CN113290302A - Quantitative prediction method for surplus height of electric arc welding additive manufacturing - Google Patents
Quantitative prediction method for surplus height of electric arc welding additive manufacturing Download PDFInfo
- Publication number
- CN113290302A CN113290302A CN202110277762.6A CN202110277762A CN113290302A CN 113290302 A CN113290302 A CN 113290302A CN 202110277762 A CN202110277762 A CN 202110277762A CN 113290302 A CN113290302 A CN 113290302A
- Authority
- CN
- China
- Prior art keywords
- molten pool
- height
- residual height
- layer
- welding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003466 welding Methods 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 46
- 239000000654 additive Substances 0.000 title claims abstract description 34
- 230000000996 additive effect Effects 0.000 title claims abstract description 34
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 22
- 238000010891 electric arc Methods 0.000 title claims abstract description 16
- 238000005253 cladding Methods 0.000 claims abstract description 69
- 230000008569 process Effects 0.000 claims abstract description 27
- 239000000463 material Substances 0.000 claims abstract description 16
- 230000000007 visual effect Effects 0.000 claims abstract description 15
- 238000012544 monitoring process Methods 0.000 claims abstract description 12
- 238000001816 cooling Methods 0.000 claims description 20
- 238000005070 sampling Methods 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 11
- 230000002787 reinforcement Effects 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 claims description 3
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 2
- 230000008034 disappearance Effects 0.000 claims description 2
- 230000008859 change Effects 0.000 abstract description 24
- 239000010410 layer Substances 0.000 description 130
- 239000011229 interlayer Substances 0.000 description 11
- 238000012360 testing method Methods 0.000 description 11
- 238000004458 analytical method Methods 0.000 description 9
- 238000009826 distribution Methods 0.000 description 7
- 230000017525 heat dissipation Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 230000035515 penetration Effects 0.000 description 4
- 239000002356 single layer Substances 0.000 description 4
- 238000009825 accumulation Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 229910001220 stainless steel Inorganic materials 0.000 description 3
- 239000010935 stainless steel Substances 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000005484 gravity Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 239000011324 bead Substances 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000002844 melting Methods 0.000 description 1
- 230000008018 melting Effects 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000007711 solidification Methods 0.000 description 1
- 230000008023 solidification Effects 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K9/00—Arc welding or cutting
- B23K9/32—Accessories
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B33—ADDITIVE MANUFACTURING TECHNOLOGY
- B33Y—ADDITIVE MANUFACTURING, i.e. MANUFACTURING OF THREE-DIMENSIONAL [3-D] OBJECTS BY ADDITIVE DEPOSITION, ADDITIVE AGGLOMERATION OR ADDITIVE LAYERING, e.g. BY 3-D PRINTING, STEREOLITHOGRAPHY OR SELECTIVE LASER SINTERING
- B33Y10/00—Processes of additive manufacturing
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B33—ADDITIVE MANUFACTURING TECHNOLOGY
- B33Y—ADDITIVE MANUFACTURING, i.e. MANUFACTURING OF THREE-DIMENSIONAL [3-D] OBJECTS BY ADDITIVE DEPOSITION, ADDITIVE AGGLOMERATION OR ADDITIVE LAYERING, e.g. BY 3-D PRINTING, STEREOLITHOGRAPHY OR SELECTIVE LASER SINTERING
- B33Y40/00—Auxiliary operations or equipment, e.g. for material handling
Landscapes
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Manufacturing & Machinery (AREA)
- Materials Engineering (AREA)
- Physics & Mathematics (AREA)
- Plasma & Fusion (AREA)
- Mechanical Engineering (AREA)
- Arc Welding In General (AREA)
Abstract
The invention relates to a quantitative prediction method of surplus height of electric arc welding additive manufacturing, which is characterized in that a CMT additive surplus height prediction system is constructed on the basis of a molten pool vision sensing system, a positioning system and a three-dimensional system, the molten pool vision sensing system acquires a molten pool image of the 1 st ms at the moment of CMT basic value, the positioning system is used for determining the position of the acquired molten pool image on a weld joint, and the current weld joint height is measured by the three-dimensional system due to the fact that the current weld joint layer has a remelting process on the next layer of the weld joint layer; and inputting the preprocessed molten pool image into a residual height prediction network to obtain predicted residual height information, and monitoring the residual height increment of each cladding layer in the electric arc additive manufacturing process by using the molten pool visual information. The method is based on the CMT single-channel multilayer molten pool form and temperature characteristic change rule, a CMT material increase residual height prediction system is constructed, and the change trend of the cladding layer residual height is predicted, so that the precision and the generalization capability of the prediction system are improved, and the method has high precision and high stability.
Description
Technical Field
The invention relates to a quantitative prediction method for the residual height of electric arc welding additive manufacturing, belonging to the technical field of online prediction of metallurgical quality.
Background
The additive manufacturing is a process of continuously melting metal wire materials to be stacked and formed under the action of a welding heat source. In the process, as the number of the cladding layers increases, the accumulated parts have the problems of serious heat accumulation, poor heat dissipation condition, difficult solidification of overheated molten pool, difficult control of the shape of the cladding layer and the like. These problems all affect the metallurgical strength, weld layer dimensional accuracy and surface accuracy of the part and therefore, monitoring of the stability of the build-up process and the quality of the formed part is important. However, under the influence of the cladding layer remelting area, the existing monitoring means mainly depend on the width of a molten pool or the overall height of accumulated parts, and the cladding layer residual height which is most concerned is difficult to measure.
Weld penetration can be used to reflect weld strength, but normally penetration is difficult to monitor during the additive process. The coupling relation exists between the weld penetration and the weld reinforcement, so that the change of the weld reinforcement reflects the change of the penetration to a certain extent during material increase, and the quality control of the material increase manufacturing process can be realized by monitoring the weld reinforcement. However, due to the influence of the cladding remelting region, the accurate value of the residual height of the cladding cannot be measured under normal conditions.
Disclosure of Invention
In order to solve the technical problems, the invention provides a quantitative prediction method of the residual height of the additive manufacturing of the electric arc welding, which analyzes the shape and the temperature field change rule of the molten pool of the same layer in different stages and the molten pool of different layers in the same stage, constructs a theoretical model of the quantitative prediction system of the residual height of the additive manufacturing of the electric arc welding based on the rule, and accurately predicts the residual height of the formed cladding layer, and the specific technical scheme is as follows:
a quantitative prediction method for the residual height of electric arc welding additive manufacturing comprises the following steps:
the method comprises the following steps: constructing a CMT additive residual height prediction system: constructing a CMT additive residual height prediction system based on a molten pool visual sensing system, a positioning system and a three-dimensional system, wherein the molten pool visual sensing system acquires a molten pool image of the CMT base value moment in 1ms, the positioning system is used for determining the position of the acquired molten pool image on a weld seam due to short sampling interval time and large sampling sample amount, the current weld seam has a remelting process on the next layer of weld seam due to the current weld seam, the three-dimensional system is used for measuring the current height of the weld seam, and the height difference between two adjacent layers on the same sampling point is the residual height;
step two: and inputting the preprocessed molten pool image into a residual height prediction network to obtain predicted residual height information, and monitoring the residual height increment of each cladding layer in the electric arc additive manufacturing process by using the molten pool visual information.
Further, the CMT material increase remaining height prediction system in the first step is also matched with a welding system and a three-dimensional scanner, wherein the welding system comprises a welding power supply, a wire feeder, a robot and a cooling system; the positioning system comprises a color CCD camera, a laser and a black and white CCD camera;
emitting laser with the central wavelength of 450nm on the upper edge part of the welding wire by a laser, wherein the laser is used for assisting positioning and corresponding the acquired molten pool image to the actual position of a welding seam; a black and white CCD camera to capture the laser spot.
Further, the specific process of the pre-processed molten pool image in the second step is as follows: in the data acquisition process, combining with the CMT welding current characteristic, an FPGA system sends two paths of synchronous signals to control a color CCD camera and a black-and-white CCD camera to acquire images in the 1 st ms at the CMT basic value moment, and the acquired molten pool image and the acquired laser point image under the low arc light interference are acquired; due to the influence of a remelting region, in order to accurately measure the residual height increment of each layer, after welding of each layer is finished, scanning the height of the current layer by using a three-dimensional scanner, fitting the variation relation between the position and the height of a welding seam by using a Matlab tool box, and unifying a molten pool image, the position of the welding seam and the corresponding residual height increment to be used as a data set of a residual height prediction network, wherein the height difference between two adjacent layers of the same welding seam position at the same sampling point is the residual height increment of the current welding seam position;
the input of the residual height prediction network is an additive molten pool image original graph and a last full connection layer.
Furthermore, the CMT material increase residual height prediction system gives an actual residual height value to the collected molten pool image for debugging the residual height prediction network, the FPGA system controls the color CCD camera to collect the molten pool image at the 1 st ms of the CMT basic value moment, then the image is input into the residual height prediction network, and the output value is the residual height of the cladding layer at the current moment, so that the residual height prediction during material increase welding is realized.
Further, the operation process of the residual height prediction network is as follows:
first, let the input of some residual block in ResNet network be XkThe output result is Xk+1At this time, it is possible to obtain:
Xk+1＝Xk+F(Xk，wk) (1)
F(Xk，wk)＝wkσ(wk-1Xk-1) (2)
wherein sigma is an activation function, the ResNet network adopts a nonlinear activation function ReLU, and the calculation formula is as follows:
and secondly, setting the loss function of the ResNet network as L, and then obtaining the loss function according to the chain derivative rule by combining the formula:
as can be seen from the equation (4),
x capable of updating weight of each network layer and transmitting gradient information back to any network layer in the backward propagation stagekIn the meantime due to
Greater than 0 is always true, so that the gradient disappearance phenomenon caused by too small backward propagation of the network or even 0 in the network training process is solved.
The invention has the beneficial effects that:
the method is based on the form and temperature characteristic change rule of the CMT single-channel multi-layer molten pool, a theoretical model of the CMT material increase surplus height prediction system is constructed, and the change trend of the surplus height of the cladding layer is predicted, so that the precision and the generalization capability of the prediction system are improved, and the method has high precision (the regression error is less than 0.05mm) and high stability.
Drawings
Fig. 1 is a schematic diagram of a CMT additive residual height prediction system of the present invention.
FIG. 2 is a flow chart of a CMT additive residual height prediction system of the present invention.
FIG. 3 is a process of feature map shrinkage due to convolution and average pooling.
Fig. 4 is a residual height prediction system applied to actual production.
Fig. 5 is a flow chart of a residual height prediction network.
Fig. 6 is an additive forming profile.
FIG. 7 is a diagram showing a multi-physical quantity characteristic of the second cladding layer,
wherein (a) the residual height varies; (b) molten pool shape in four states; (c) the average temperature change of the weld seam in the fixed region of the weld pool.
FIG. 8 is a diagram showing physical quantities of cladding layers,
wherein: (a) molten pool shape when each layer is inclined; (b) the height of each layer is changed; (c) the molten pool morphology of the different cladding layers at 50mm from the arcing point.
Figure 9 is a graph of typical layer regression results and error,
wherein: (a) a second layer; (b) a fifth layer; (c) an eighth layer; (d) the tenth layer.
FIG. 10 is a graph showing the regression results of the height of each layer at the time of interlayer cooling for 3 min.
FIG. 11 is a graph showing regression results of the residual height of the cladding layer at different interlayer cooling times,
wherein: (a) cooling time is 1 min; (b) cooling time is 2 min; (c)3 min; (d) cooling time is 4 min.
Detailed Description
The present invention will now be described in further detail with reference to the accompanying drawings.
The CMT additive surplus height prediction system consists of a molten pool visual sensing system, a positioning system and a three-dimensional system. As shown in figure 2, a molten pool vision sensing system acquires a molten pool image of the 1 st ms at the moment of CMT basic value, a positioning system is used for determining the position of the acquired molten pool image on a weld seam due to short sampling interval time and large sampling sample amount, a three-dimensional system is used for measuring the current height of the weld seam due to the remelting process of the current cladding layer on the next cladding layer, and the height difference between two adjacent layers on the same sampling point is the residual height. And then inputting the preprocessed molten pool image into a residual height prediction network designed in the chapter to obtain predicted residual height information, thereby realizing monitoring of residual height increment of each cladding layer in the electric arc additive manufacturing process by using molten pool visual information.
The device for monitoring the residual quantity of the cladding layer in the CMT arc additive manufacturing process comprises three parts: welding system, CMT increase material remaining height prediction system and computer. The welding system includes a welding power supply, a wire feeder, a robot, and a cooling system. The CMT additive residual height prediction system comprises a color CCD camera (Basler ace A640-750uc) fixed on a robot, an included angle between the color CCD camera and a welding gun is about 40 degrees, a laser, a positioning system consisting of a black and white CCD camera (Basler ace A1920-155um) fixed on a welding platform and a three-dimensional scanner, and is shown in figure 1. The computer is the core of the molten pool sensing and predicting the residual height.
In order to correspond the acquired weld pool image to the actual position of the weld seam, a laser is used for auxiliary positioning. A laser with a center wavelength of 450nm is used to irradiate the upper edge part of the welding wire, and a black and white CCD camera is arranged to capture a laser spot, as shown in figure 1. In the data acquisition process, the FPGA system sends two paths of synchronous signals by combining the CMT welding current characteristic, the color CCD camera and the black-and-white CCD camera are controlled to acquire images in the 1 st ms at the moment of CMT basic value, and the acquired molten pool image and the acquired laser point image under the low arc light interference are shown in figure 2. In addition, due to the influence of the remelting region, in order to accurately measure the extra height increment of each layer, after welding of each layer is finished, the height of the current layer is scanned by a Wiiboox three-dimensional scanner, a Matlab toolbox is used for fitting the change relation between the position and the height of the welding line, and the height difference of the same welding line positions of two adjacent layers of the same sampling point is the extra height increment of the current welding line position. Therefore, the weld pool image, the weld seam position and the corresponding residual height increment can be unified to be used as a data set of the residual height prediction network.
The input of the residual height prediction network structure is an additive molten pool image original graph and a full connection layer of the last layer, and only the value of the residual height is transmitted. Table 1 gives specific information of the network part. FIG. 3 shows the process of feature map shrinkage due to convolution and average pooling. The residual height prediction network structure uses a basic network framework which is Resnet-34, MSE is used as a loss function, and MAE is used as an evaluation criterion of a regression result.
TABLE 1 network architecture
The residual height prediction is a supervised regression problem, and the CMT material increase residual height prediction system is used for giving a residual height actual value to the collected molten pool image and debugging a residual height prediction network. In practical production application, the FPGA system controls the color CCD camera to collect a molten pool image in the 1 st ms of CMT basic value moment, then the image is input into the residual height prediction network, and the output value is the residual height of the cladding layer at the current moment, so that the residual height prediction during the additive welding is realized. A diagram of a system for predicting the residual height in actual production is shown in fig. 4.
Therefore, the remaining height prediction network flow chart is shown in fig. 5. The image collected by the visual system of the molten pool is sent into the neural network after being preprocessed. And predicting a residual height value according to the molten pool image by the regression network so as to achieve the purpose of residual height monitoring.
1 single pass multilayer CMT cladding layer weld pool quality analysis
Taking CMT-based stainless steel single-channel multi-layer additive manufacturing as an example, the change situation of the residual quantity of the cladding layer in the electric arc additive manufacturing process is monitored. The basic welding process parameters used are shown in table 4.2.
TABLE 4.2 CMT additive manufacturing Process parameters
1.1 Single-layer cladding layer weld pool quality analysis
The stacking was carried out in the same direction, and the height of each layer was as shown in FIG. 6(b), wherein the interlayer cooling time was 3 min. The height of the arcing end is higher and the height of the arcing end is lower than that of the middle section of the cladding layer. As shown in fig. 6(a), (c), the acquired weld pool images are also obviously different, but the problems can be solved by adjusting the welding parameters of the arc starting end and the arc extinguishing end, so that the weld pool image of the cladding layer 20-70mm away from the arc starting end is taken for analysis.
Because of the change of the heat dissipation condition around the molten pool and the influence of other factors, the molten pool shape is changed dynamically all the time, and the molten pool image at the position 20-70mm away from the arc end of a certain layer of cladding layer can be discussed in four stages according to the change of the molten pool length. As shown in fig. 7(a), the change in the molten pool state will be described by taking the molten pool of the second layer cladding layer as an example. The first stage is 20mm-40.5mm, the front part of the molten pool leaves the arcing region, but the rear part of the molten pool does not leave the arcing region, and the region can be observed to be in a red hot state due to more cladding amount and poor heat dissipation condition of the arcing region. The second stage is 40.5mm-47mm, the rear part of the molten pool leaves the arc striking area, the molten pool shape begins to shrink, and the length and the width of the molten pool are minimum at the stage. The third stage is 47mm to 66mm, the heat dissipation condition is deteriorated due to the influence of heat accumulation, the molten pool is in a high-temperature state for a long time, and the flowing of the molten pool is facilitated, so that the length of the molten pool is lengthened, and the width of the molten pool is widened. The fourth stage is 66mm to 70mm, and in the fourth stage, because the molten pool is in the stage of the cladding layer inclination, the molten pool flows downwards under the influence of gravity, the molten pool is longer from the position of a welding gun, the dry elongation is longer, and the length of the molten pool is changed.
Fig. 7(c) shows the average temperature of the weld bead in a region at a fixed distance from the welding gun (this region is the red box region in fig. 7 (b)) as a function of the welding time. It can be seen from fig. 7(c) that the change in average temperature of this region undergoes a plateau-fall-rise-fall law, corresponding to the four stages of the previous analysis. The change of the temperature field distribution helps the regression of the remaining height of the cladding layer.
1.2 quality analysis of different cladding layer weld pools
In fact, the difference between the molten pool morphologies of different cladding layers is greater. The first cladding layer is in direct contact with the base material, and the heat dissipation conditions are best, so that the difference between the length and area of the molten pool and the other layers is greatest. The subsequent molten pool quality analysis of the single-pass multilayer is not particularly described, and the molten pools of the cladding layers after the first layer are all used as research objects.
As shown in fig. 8(b), the second to tenth layers all start to be tilted after undergoing a plateau, and the position where the tilting occurs is closer to the starting arc end as the number of layers increases, and the red dotted line portion in fig. 8(b) is a boundary line of four stages of the molten pool of each cladding layer, where the last boundary line is the position on the cladding layer when each layer starts to be tilted, and fig. 8(a) is an image of the molten pool when each layer starts to be tilted. As the number of layers increases, the inclination angle of each layer increases, and the width, length, area and dry elongation of the welding wire at the beginning of the inclination are completely different along with the influence of heat dissipation conditions and other factors. FIG. 8(c) is a diagram showing the weld pool at 50mm from the arc end of each layer, and the weld pool patterns of the second to sixth layers are in the third stage, in which the weld layer is not yet inclined and the dry elongation of the wire is hardly changed. And the molten pool forms from the seventh layer to the tenth layer are in the fourth stage, at the moment, the molten pool flows downwards due to the gravity action caused by the inclined cladding layer, the dry elongation of the welding wire is lengthened, and the longer the dry elongation of the welding wire is along with the increase of the layer number.
In summary, the width, length, area and temperature field of the molten pool at different positions of the same layer of cladding layer are dynamically changed, the molten pool at the same position of different layers of cladding layer is also different in shape and dry elongation of the welding wire due to inclination and the like, and the visual shape of the molten pool is different due to the changes, so that the difference is favorable for predicting the regression of the network to the residual height of the cladding layer.
2 residual height prediction network experiment result and analysis
And repeating multiple groups of experiments under the condition that the interlayer cooling time is 3min, acquiring the molten pool images of the cladding layers by using the equipment, and obtaining the corresponding residual height values. Selecting molten pool images of the same layer of cladding layer to randomly form a training set and a test set according to a ratio approximate to 4: 1, then training a network on NVIDIARTX 2080TI GPU, performing residual height regression on the test set, and comparing the residual height regression with an actual value. This process was repeated 4 times, and the number of randomly selected layers and the number of training and testing sets are shown in Table 4.3. Furthermore, all the cladding layer molten pool images are combined into a training set and a testing set according to the proportion of 4: 1, and the training process and the testing process are the same. The number of training and test sets is shown in Table 4.3.
TABLE 4.3 residual height regression dataset size
2.1 Single-layer cladding layer prediction error analysis
As shown in fig. 9, the regression results of the residual heights of four randomly selected cladding layers are shown. The black background graph shows the height variation of one layer from the next along the direction of welding. And after each layer is cooled, extracting the overall height of the part in the current layer by using a three-dimensional scanner. And (3) drawing a curve of the height of the part along with the change of the position of the welding seam, and fitting the curve by using an MATLAB tool box to obtain a corresponding functional relation. And then substituting the position of the molten pool image on the weld joint calculated by the laser positioning system into the function to obtain the corresponding part height of the position at the upper layer and the lower layer. The height difference is the corresponding residual height of the weld pool image. The blue line is the true value of the residual height, the red line is the residual height predicted by the network, and the change of the absolute value of the error of the blue line and the absolute value of the error of the red line is shown as the purple line. The average error of the four-layer regression results is shown in table 4. The regression calculation time for a single weld puddle image was 61 ms.
TABLE 4 prediction error of different cladding layers
The molten pool in the middle section of the cladding layer changes due to the change of the surrounding heat dissipation conditions and other factors, and the shape of the molten pool and the distribution of the temperature field of the molten pool change accordingly, and the two factors are main factors directly influencing the forming characteristics. For high-temperature objects, the color distribution of the objects in the CCD is directly influenced by the temperature of the objects, so that the visual image gives consideration to the morphological characteristics and the temperature characteristics of the molten pool. The network can regress the height of the single-layer cladding layer according to the visual molten pool shape difference.
2.2 error analysis of prediction results for different cladding layers
FIG. 10 shows the regression results of the residual heights of the layers at a time of cooling between layers for 3min, and the average error of the regression results of the ten layers of residual heights is 0.0294 mm.
For molten pools of different cladding layers, the inclination angle of each layer is increased along with the increase of the number of layers, so that the positions of the cladding layers at which the inclination starts are different, and the visual morphology of the molten pools is obviously different. And for the same position of different cladding layers, the visual form of a molten pool and the dry elongation of the welding wire are changed due to factors such as inclination, and the difference enables the network to return to the residual heights of different cladding layers.
2.3 remaining height prediction network generalization ability to welding Process parameters
The single-pass multi-layer same-direction accumulation process is a multivariable strong coupling process, a plurality of factors (current, welding speed, interlayer cooling time and the like) are coupled together to jointly influence the residual height of the cladding layer, and the interlayer cooling time plays an important role in the residual height of the cladding layer. Therefore, in order to evaluate the generalization ability of the depth residual error network, the interlayer cooling time is changed to 1min, 2min, 3min and 4min under the condition of keeping other welding parameters unchanged, and multiple groups of experiments are repeated. Randomly forming a training set and a testing set by each layer of molten pool images acquired by a vision system according to a ratio of approximately 4: 1, wherein the number of the training set is 21000, the number of the testing set is 5290, and the training process and the testing process are the same as those in 1.1.
FIG. 11 shows regression results of the residual height of the cladding layer at different interlayer cooling times. The model has good performance, and the regression average error of the overall residual height is 0.0466 mm. As shown in fig. 11(b), the eighth, ninth, and tenth cladding layers with a cooling time of 2min had poor regression effect, and the average error was 0.2405 mm. The reasons for poor regression effect are that the undercut phenomenon occurs on the eighth cladding layer, the flow of the molten pool occurs on the left side of the ninth cladding layer and the tenth cladding layer, and the two phenomena change the molten pool form and the distribution of the temperature field of the molten pool. Therefore, the network regression capability is not influenced by the interlayer cooling time.
And (4) conclusion:
in the electric arc additive manufacturing process, the monitoring of the increment of the residual height of the cladding layer is indispensable, but the monitoring is often difficult due to the influence of the remelting zone of the cladding layer. Aiming at the problem that the monitoring of the surplus height of a cladding layer is difficult in the electric arc material increase process, a CMT electric arc material increase surplus height quantitative prediction system is constructed on the basis of the change rule of the shape and the temperature characteristics of a molten pool in CMT same-direction welding. In the early data acquisition stage, in order to unify a molten pool image, a welding seam position and a residual height increment, a molten pool vision system is adopted to acquire a molten pool image of a base value moment in 1ms, and meanwhile, a CMT cladding layer residual height trend change model is constructed based on three-dimensional point cloud distribution conditions of different layers, and the residual height increment is extracted.
The method is characterized in that CMT stainless steel single-pass multilayer welding is taken as a technological background, a equidirectional welding mode is adopted, data sets consisting of single-layer and different-layer molten pool images and corresponding residual heights under the same interlayer cooling time and different-layer cooling time are collected and trained, and the test results of the test sets prove that the network designed in the chapter has high precision (the regression error is less than 0.05mm) and high stability. And analyzing by combining the molten pool shape and temperature field distribution of the same layer in different stages, the molten pool shape and regression result error of different layers in the same stage, and obtaining the conclusion that the molten pool shape and the molten pool temperature field distribution are main factors influencing the regression result.
In the embodiment, a single-channel multilayer equidirectional welding of CMT stainless steel is taken as an example, a molten pool image-residual height data set under different interlayer cooling time is acquired on the basis of a molten pool vision system and a CMT cladding layer residual height trend change model, and an arc additive residual height quantitative prediction system based on molten pool morphology-temperature characteristics predicts the residual height change trend of the cladding layer, so that the precision and the generalization capability of the prediction system are improved.
In light of the foregoing description of the preferred embodiment of the present invention, many modifications and variations will be apparent to those skilled in the art without departing from the spirit and scope of the invention. The technical scope of the present invention is not limited to the content of the specification, and must be determined according to the scope of the claims.
Claims (5)
1. A quantitative prediction method for the residual height of electric arc welding additive manufacturing is characterized by comprising the following steps: the method comprises the following steps:
the method comprises the following steps: constructing a CMT additive residual height prediction system: constructing a CMT additive residual height prediction system based on a molten pool visual sensing system, a positioning system and a three-dimensional system, wherein the molten pool visual sensing system acquires a molten pool image of the CMT base value moment in 1ms, the positioning system is used for determining the position of the acquired molten pool image on a weld seam due to short sampling interval time and large sampling sample amount, the current weld seam has a remelting process on the next layer of weld seam due to the current weld seam, the three-dimensional system is used for measuring the current height of the weld seam, and the height difference between two adjacent layers on the same sampling point is the residual height;
step two: and inputting the preprocessed molten pool image into a residual height prediction network to obtain predicted residual height information, and monitoring the residual height increment of each cladding layer in the electric arc additive manufacturing process by using the molten pool visual information.
2. The quantitative prediction method of weld reinforcement for arc welding according to claim 1, wherein: the CMT material increase residual height prediction system in the first step is also matched with a welding system and a three-dimensional scanner, wherein the welding system comprises a welding power supply, a wire feeder, a robot and a cooling system; the positioning system comprises a color CCD camera, a laser and a black and white CCD camera;
emitting laser with the central wavelength of 450nm on the upper edge part of the welding wire by a laser, wherein the laser is used for assisting positioning and corresponding the acquired molten pool image to the actual position of a welding seam; a black and white CCD camera to capture the laser spot.
3. The quantitative prediction method of weld reinforcement for arc welding according to claim 1, wherein: the concrete process of the pretreated molten pool image in the second step is as follows: in the data acquisition process, combining with the CMT welding current characteristic, an FPGA system sends two paths of synchronous signals to control a color CCD camera and a black-and-white CCD camera to acquire images in the 1 st ms at the CMT basic value moment, and the acquired molten pool image and the acquired laser point image under the low arc light interference are acquired; due to the influence of a remelting region, in order to accurately measure the residual height increment of each layer, after welding of each layer is finished, scanning the height of the current layer by using a three-dimensional scanner, fitting the variation relation between the position and the height of a welding seam by using a Matlab tool box, and unifying a molten pool image, the position of the welding seam and the corresponding residual height increment to be used as a data set of a residual height prediction network, wherein the height difference between two adjacent layers of the same welding seam position at the same sampling point is the residual height increment of the current welding seam position;
the input of the residual height prediction network is an additive molten pool image original graph and a last full connection layer.
4. The quantitative prediction method of weld reinforcement for arc welding according to claim 3, wherein: the CMT material increase residual height prediction system gives a residual height actual value to the collected molten pool image to debug the residual height prediction network, the FPGA system controls the color CCD camera to collect the molten pool image in the 1 st ms at the moment of CMT basic value, then the image is input into the residual height prediction network, and the output value is the residual height of the cladding layer at the current moment, so that the residual height prediction during material increase welding is realized.
5. The quantitative prediction method of weld reinforcement for arc welding according to claim 4, wherein: the ResNet operation process of the basic network adopted by the residual height prediction network is as follows:
first, let the input of some residual block in ResNet network be XkThe output result is Xk+1At this time, it is possible to obtain:
Xk+1＝Xk+F(Xk,wk) (1)
F(Xk,wk)＝wkσ(wk-1Xk-1) (2)
wherein sigma is an activation function, the ResNet network adopts a nonlinear activation function ReLU, and the calculation formula is as follows:
and secondly, setting the loss function of the ResNet network as L, and then obtaining the loss function according to the chain derivative rule by combining the formula:
as can be seen from the equation (4),
x capable of updating weight of each network layer and transmitting gradient information back to any network layer in the backward propagation stagekIn the meantime due to
Greater than 0 is always true, so that the gradient disappearance phenomenon caused by too small backward propagation of the network or even 0 in the network training process is solved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110277762.6A CN113290302A (en) | 2021-03-15 | 2021-03-15 | Quantitative prediction method for surplus height of electric arc welding additive manufacturing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110277762.6A CN113290302A (en) | 2021-03-15 | 2021-03-15 | Quantitative prediction method for surplus height of electric arc welding additive manufacturing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113290302A true CN113290302A (en) | 2021-08-24 |
Family
ID=77319412
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110277762.6A Pending CN113290302A (en) | 2021-03-15 | 2021-03-15 | Quantitative prediction method for surplus height of electric arc welding additive manufacturing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113290302A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113441815A (en) * | 2021-08-31 | 2021-09-28 | 南京南暄励和信息技术研发有限公司 | Electric arc additive manufacturing layer width and residual height cooperative control method based on deep learning |
| CN114101855A (en) * | 2021-12-29 | 2022-03-01 | 福州大学 | Electric arc additive manufacturing and testing method for duplex stainless steel |
| CN115294105A (en) * | 2022-09-28 | 2022-11-04 | 南京理工大学 | Multilayer multi-pass welding remaining height prediction method |
| CN115609110A (en) * | 2022-11-22 | 2023-01-17 | 南京理工大学 | Electric arc composite additive melting depth prediction method based on multimode fusion |
| FR3135637A1 (en) * | 2022-05-17 | 2023-11-24 | Fse | Device and method for controlling a material supply in additive manufacturing |
| CN117340280A (en) * | 2023-12-05 | 2024-01-05 | 成都斐正能达科技有限责任公司 | LPBF additive manufacturing process monitoring method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111738369A (en) * | 2020-08-24 | 2020-10-02 | 南京知谱光电科技有限公司 | Weld penetration state and penetration depth real-time prediction method based on visual characteristics of molten pool |
| CN111739019A (en) * | 2020-07-29 | 2020-10-02 | 南京知谱光电科技有限公司 | Material increase residual height prediction method based on long-range prediction of molten pool image |
| CN111932539A (en) * | 2020-10-13 | 2020-11-13 | 南京知谱光电科技有限公司 | Molten pool image and depth residual error network-based height and penetration collaborative prediction method |
| CN112017186A (en) * | 2020-11-02 | 2020-12-01 | 南京知谱光电科技有限公司 | Material increase and residual height prediction method based on molten pool image and depth residual error network |
-
2021
- 2021-03-15 CN CN202110277762.6A patent/CN113290302A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111739019A (en) * | 2020-07-29 | 2020-10-02 | 南京知谱光电科技有限公司 | Material increase residual height prediction method based on long-range prediction of molten pool image |
| CN111738369A (en) * | 2020-08-24 | 2020-10-02 | 南京知谱光电科技有限公司 | Weld penetration state and penetration depth real-time prediction method based on visual characteristics of molten pool |
| CN111932539A (en) * | 2020-10-13 | 2020-11-13 | 南京知谱光电科技有限公司 | Molten pool image and depth residual error network-based height and penetration collaborative prediction method |
| CN112017186A (en) * | 2020-11-02 | 2020-12-01 | 南京知谱光电科技有限公司 | Material increase and residual height prediction method based on molten pool image and depth residual error network |
Non-Patent Citations (1)
| Title |
|---|
| 杨志超: "面向虚拟焊接质量评估的深度学习方法及应用研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅰ辑》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113441815A (en) * | 2021-08-31 | 2021-09-28 | 南京南暄励和信息技术研发有限公司 | Electric arc additive manufacturing layer width and residual height cooperative control method based on deep learning |
| CN113441815B (en) * | 2021-08-31 | 2021-11-16 | 南京南暄励和信息技术研发有限公司 | Electric arc additive manufacturing layer width and residual height cooperative control method based on deep learning |
| CN114101855A (en) * | 2021-12-29 | 2022-03-01 | 福州大学 | Electric arc additive manufacturing and testing method for duplex stainless steel |
| CN114101855B (en) * | 2021-12-29 | 2023-02-14 | 福州大学 | Electric arc additive manufacturing and testing method for duplex stainless steel |
| FR3135637A1 (en) * | 2022-05-17 | 2023-11-24 | Fse | Device and method for controlling a material supply in additive manufacturing |
| CN115294105A (en) * | 2022-09-28 | 2022-11-04 | 南京理工大学 | Multilayer multi-pass welding remaining height prediction method |
| CN115609110A (en) * | 2022-11-22 | 2023-01-17 | 南京理工大学 | Electric arc composite additive melting depth prediction method based on multimode fusion |
| CN115609110B (en) * | 2022-11-22 | 2023-12-15 | 南京理工大学 | Electric arc composite additive penetration prediction method based on multimode fusion |
| CN117340280A (en) * | 2023-12-05 | 2024-01-05 | 成都斐正能达科技有限责任公司 | LPBF additive manufacturing process monitoring method |
| CN117340280B (en) * | 2023-12-05 | 2024-02-13 | 成都斐正能达科技有限责任公司 | LPBF additive manufacturing process monitoring method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113290302A (en) | Quantitative prediction method for surplus height of electric arc welding additive manufacturing | |
| CN105855743B (en) | A kind of Weld pool dynamic process on-line monitoring system and method | |
| CN108290219B (en) | Additive manufacturing method and apparatus | |
| CN111932539B (en) | Molten pool image and depth residual error network-based height and penetration collaborative prediction method | |
| JP4953453B2 (en) | Method and apparatus for optimizing a forging process | |
| CN112017186A (en) | Material increase and residual height prediction method based on molten pool image and depth residual error network | |
| KR20230004570A (en) | On-site inspection method based on digital data model of welding | |
| CN114012210B (en) | Deposition quality judgment system and method in electric arc additive process | |
| CN111319260B (en) | Method for monitoring a molten bath | |
| JP2021516160A (en) | A method for automatically determining the optimum welding parameters for performing work piece welding | |
| CN111738369A (en) | Weld penetration state and penetration depth real-time prediction method based on visual characteristics of molten pool | |
| CN115697610A (en) | Method for analyzing a laser machining process, system for analyzing a laser machining process and laser machining system with such a system | |
| EP3666427A1 (en) | Method for melt pool monitoring using geometric length | |
| CN111319265B (en) | Molten pool monitoring method using algebraic connectivity | |
| CN115294105B (en) | Multilayer multi-pass welding remaining height prediction method | |
| JP5611177B2 (en) | Ablation abnormality detection device and anomaly detection method | |
| CN111318696B (en) | Molten pool monitoring method using Green's theorem | |
| CN114623935A (en) | Steel ladle quasi-throwing and quasi-stopping model application method based on infrared thermal imager imaging data | |
| JP2576298B2 (en) | Pipe length measuring method and apparatus | |
| CN111318697B (en) | Method for monitoring a weld pool using fractal dimension | |
| Dellarre et al. | Qualify a near-infrared camera to detect thermal deviation during aluminum alloy Wire Arc Additive Manufacturing | |
| JP7463257B2 (en) | Welding quality evaluation method and welding equipment | |
| Ma | Vision-sensing and bead width control of the first layer of multi-layered part in GTAW based Additive Manufacturing | |
| Wu et al. | Penetration prediction of narrow-gap laser welding based on coaxial high dynamic range observation and machine learning | |
| Curiel Braco et al. | Methodology for the path definition in multi-layer gas metal arc welding (GMAW) |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210824 |
|
| RJ01 | Rejection of invention patent application after publication |