dimensional space are accurately obtained, effectively overcoming the technical limitations of low efficiency, large subjective error and difficulty in fully reflecting the three-dimensional morphology deviation of traditional measurement methods.

[0009] Preferably, the surface defect features of the two-dimensional texture image are extracted based on a convolutional neural network model to determine a defect feature dataset, including: The two-dimensional texture image is preprocessed to obtain a target texture image; The target texture image is identified and segmented based on a convolutional neural network model to obtain a defect segmentation mask; Based on the defect segmentation mask, a geometric morphological parameter set of the surface defect features is calculated; Based on the geometric morphological parameter set, a defect feature dataset is generated.

[0010] By adopting the technical scheme, the defects such as collapse, bulge and crack are accurately recognized from the surface texture of the printing layer, and the geometric parameter set is quantified, so that the intelligent perception and digital representation of the apparent quality of the printed concrete are realized, stress concentration and interlayer bonding performance degradation caused by surface defect expansion are effectively inhibited, and the integrity and long-term stability of the printed concrete member are improved.

[0011] Preferably, the specific training process of the deformation trend prediction model is as follows: The printing parameters, the deviation field parameters, the defect feature parameters and the thixotropy parameters are obtained from the historical database and preprocessed to construct a training sample set, wherein the training sample set includes a training set and a test set; The training set is input into an initial prediction model for training to obtain the deformation trend prediction model; The test set is input into the deformation trend prediction model to obtain a deformation trend set; It is judged whether the deformation trend set meets a preset condition, and if so, the current prediction model is taken as the deformation trend prediction model.

[0012] By adopting the technical scheme, the deformation trend prediction model is trained by using multi-source historical printing data, the complex coupling relationship between the printing parameters, the real-time deviation, the surface defects and the material thixotropy is fully mined, the model can accurately learn the dynamic deformation law in the concrete printing process, thereby improving the prediction accuracy and generalization ability of the model under complex working conditions, and further improving the accuracy of the deformation trend set prediction, thereby fundamentally guaranteeing the forming precision and overall structural stability of the concrete member.

[0013] Preferably, the printing parameters, the deviation field parameters, the defect feature parameters and the thixotropy parameters are obtained from the historical database and preprocessed to construct a training sample set, including: The printing parameters, the deviation field parameters, the defect feature parameters and the thixotropy parameters are data cleaned to obtain an initial sample set; the parameters in the initial sample set are standardized to obtain an intermediate sample set; The parameters in the intermediate sample set are time-aligned and feature-recombined to obtain the training sample set.

[0014] By adopting the technical scheme, the multi-source heterogeneous historical data is cleaned, standardized, time-aligned and feature-recombined to construct a high-quality training sample set, thereby effectively solving the model training difficulty problem caused by different data dimensions, time sequence dislocation and feature isolation, providing accurate and consistent input for the deformation trend prediction model, and improving the convergence speed and prediction accuracy of the model.

[0015] Preferably, judging whether the deformation trend set meets the preset condition further includes: If the deformation trend set does not satisfy the preset condition, the hyperparameters of the initial prediction model are obtained, and the hyperparameters are adjusted by a preset step size;The adjusted hyperparameters are used as new hyperparameters, and the step of inputting the training set into the initial prediction model for training is jumped to obtain the deformation trend prediction model until the deformation trend set satisfies the preset condition.

[0016] By adopting the above technical scheme, the hyperparameter self-adaptive adjustment mechanism is introduced in the model training process, the model parameters are automatically iteratively optimized when the prediction output does not satisfy the accuracy requirement, the defects of low efficiency, dependence on experience and easy to fall into local optimum of traditional manual parameter adjustment are effectively overcome, and the convergence speed, generalization ability and reliability of the final output result of the deformation trend prediction model are effectively improved.

[0017] Preferably, the deformation trend set includes path offset, extrusion speed increment and layer height increment. The printing parameters include extrusion speed parameters, printing path parameters and layer height parameters. Adjusting the printing parameters of the to-be-printed concrete layer based on the deformation trend set to generate the target concrete component, comprising: Compensate the printing path parameters based on the path offset to generate the target printing path; Adjust the extrusion speed parameters based on the extrusion speed increment to generate the target extrusion speed; Adjust the layer height parameters based on the layer height increment to generate the target layer height; Print the next concrete layer using the target printing path, the target extrusion speed and the target layer height to generate the target concrete component.

[0018] By adopting the above technical scheme, the deformation trend prediction result is quantified as the accurate compensation amount of the printing path, the extrusion speed and the layer height, the printing deviation caused by the change of material rheological property, environmental disturbance and cumulative error can be actively compensated, the transmission and expansion of the interlayer misplacement are effectively inhibited, the geometric accuracy in three-dimensional space and the interlayer bonding quality of the concrete component are fundamentally guaranteed, and the stability of the printing process and the integrity of the formed component are improved.

[0019] The second object of the present application is to provide a concrete 3D printing adaptive correction system, which has the characteristics of improving the stability of the printed concrete component.

[0020] The above-mentioned second object of the present application is achieved by the following technical scheme: A concrete 3D printing adaptive correction system, comprising: A data acquisition module for acquiring three-dimensional point cloud data, two-dimensional texture images and printing parameters of the printed concrete layer in real time;A calculation module for determining the deviation field matrix based on the three-dimensional point cloud data and the theoretical digital model; The data extraction module is configured to extract surface defect features of the two-dimensional texture image based on the convolutional neural network model, and determine a defect feature dataset; The data prediction module is configured to input the bias field matrix, the defect feature dataset, and the printing parameter into a deformation trend prediction model, and determine a deformation trend set of the to-be-printed concrete layer. The data adjustment module is configured to adjust the printing parameter of the to-be-printed concrete layer based on the deformation trend set, and generate a target concrete component.

[0021] By adopting the above technical solutions, the data acquisition module is configured to acquire the three-dimensional morphology, the surface texture, and the printing parameter of the concrete printing layer in real time, combine the design reference provided by the theoretical digital model, construct a bias field and a defect feature dataset fused with multi-source sensing information, and input the bias field and the defect feature dataset into a deformation trend prediction model, so that the data prediction module generates a deformation trend set for guiding the printing parameter of the next layer, and finally realizes closed-loop adaptive correction. The problems of inconsistent extrusion shape, accumulated layer misplacement, printing body deformation, size out-of-tolerance, and structure instability caused by the time-varying rheological properties of concrete and environmental disturbances in traditional open-loop printing are effectively solved, the forming precision and stability of concrete 3D printing are improved, and the work efficiency is effectively improved while the labor cost is reduced when the application is applied to the building scene.

[0022] The third object of the present application is to provide an electronic device with the characteristic of improving the stability of the printed and formed concrete component.

[0023] The third object of the present application is achieved by the following technical solutions: An electronic device includes a memory and a processor, and the memory stores a computer program capable of being loaded and executed by the processor to perform any of the above-mentioned concrete 3D printing adaptive correction methods.

[0024] The fourth object of the present application is to provide a computer storage medium capable of storing a corresponding program, with the characteristic of facilitating the improvement of the stability of the printed and formed concrete component.

[0025] The fourth object of the present application is achieved by the following technical solutions: A computer readable storage medium stores a computer program capable of being loaded and executed by a processor to perform any of the above-mentioned concrete 3D printing adaptive correction methods.

[0026] In summary, the present application has at least one of the following beneficial technical effects: 1. The present application realizes closed-loop adaptive correction by real-time sensing of the printing layer state and comparison with the theoretical model, effectively suppresses the shape deviation and layer misplacement caused by material time-varying and environmental disturbance, and improves the forming precision and structural integrity of concrete 3D printing. 2. The application uses an iterative closest point algorithm to realize point cloud registration, unifies the coordinate space of the measured data and the theoretical model, calculates and decomposes the three-dimensional deviation of each point to the model surface, accurately identifies the early deformation trend of the printed body in each direction, and thus ensures the structural stability of the concrete member during printing; 3. The application preprocesses the multi-source historical data, constructs high-quality training samples, provides accurate input for the deformation trend prediction model, and thus improves the convergence speed and prediction accuracy of the deformation trend prediction model. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a step flowchart of a concrete 3D printing adaptive correction method provided by an embodiment of the application.

[0028] Figure 2 is a step flowchart of another concrete 3D printing adaptive correction method provided by an embodiment of the application.

[0029] Figure 3 is a structural block diagram of a concrete 3D printing adaptive correction system provided by an embodiment of the application. DETAILED DESCRIPTION

[0030] The embodiment of the application provides a concrete 3D printing adaptive correction method, which is used to solve the technical problem that 3D printing technology currently mainly relies on fixed parameters such as a pre-set printing path and an extrusion speed to complete printing, and lacks real-time monitoring and dynamic adjustment capability during printing. This means that in actual application, it cannot perceive the dynamic change of the rheological properties of concrete mixture caused by factors such as raw material batch and temperature, thereby causing deviation between the extrusion shape and the design target, and meanwhile, the slight misplacement or deformation of each layer during printing will be accumulated through the layer stacking effect, finally causing the instability of the upper layer printing, and leading to the technical problem of the failure of the overall structure. The method has the technical effect of improving the stability of the printed concrete member.

[0031] To make the application purpose, features and advantages of the application more obvious and easy to understand, the technical solutions in the embodiments of the application will be clearly and completely described below in combination with the drawings in the embodiments of the application. Obviously, the following described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0032] It should be noted that the terms "first", "second", and the like in the present disclosure are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the disclosure described herein can be implemented in an order other than that illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure.

[0033] In addition, the term "and / or" herein is only a description of the association relationship between the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent three cases of A alone, A and B together, and B alone. In addition, the character " / " herein generally represents an "or" relationship between the associated objects unless otherwise specified.

[0034] Embodiment one: Please refer to Figure 1 The present disclosure provides a concrete 3D printing self-adaptive correction method, comprising: Step 101, real-time acquisition of three-dimensional point cloud data, two-dimensional texture image and printing parameters of the printed concrete layer.

[0035] The three-dimensional point cloud data is the high-precision three-dimensional topographic data of the printed concrete layer surface obtained in real time by a laser line scanner installed at the rear of the printing head. The model of the laser scanner includes but is not limited to Keyence LJ-V7080, SICK TriSpector1000, etc. The laser line scanner scans at a preset frequency, which includes but is not limited to 10Hz, 20Hz, 50Hz, etc. to ensure coverage of the latest printed 3-5 layers of concrete surface, and the measurement accuracy is not less than ±0.025mm.

[0036] The two-dimensional texture image is a gray or color image of the concrete layer surface collected synchronously by a high-resolution industrial camera arranged in cooperation with the laser scanner, which is used to represent the surface texture, defect distribution and apparent state of the material.

[0037] It is worth mentioning that the industrial camera is equipped with a 35mm fixed focus lens, installed near the laser scanner, with an optical axis at a 45° angle to the printing platform plane, and the acquisition frame rate is synchronized with the printing layer. The printing platform refers to the horizontal reference plane used to carry the printed components in the concrete 3D printing system.

[0038] The printing parameters refer to a set of adjustable variables used to control the running state of the printing equipment and the material output characteristics during the concrete 3D printing process, including but not limited to extrusion speed, printing path coordinates, layer height, material flow rate, printing head moving speed, environmental temperature and humidity, etc. These printing parameters are recorded in real time by the printing control system and uploaded to the data processing center.

[0039] Preferably, the printing parameters of the embodiment are extrusion speed parameter, printing path parameter and layer height parameter.

[0040] The extrusion speed parameter refers to the key process parameter that controls the amount of concrete material extruded from the nozzle of the printing head per unit time, which determines the width and height of the printed line, and is usually achieved by controlling the speed of the extruder servo motor.

[0041] The printing path parameter refers to the coordinate sequence instruction that controls the motion trajectory of the printing head in three-dimensional space.

[0042] The layer height parameter refers to the designed height difference between the top surfaces of adjacent two printed concrete layers in the vertical direction.

[0043] It is worth mentioning that the extrusion speed parameter is calculated by reading the motor speed in real time through the encoder feedback signal of the extruder servo motor, combined with the lead of the screw or the cross-sectional area of the piston. The printing path parameter is directly obtained by the multi-axis position loop feedback of the motion controller, which records the absolute coordinates of the printing head in X, Y and Z axes in real time. The layer height parameter is directly determined by the position instruction value of the Z-axis servo motor.

[0044] The printing control system refers to the computer integrated control unit in the concrete 3D printing equipment for coordinating and executing the printing process. It can accurately drive the three-axis servo motor through the multi-axis motion controller, control the printing head to move along the preset path; adjust the speed and torque of the extruder motor through the servo driver to achieve precise control of the material flow rate; monitor, record and upload the printing parameters to the data processing center in real time; synchronize with the laser scanner and industrial camera, pause the motion and trigger data acquisition after each layer is printed; receive adjustment instructions issued by the data processing center, and dynamically adjust the printing parameters of the next layer.

[0045] In the embodiment of the application, three-dimensional point cloud data, two-dimensional texture images and printing parameters of the printed concrete layer are collected in real time.

[0046] Step 102, based on the three-dimensional point cloud data and the theoretical digital model, determine the deviation field matrix.

[0047] The theoretical digital model is a high-precision three-dimensional digital model that is generated in advance based on building information model (BIM) or computer-aided design (CAD) software and is completely consistent with the target component to be printed. The model defines the ideal geometric shape, size and spatial coordinates of the component, and is usually stored in STL or NURBS surface format.

[0048] The deviation field matrix is a structured deviation data set obtained by registering and calculating the deviation between the measured three-dimensional point cloud data and the theoretical digital model, and the matrix represents the deviation of each position on the printing layer surface in the three-dimensional space from the theoretical model in a grid form.

[0049] In the embodiment of the present application, the deviation field matrix is determined based on the three-dimensional point cloud data and the theoretical digital model.

[0050] In step 103, the surface defect features of the two-dimensional texture image are extracted based on the convolutional neural network model to determine the defect feature data set.

[0051] The defect feature data set is the quantitative characterization data about the surface defects extracted by identifying and segmenting the two-dimensional texture image of the concrete surface using the convolutional neural network model, including but not limited to defect type, location, area, contour parameters, and other geometric features.

[0052] The convolutional neural network model is a pre-trained deep learning model with an encoder-decoder structure that can perform pixel-level segmentation on typical defects such as cracks, holes, and bumps in images.

[0053] In the embodiment of the present application, the pre-trained convolutional neural network model is used to extract the defect feature data set from the surface defect features of the collected two-dimensional texture image.

[0054] In step 104, the deviation field matrix, the defect feature data set, and the printing parameters are input into the deformation trend prediction model to determine the deformation trend set of the to-be-printed concrete layer.

[0055] The deformation trend prediction model is a pre-trained machine learning model with a long short-term memory (LSTM) architecture that can learn the time series coupling relationship between the printing parameters, real-time deviation, surface defects, and material properties.

[0056] The deformation trend set is a quantitative output of the future printing layer deformation development trend predicted by the deformation trend prediction model based on the current printing state, including but not limited to path offset, extrusion speed increment, layer height increment, print head movement acceleration adjustment value, and path smoothness parameter.

[0057] Preferably, the deformation trend set of the present embodiment includes path offset, extrusion speed increment, and layer height increment.

[0058] The path offset refers to the compensation adjustment amount required for the next layer printing path in the horizontal plane (X and Y directions) according to the output of the deformation trend prediction model, with units of millimeters, used to correct the horizontal position deviation caused by the deformation or misplacement of the printed layer, ensuring that the printing contour is consistent with the design model.

[0059] The extrusion speed increment refers to the adjustment amount required for the material extrusion speed in the next layer printing process according to the output of the deformation trend prediction model, with the unit of millimeters per second, which is used to dynamically adjust the material output to adapt to the changes in the extrusion amount caused by the changes in layer height, path offset or material rheological property, so as to prevent over-extrusion or shortage of materials.

[0060] The layer height increment refers to the compensation adjustment amount required for the vertical height (Z direction) of the next layer printing according to the output of the deformation trend prediction model, with the unit of millimeters, which is used to correct the accumulated error of the layer height caused by the unevenness of the current layer surface or the overall deformation, so as to ensure the dimensional accuracy and interlayer bonding quality of the printed component in the vertical direction.

[0061] It is worth mentioning that by inputting the deviation field matrix, defect feature dataset and printing parameters of the current layer into the pre-trained deformation trend prediction model, the deformation trend of future multiple layers can be predicted, and the deformation trend set containing the path offset amount in each direction, the extrusion speed adjustment amount and the layer height compensation amount can be output. Compared with the prior art which can only make adjustments to the current error, the embodiment can effectively overcome the problem of concrete response lag.

[0062] For example, the deformation trend prediction model sets the prediction step N = 5 layers, and the control step The model predicts the deformation trend of the next 5 layers based on the deviation field matrix, defect feature dataset and printing parameters of the current layer, but only optimizes the calculation of the optimal adjustment parameters of the next 2 layers as the output of the deformation trend set. Through this setting, the predictability is ensured, and the computational complexity is controlled, achieving a balance between prediction effect and computational efficiency.

[0063] In the embodiment of the present application, the deviation field matrix, defect feature dataset and printing parameters are input into the deformation trend prediction model to determine the path offset amount, extrusion speed increment and layer height increment of the concrete layer to be printed.

[0064] Step 105, adjusting the printing parameters of the concrete layer to be printed based on the deformation trend set to generate a target concrete component.

[0065] The concrete layer to be printed refers to the concrete material layer that will be printed in the current printing period, which is the direct object that is adjusted and controlled in real time during the printing process of the layer.

[0066] It is worth mentioning that the layer to be printed is a key transition layer that connects the printed part and the subsequent printed layer, and its printing quality directly affects the geometric accuracy and mechanical properties of the overall structure.

[0067] The target concrete component refers to the concrete component that is finally printed and formed by the self-adaptive deviation correction method of the present application.

[0068] It is worth mentioning that a linear concrete wall column with a size of 1000mm (length) x 200mm (width) x 800mm (height) is printed by using the traditional open loop and the method of the embodiment respectively, and the printing results are shown in Table 1.

[0069] Table 1 printing results It can be seen that, since the target concrete member of the embodiment is generated by the process of layer-by-layer real-time monitoring, predicting deformation trend and dynamically adjusting printing parameters, compared with the member printed by mainly relying on the fixed parameters such as pre-set printing path and extrusion speed in the prior art, the target concrete member of the embodiment has higher geometric precision, smoother surface and better stability, which can effectively improve the safety performance of the overall structure.

[0070] In the embodiment of the application, the printing parameters of the to-be-printed concrete layer are adjusted based on the deformation trend set to generate the target concrete member.

[0071] Embodiment two: Please refer to Figure 2 The application provides another concrete 3D printing self-adaptive deviation correction method, which comprises the following steps: Step 201, real-time collection of three-dimensional point cloud data, two-dimensional texture image and printing parameters of the printed concrete layer.

[0072] The specific implementation process of step 201 is similar to that of step 101, and will not be described here.

[0073] In the embodiment of the application, the three-dimensional point cloud data, two-dimensional texture image and printing parameters of the printed concrete layer are collected in real time.

[0074] Step 202, registration of the three-dimensional point cloud data and the three-dimensional surface of the theoretical digital model by using the iterative closest point algorithm to obtain a spatial transformation matrix.

[0075] The iterative closest point algorithm is a registration algorithm for realizing the optimal spatial transformation between two point sets through iterative calculation, and is used for accurately aligning the measured point cloud data and the theoretical model.

[0076] The three-dimensional surface refers to the continuous geometric expression of the outer surface of the concrete member defined in the theoretical digital model, which is usually composed of triangular meshes or NURBS surfaces, and completely characterizes the ideal geometric shape and spatial profile of the design target. The surface serves as a theoretical reference for registration and deviation calculation, and provides a spatial reference system for the point cloud data.

[0077] The spatial transformation matrix is ​​the optimal transformation parameter matrix calculated by the iterative nearest point algorithm. It includes the rotation matrix R and the translation vector T, and is used to transform the measured point cloud data into the coordinate system of the theoretical digital model.

[0078] Understandably, for each point in the measured point cloud, the nearest neighbor point is found on the 3D surface of the theoretical digital model. The rotation matrix R and translation vector T are calculated by SVD decomposition to minimize the distance between the two point sets. The measured point cloud is then aligned with the theoretical model using the current transformation matrix. This process is repeated iteratively until the preset convergence condition is met, and the final spatial transformation matrix is ​​output to complete the registration.

[0079] The preset convergence condition can be either the distance error being less than a threshold or reaching the maximum number of iterations, which will not be elaborated here.

[0080] In this embodiment of the invention, the three-dimensional point cloud data and the three-dimensional surface of the theoretical digital model are registered by the iterative nearest point algorithm to obtain the spatial transformation matrix, thereby achieving high-precision spatial alignment between the measured data and the theoretical model.

[0081] Step 203: Perform coordinate transformation on the 3D point cloud data based on the spatial transformation matrix to obtain the target point cloud dataset.

[0082] The target point cloud dataset refers to the set of resampled points after the original 3D point cloud data has been transformed into the coordinate system of the theoretical digital model through a spatial transformation matrix.

[0083] It is worth mentioning that for any point P in the point cloud original Given coordinates [x, y, z], and matrix operations P... transformed =R×P original By adding T, its new coordinates in the theoretical digital model coordinate system can be calculated, thus obtaining the transformed point P. transformed By iterating through all points in the point cloud and performing this transformation operation, a target point cloud dataset that is perfectly aligned with the theoretical model space is finally generated. Obtaining this point cloud dataset ensures the comparability of the measured data and the theoretical benchmark in the same coordinate space.

[0084] It is worth mentioning that R is a rotation matrix calculated by the iterative nearest point algorithm, used to characterize the rotation transformation from the measured coordinate system to the theoretical model coordinate system; T is a translation vector, used to characterize the translation transformation between the two coordinate systems.

[0085] In this embodiment of the invention, the calculated spatial transformation matrix is ​​applied to the original three-dimensional point cloud data, and each point cloud data point is transformed into the coordinate system of the theoretical digital model through matrix operations to obtain a target point cloud dataset aligned with the theoretical model space.

[0086] Step 204, calculate the deviation of each data point in the target point cloud data set to the nearest point of the three-dimensional surface and decompose it to obtain the Z-axis deviation value, the X-axis deviation value and the Y-axis deviation value.

[0087] The deviation decomposition refers to the process of decomposing the Euclidean distance deviation of each data point to the nearest point of the theoretical model surface into a normal deviation and a tangent deviation along the local coordinate system of the theoretical model surface.

[0088] It should be noted that the normal deviation is the Z-axis deviation, and the tangent deviation is the X and Y axis deviation.

[0089] The Z-axis deviation value refers to the deviation component along the normal direction of the nearest point on the theoretical model surface, and is used to represent the bulging or collapse defects of the printing layer in the vertical direction.

[0090] The X-axis deviation value and the Y-axis deviation value respectively refer to the deviation components in two mutually orthogonal directions in the tangent plane at the nearest point on the theoretical model surface, and are used to represent the shrinkage, expansion or misplacement deformation of the printing layer in the horizontal plane.

[0091] It is worth mentioning that the KD-tree acceleration algorithm is used to find the nearest point of the theoretical model surface for each target point cloud data point, the Euclidean distance between the data point and the nearest point is calculated as the total deviation, and then the total deviation vector is decomposed according to the local coordinate system at the nearest point to obtain the Z-axis deviation value, the X-axis deviation value and the Y-axis deviation value, so as to accurately quantify the deviation of the printing structure in each direction in the three-dimensional space.

[0092] In the embodiment of the present application, the deviation of each data point in the target point cloud data set to the nearest point of the three-dimensional surface is calculated and decomposed to obtain the Z-axis deviation value, the X-axis deviation value and the Y-axis deviation value.

[0093] Step 205, determine the deviation field matrix based on the Z-axis deviation value, the X-axis deviation value and the Y-axis deviation value.

[0094] The specific implementation of generating the deviation field matrix is as follows: First, the printing layer surface is uniformly meshed in the XY plane according to a preset resolution, which can be 1mm*1mm, to form a regular two-dimensional grid array, each grid cell represents a 1mm 2 region, and the coordinates of the center point of the cell are taken as the identifier of the cell.

[0095] Second, for each grid cell, all target point cloud data points are traversed, and all data points falling within the current grid cell range are selected by spatial position judgment, and if no data point falls within a certain grid cell, adjacent cell interpolation or empty value marking is used for processing.

[0096] Thirdly, for each grid cell containing data points, the X-axis deviation value, Y-axis deviation value and Z-axis deviation value corresponding to all data points falling into the cell are extracted, and the weighted average values of the three groups of deviation values can be calculated by using a bilinear interpolation algorithm: taking the reciprocal of the distance between each data point and the center of the grid as the weight to calculate the weighted average of the deviation values, so as to accurately reflect the overall deviation in the grid cell.

[0097] Fourthly, the calculated X-axis deviation average value, Y-axis deviation average value and Z-axis deviation average value are combined into a three-dimensional deviation vector [X avg ,Y avg ,Z avg ], which is stored as the deviation data of the current grid cell in the corresponding position of the matrix, and after traversing all grid cells, a structured deviation field matrix completely characterizing the three-dimensional deviation distribution of the surface of the printing layer is generated.

[0098] It is worth mentioning that the deviation field matrix is stored in the form of a two-dimensional array, the row and column indexes of the array correspond to the XY coordinates of the grid, and each element stores a three-dimensional deviation vector.

[0099] In the embodiment of the application, the surface of the printing layer is divided into grids according to a preset resolution, for each grid cell, the average values of the X-axis, Y-axis and Z-axis deviation values of all point cloud data points falling into the cell are calculated by using a bilinear interpolation algorithm, and the three average values are stored as the deviation vector of the grid cell, and finally a deviation field matrix completely characterizing the three-dimensional deviation distribution of the surface of the printing layer is generated.

[0100] Step 206: Extracting surface defect features of the two-dimensional texture image based on the convolutional neural network model to determine a defect feature data set.

[0101] Preferably, step 206 includes the following sub-steps: Step 11: Preprocessing the two-dimensional texture image to obtain a target texture image.

[0102] The target texture image refers to a standardized image suitable for input of the convolutional neural network model after preprocessing, which has a unified size, enhanced contrast and clear texture features after noise reduction.

[0103] Preprocessing refers to a series of image processing operations performed on the original two-dimensional texture image, aiming to improve image quality and unify data format, including but not limited to size normalization, contrast enhancement and noise filtering.

[0104] In the embodiment of the application, the two-dimensional texture image is preprocessed to obtain a target texture image.

[0105] Step 12, identifying and segmenting the target texture image based on the convolutional neural network model to obtain a defect segmentation mask.

[0106] The defect segmentation mask refers to a binary image generated after processing the target texture image by the convolutional neural network model, wherein the region with a pixel value of 1 is identified as a surface defect, and the region with a pixel value of 0 is identified as a normal concrete surface.

[0107] It can be understood that the preprocessed target texture image is input into the pre-trained U-Net convolutional neural network model. The model first extracts multi-scale features of the image step by step through the encoder path, then restores the spatial resolution step by step through the decoder path, and performs jump connection with the feature map of the corresponding stage of the encoder to fuse the detail information, and finally outputs a defect segmentation mask with the same size as the input image through a 1x1 convolutional layer and a Sigmoid activation function.

[0108] In the embodiment of the present application, the target texture image is identified and segmented based on the convolutional neural network model to obtain a defect segmentation mask.

[0109] Step 13, calculating the geometric morphological parameter set of the surface defect feature based on the defect segmentation mask.

[0110] The geometric morphological parameter set refers to a set of feature values used to quantitatively describe the geometric characteristics of the surface defect, including but not limited to defect area, perimeter, major axis length, minor axis length, aspect ratio, circularity, and contour moment, which are used to accurately represent the size, shape, and distribution characteristics of the defect.

[0111] The specific process of calculating the geometric morphological parameter set of the surface defect feature is as follows: First, perform connected component analysis on the defect segmentation mask, identify and label each independent defect region using the 8-neighborhood connection criterion, assign a unique label to each connected component, and form a labeled image. Through this step, the pixel-level defect identification in the binary mask is converted into an independent defect object with clear boundaries.

[0112] Second, calculate the pixel area of each labeled defect region. Convert the pixel area to the actual physical area by using the pre-calibrated pixel-physical size conversion coefficient. This conversion coefficient can be obtained by shooting a calibration board with known size and calculating the ratio of pixel size to physical size.

[0113] Third, the precise contour of each defect area is extracted, and based on the contour, the following features are calculated: perimeter: the actual physical perimeter (unit: mm) is obtained by accumulating the Euclidean distance between adjacent pixels on the contour; minimum circumscribed rectangle: the minimum area rectangle that can completely enclose the defect contour is calculated to obtain its long axis length and short axis length (unit: mm), and the rectangle direction angle; aspect ratio: the ratio of the long axis length to the short axis length of the minimum circumscribed rectangle is calculated to describe the length of the defect.

[0114] Fourth, the area label, center point coordinates, physical area, physical perimeter, long axis length, short axis length and aspect ratio of each defect are combined into a feature vector, and the feature vectors of all defects are summarized to generate a structured defect geometric parameter data set.

[0115] In the embodiment of the present application, based on the defect segmentation mask, the geometric parameter set of the surface defect feature is calculated.

[0116] Step 14, based on the geometric parameter set, a defect feature data set is generated.

[0117] The defect feature data set refers to a structured data set constructed based on the geometric parameter set, which is used to comprehensively represent the type, size, shape and distribution characteristics of the concrete surface defects.

[0118] It is worth mentioning that the parameters of the geometric parameter set are first standardized, and then the principal component analysis method is used to reduce the dimension of the features of the standardized parameters, extract the principal components that best represent the defect features from multiple geometric parameters, reduce data redundancy while retaining main shape information, then generate a feature vector for each defect, the feature vector contains basic size features, shape features, defect type identification, spatial distribution features, and finally, the feature vectors of all defects are indexed according to the time stamp and spatial position to construct a structured defect feature data set.

[0119] It is worth mentioning that the defect feature data set not only contains detailed feature descriptions of each defect, but also records the distribution of the defects in the printed layer, enriches the diversity of the data, and effectively improves the model prediction accuracy.

[0120] In the embodiment of the present application, based on the geometric parameter set, a defect feature data set is generated.

[0121] Step 207, input the deviation field matrix, the defect feature data set and the printing parameters into the deformation trend prediction model to determine the deformation trend set of the to-be-printed concrete layer.

[0122] Preferably, the specific training process of the deformation trend prediction model includes the following sub-steps: Step 21, obtaining the printing parameters, deviation field parameters, defect feature parameters and thixotropy parameters from the historical database for preprocessing to construct a training sample set, wherein the training sample set includes a training set and a test set.

[0123] It should be noted that the training sample set is divided into the training set and the test set by a preset ratio, which can be 8:2 or 7:3.

[0124] Preferably, step 21 includes the following sub-steps: Step 21a, data cleaning is performed on the printing parameters, deviation field parameters, defect feature parameters and thixotropy parameters to obtain an initial sample set.

[0125] The printing parameters refer to a set of adjustable variables recorded in the historical printing process to control the running state of the printing equipment and the output characteristics of the material.

[0126] The deviation field parameters refer to a set of structured deviation data calculated based on the historical point cloud data and the theoretical model.

[0127] The defect feature parameters refer to quantitative characterization data of surface defects extracted from historical two-dimensional texture images.

[0128] The thixotropy parameters refer to parameters representing the change of the rheological properties of the concrete material over time, including apparent viscosity, yield stress, thixotropic recovery time and shear thinning index.

[0129] It is worth mentioning that the thixotropy parameters can be obtained by, but not limited to, real-time measurement by an online rheometer installed on an extruder, offline measurement by a laboratory rotary rheometer on a concrete sample mixed at the same time, and inputting the measurement results into the system after associating with the time stamp and estimating based on material proportioning and environmental temperature and humidity data through experience or mechanism model.

[0130] It is worth mentioning that the range of the printing parameters is checked to eliminate data obviously beyond the physically feasible range, the deviation field parameters are subjected to outlier detection, and the three-sigma rule is used to identify and eliminate abnormal data points significantly deviating from the mean, the missing values in the defect feature parameters are filled according to the defect type: for crack defects, the mean value of the same type of defects is used for filling, for hole defects, the median value is used for filling, for continuous parameter missing, linear interpolation method is used to complete, and noise filtering is performed on the thixotropy parameters, and a sliding average window is used to smooth the data fluctuations to eliminate random noise interference in the sensor acquisition process, to generate the initial sample set.

[0131] In the embodiment of the present application, data cleaning is performed on the printing parameters, deviation field parameters, defect feature parameters and thixotropy parameters to obtain an initial sample set.

[0132] Step 21b, standardizing the parameters in the initial sample set to obtain an intermediate sample set.

[0133] Standardization refers to numerical transformation of parameters in the initial sample set with different dimensions and orders of magnitude to convert them into standard normal distribution with a mean of 0 and a standard deviation of 1, so as to eliminate the influence of dimensional differences between features on model training.

[0134] The intermediate sample set refers to a data set with uniform dimensions obtained after standardization.

[0135] It is worth mentioning that Z-score standardization can be used for printing parameters and deviation field parameters, and the calculation formula is: In the formula, x is the original value, μ is the mean of the parameter, and σ is the standard deviation.

[0136] For category type features such as defect type in defect feature parameters, one-hot encoding can be used to convert them into binary vector representation to eliminate the numerical partial order relationship between category labels; numerical type features such as area and perimeter can be standardized by Z-score.

[0137] For thixotropy parameters, since they have time sequence correlation, sliding window standardization can be used to perform Z-score transformation in each window to retain their dynamic change characteristics.

[0138] In the embodiment of the present application, the parameters in the initial sample set are standardized to obtain an intermediate sample set.

[0139] Step 21c, time sequence alignment and feature reorganization of the parameters in the intermediate sample set to obtain a training sample set.

[0140] Time sequence alignment refers to integrating parameters from different data sources with different timestamps onto a unified time axis through time synchronization technology to ensure that data at the same time point corresponds to the same printing state.

[0141] Feature reorganization refers to reorganizing multi-source parameters after time sequence alignment according to printing layer sequence and feature dimension to form a structured sample suitable for model input.

[0142] In the embodiment of the present application, the parameters in the intermediate sample set are aligned according to the timestamp, and records with time sequence interruption or asynchronization are removed. Data at each reference time point constitutes a sample to obtain a training sample set.

[0143] Step 22, inputting the training set into an initial prediction model for training to obtain a deformation trend prediction model.

[0144] The initial prediction model refers to an untrained multi-layer long short-term memory network model, the input layer dimension of which matches the feature dimension, and the output layer dimension of which is consistent with the adjustment parameter dimension of the deformation trend set.

[0145] In the embodiment of the present application, the training set is input into the initial prediction model for training to obtain a deformation trend prediction model.

[0146] Step 23, input the test set into the deformation trend prediction model to obtain a deformation trend set.

[0147] The test set refers to a data sample set reserved from the historical database and not participating in model training, which is used to objectively evaluate the generalization performance and prediction accuracy of the deformation trend prediction model.

[0148] The test set data after standardization, i.e., the printing parameters, the bias field parameters, the defect feature parameters and the thixotropy parameters, are input into the trained deformation trend prediction model, and the model predicts the corresponding deformation trend set sample by sample according to the input multi-source features.

[0149] In the embodiment of the present application, the test set is input into the deformation trend prediction model to obtain a deformation trend set.

[0150] Step 24, judge whether the deformation trend set meets the preset condition, if yes, the current prediction model is taken as the deformation trend prediction model.

[0151] The preset condition refers to the quantitative index threshold set for evaluating the performance of the deformation trend prediction model, including the prediction accuracy, error range and stability requirement, etc.

[0152] The specific implementation mode of judging whether the deformation trend set meets the preset condition is as follows: The performance indicators that can be included between the test set prediction result and the true value are as follows: First, the root mean square error indicator: measures the deviation degree of the predicted value and the true value. The root mean square error of the path offset and the layer height increment can be less than 1.0 mm, and the root mean square error of the extrusion speed increment can be less than 0.5 mm / s; Second, the average absolute percentage error indicator: evaluates the relative error of the prediction. The average absolute percentage error can be less than 10%.

[0153] The calculation mode of the root mean square error indicator is as follows: In the formula, RMSE ΔX is the root mean square error of the X-direction path offset prediction value, ΔX is the X-direction path offset, n is the total number of samples, ΔX pred,i is the X-direction path offset predicted by the i-th sample, ΔX true,iis the true X-direction path offset for the i-th sample.

[0154] The average absolute percentage error indicator is calculated as follows: where RAPE ΔX is the average absolute percentage error of the X-direction path offset.

[0155] It should be noted that if ΔX true,i = 0, the sample is ignored.

[0156] It is worth mentioning that the above is an example of the X-direction path offset, the Y-direction path offset, the layer height increment and the extrusion speed increment are the same, and the true data is the known and correct offset value obtained from the historical database.

[0157] If all the above indicators reach the preset threshold and the stability test passes, it is determined that the deformation trend set meets the preset condition, and the current trained model is used as the final deformation trend prediction model for practical application.

[0158] In the embodiment of the present application, it is determined whether the deformation trend set meets the preset condition, and if so, the current prediction model is used as the deformation trend prediction model.

[0159] Optionally, determining whether the deformation trend set meets the preset condition further includes the following sub-steps: Step 24a, if the deformation trend set does not meet the preset condition, the hyperparameters of the initial prediction model are obtained and adjusted by a preset step size.

[0160] Hyperparameters refer to parameters set before the start of model training, which cannot be learned through the training process itself. They control the structure, complexity and training process of the model.

[0161] The preset step size refers to the fixed change amount set for hyperparameter adjustment, which is used to systematically explore the hyperparameter space to optimize the model performance.

[0162] When the deformation trend set does not meet the preset condition, the learning rate, the number of hidden layer units, the dropout rate, the batch size L2 regularization coefficient and the current values of these numbers can be obtained from the initial prediction model.

[0163] All of the above data can be adjusted by the preset step size, or only one can be adjusted. Through these processes, the prediction accuracy of the model is improved.

[0164] In the embodiment of the present application, if the deformation trend set does not meet the preset condition, the hyperparameters of the initial prediction model are obtained and adjusted by a preset step size.

[0165] Step 24b: Use the adjusted hyperparameters as the new hyperparameters, and jump to the step of inputting the training set into the initial prediction model for training to obtain the deformation trend prediction model, until the deformation trend set meets the preset conditions.

[0166] The new hyperparameters refer to the optimized single hyperparameters or combinations of hyperparameters obtained after adjustment in step 24a.

[0167] Understandably, the adjusted hyperparameters in step 24a are used as new configuration parameters to update the network structure and training settings of the initial prediction model. Then, the program execution jumps to step 22 to retrain the model using the updated hyperparameters and the original training set. This iterative training process uses the same loss function and optimizer, but uses the model parameters obtained from the previous training as the initialization starting point to accelerate convergence. Finally, steps 23 and 24 are executed on the retrained model to check whether the deformation trend set meets the preset conditions. In this way, it is ensured that the final deformation trend prediction model meets the accuracy requirements of the actual printing scenario.

[0168] It is worth mentioning that, to ensure efficiency, a maximum number of iterations can be set. If the iterations are exceeded and convergence is still not achieved, an optimization failure message will be displayed, requiring manual intervention.

[0169] In this embodiment of the invention, the adjusted hyperparameters are used as new hyperparameters, and the process jumps to the step of inputting the training set into the initial prediction model for training to obtain the deformation trend prediction model, until the deformation trend set meets the preset conditions.

[0170] Step 208: Adjust the printing parameters of the concrete layer to be printed based on the deformation trend set to generate the target concrete component.

[0171] Preferably, step 208 includes the following sub-steps: Step 31: Generate the target printing path by compensating the printing path parameters based on the path offset.

[0172] The target printing path refers to the optimized path coordinate sequence used for actual printing, obtained by compensating and adjusting the original preset printing path according to the path offset.

[0173] Understandably, this involves obtaining the coordinate sequence of the original preset print path, which includes a series of ordered (X... original ,Y original Z original For each horizontal plane coordinate point in the original path, vector compensation is performed using path offsets ΔX and ΔY: Calculate the new X coordinate after compensation: X target =X original +ΔX, where, X target Let's define the new X coordinate; Calculate the new Y coordinate after compensation: Y target = Y original + ΔY, wherein Y target is the new Y coordinate; Thus, all the compensated coordinate points can be recombined in the original order to generate the target printing path. The target printing path, by compensating for the predicted deformation deviation in advance, ensures that the movement trajectory of the printing head in the X-Y plane can offset the expected deformation error, thereby ensuring the consistency of the printing contour with the design model.

[0174] In the embodiments of the present application, the printing path parameters are compensated based on the path offset to generate the target printing path.

[0175] Step 32, adjust the extrusion speed parameter based on the extrusion speed increment to generate the target extrusion speed.

[0176] The extrusion speed increment refers to the change amount of the material extrusion speed output by the deformation trend prediction model for adjusting the next layer printing process.

[0177] The extrusion speed increment is added to the original extrusion speed parameter to obtain the target extrusion speed.

[0178] It is worth mentioning that the target extrusion speed calculated can also be subjected to rationality verification, such as checking whether it exceeds the feasible speed range of the extruder servo motor, checking the matching with the printing head movement speed, ensuring the coordination between the material output and the movement trajectory, etc. If the verification passes, the target extrusion speed is sent to the extruder controller as the final parameter, and if the verification fails, the device limit value is truncated and a warning is triggered for manual verification.

[0179] In the embodiments of the present application, the extrusion speed parameter is adjusted based on the extrusion speed increment to generate the target extrusion speed, which can effectively prevent the problems of material over-extrusion or shortage caused by changes in layer height, path offset or material rheological property, and ensure the stability of the printing process and the forming quality.

[0180] Step 33, adjust the layer height parameter based on the layer height increment to generate the target layer height.

[0181] The target layer height refers to the optimized layer height for actual printing obtained by compensating and adjusting the original layer height parameter based on the layer height increment.

[0182] The original layer height parameter is added to the layer height increment to obtain the target layer height.

[0183] It is worth mentioning that the calculated target layer height can also be reasonably checked, such as checking whether it exceeds the Z-axis stroke range and mechanical limit of the printing device and whether it meets the minimum layer height requirement. If the check passes, the target layer height is sent to the extruder controller as the final parameter, and if the check fails, the device limit value is truncated, and a warning is triggered for manual verification.

[0184] It is worth mentioning that by dynamically correcting the accumulated error of the layer height caused by the unevenness of the current layer surface or the overall deformation, the dimensional accuracy of the printed component in the vertical direction and the interlayer bonding quality can be ensured, effectively avoiding the accumulation of height direction deviation in the printing process, thereby improving the stability and integrity of the target construction.

[0185] In the embodiment of the present application, the layer height parameter is adjusted based on the layer height increment to generate a target layer height.

[0186] Step 34, using the target printing path, the target extrusion speed and the target layer height to print the next concrete layer to generate a target concrete component.

[0187] It can be understood that the printing control system receives the target printing path, the target extrusion speed and the target layer height, and then drives the three-axis motion platform to move the print head according to the target printing path through the motion control module, while synchronously controlling the extruder to output concrete material according to the target extrusion speed, and positioning the nozzle to the Z-axis height corresponding to the target layer height at the beginning of each layer printing. Finally, by repeatedly executing the closed-loop control process of steps 201 to 208 layer by layer, the printing parameters of each layer are continuously adjusted adaptively, and finally a target concrete structure with high geometric accuracy, good interlayer bonding and strong structural integrity is generated, effectively solving the structural instability problem caused by error accumulation in traditional open-loop printing.

[0188] In the embodiment of the present application, the target printing path, the target extrusion speed and the target layer height are used to print the next concrete layer to generate a target concrete component.

[0189] Embodiment three: Please refer to Figure 3 The present application provides a kind of concrete 3D printing adaptive correction system, comprising: Data acquisition module 101, for real-time acquisition of three-dimensional point cloud data of printed concrete layer, two-dimensional texture image and printing parameter.

[0190] Calculation module 102, for determining deviation field matrix based on three-dimensional point cloud data and theoretical digital model.

[0191] Data extraction module 103, for extracting surface defect features of two-dimensional texture image based on convolutional neural network model, and determining defect feature data set.

[0192] The data prediction module 104 is configured to input the bias field matrix, the defect feature dataset and the printing parameter into a deformation trend prediction model to determine a deformation trend set of the to-be-printed concrete layer.

[0193] The data adjustment module 105 is configured to adjust the printing parameter of the to-be-printed concrete layer based on the deformation trend set to generate a target concrete component.

[0194] Since the above is a system corresponding to the concrete 3D printing adaptive correction method, the implementation principle is the same as the concrete 3D printing adaptive correction method, which will not be described here.

[0195] Embodiment four: The electronic device of the embodiment of the present application comprises a memory and a processor, and the memory stores a computer program; when the computer program is executed by the processor, the processor executes the concrete 3D printing adaptive correction method of any one of the above embodiments.

[0196] The memory can be an electronic memory such as a flash memory, an EEPROM (Electrically Erasable Programmable Read-Only Memory), an EPROM, a hard disk or a ROM. The memory has a storage space for program codes for executing any of the method steps described above. For example, the storage space for program codes can include individual program codes for implementing various steps in the above method, respectively. These program codes can be read from or written into one or more computer program products. These computer program products include program code carriers such as hard disks, compact discs (CDs), memory cards or floppy disks. The program codes can be compressed in an appropriate form, for example. These codes, when executed by a computing processing device, cause the computing processing device to perform various steps in the methods described above.

[0197] Embodiment five: The embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed to implement the concrete 3D printing adaptive correction method of any one of the above embodiments.

[0198] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0199] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the described device embodiments are merely schematic. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0200] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0201] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can be physically present separately, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0202] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application, essentially or the part that contributes to the prior art, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0203] The above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions described in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A concrete 3D printing adaptive deviation correction method, characterized in that, The method comprises the following steps: real-time acquisition of three-dimensional point cloud data, two-dimensional texture images and printing parameters of the printed concrete layer; determination of a deviation field matrix based on the three-dimensional point cloud data and a theoretical digital model; extraction of surface defect features of the two-dimensional texture images based on a convolutional neural network model to determine a defect feature dataset; input of the deviation field matrix, the defect feature dataset and the printing parameters into a deformation trend prediction model to determine a deformation trend set of the to-be-printed concrete layer; adjustment of the printing parameters of the to-be-printed concrete layer based on the deformation trend set to generate a target concrete component.

2. The concrete 3D printing adaptive deviation correction method according to claim 1, characterized in that, The method for determining the deviation field matrix based on the three-dimensional point cloud data and the theoretical digital model comprises the following steps: registration of three-dimensional surfaces of the three-dimensional point cloud data and the theoretical digital model by an iterative closest point algorithm to obtain a spatial transformation matrix; coordinate transformation of the three-dimensional point cloud data based on the spatial transformation matrix to obtain a target point cloud dataset; calculation of deviations of each data point in the target point cloud dataset to the closest point of the three-dimensional surface and decomposition to obtain Z-axis deviation values, X-axis deviation values and Y-axis deviation values; determination of the deviation field matrix based on the Z-axis deviation values, the X-axis deviation values and the Y-axis deviation values.

3. The concrete 3D printing adaptive deviation correction method according to claim 1, characterized in that, The method for extracting surface defect features of the two-dimensional texture images based on the convolutional neural network model to determine the defect feature dataset comprises the following steps: preprocessing of the two-dimensional texture images to obtain target texture images; recognition and segmentation of the target texture images based on the convolutional neural network model to obtain a defect segmentation mask; calculation of a geometric morphological parameter set of the surface defect features based on the defect segmentation mask; generation of the defect feature dataset based on the geometric morphological parameter set.

4. The concrete 3D printing adaptive deviation correction method according to claim 1, wherein, The specific training process of the deformation trend prediction model comprises the following steps: acquisition of printing parameters, deviation field parameters, defect feature parameters and thixotropy parameters from a historical database for preprocessing to construct a training sample set, wherein the training sample set comprises a training set and a test set; input of the training set into an initial prediction model for training to obtain the deformation trend prediction model; input of the test set into the deformation trend prediction model to obtain a deformation trend set; determination of whether the deformation trend set meets a preset condition, and if so, the current prediction model is taken as the deformation trend prediction model.

5. The concrete 3D printing adaptive deviation correction method according to claim 4, characterized in that, The method for preprocessing the printing parameters, the deviation field parameters, the defect feature parameters and the thixotropy parameters from the historical database to construct the training sample set comprises the following steps: data cleaning of the printing parameters, the deviation field parameters, the defect feature parameters and the thixotropy parameters to obtain an initial sample set; standardization processing of the parameters in the initial sample set to obtain an intermediate sample set; time sequence alignment and feature recombination of the parameters in the intermediate sample set to obtain the training sample set.

6. The concrete 3D printing adaptive deviation correction method according to claim 4, characterized in that, The method for determining whether the deformation trend set meets the preset condition further comprises the following steps: if the deformation trend set does not meet the preset condition, the hyperparameters of the initial prediction model are obtained and adjusted by a preset step size; the adjusted hyperparameters are taken as new hyperparameters, and the step of inputting the training set into the initial prediction model for training to obtain the deformation trend prediction model is repeated until the deformation trend set meets the preset condition.

7. The concrete 3D printing adaptive deviation correction method according to claim 1, wherein the deformation trend set comprises a path offset, an extrusion speed increment and a layer height increment. ​ The printing parameters include an extrusion speed parameter, a printing path parameter, and a layer height parameter; adjust the printing parameters of the to-be-printed concrete layer based on the deformation trend set to generate the target concrete component, including: compensate for the printing path parameter based on the path offset to generate a target printing path; adjust the extrusion speed parameter based on the extrusion speed increment to generate a target extrusion speed; adjust the layer height parameter based on the layer height increment to generate a target layer height; print the next concrete layer using the target printing path, the target extrusion speed, and the target layer height to generate the target concrete component.

8. A concrete 3D printing adaptive deviation correction system, characterized in that, comprising: a data acquisition module configured to acquire three-dimensional point cloud data, two-dimensional texture images, and printing parameters of the printed concrete layer in real time; a calculation module configured to determine a deviation field matrix based on the three-dimensional point cloud data and a theoretical digital model; a data extraction module configured to extract surface defect features of the two-dimensional texture images based on a convolutional neural network model to determine a defect feature dataset; a data prediction module configured to input the deviation field matrix, the defect feature dataset, and the printing parameters into a deformation trend prediction model to determine a deformation trend set of the to-be-printed concrete layer; a data adjustment module configured to adjust the printing parameters of the to-be-printed concrete layer based on the deformation trend set to generate the target concrete component.

9. An electronic device, comprising: comprising a memory and a processor, the memory storing a computer program capable of being loaded and executed by the processor to perform the adaptive deviation correction method for concrete 3D printing according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, a computer program capable of being loaded and executed by the processor to perform the adaptive deviation correction method for concrete 3D printing according to any one of claims 1 to 7.

Images