Patent Application
20230286075
The present invention relates to a building plan assistance method and a building plan assistance device when a built object is manufactured by weld beads.
In recent years, there is a growing need for manufacturing a component by additive manufacturing using a 3D printer. Researches and developments have been made toward practical applications of building using a metal material. A 3D printer for additive manufacturing of a metal material produces a built object having a desired shape by melting and solidifying a metal powder or a metal wire by use of a heat source such as a laser or an arc, and depositing the weld metal (weld beads).
However, in the additive manufacturing using the metal material, material properties such as metal structure and hardness tend to change according to manufacturing conditions. The properties of the metal material forming the built object may significantly vary from expected properties. Therefore, in an existing welding technique, the manufacturing conditions are adjusted based on empirical knowledge, trial and error, etc. such that the desired shape and properties can be obtained by predicting properties of the built object when the built object is manufactured under specified manufacturing conditions.
Further, in order to embody utilization of information from the above-mentioned experience and trial and error on a computer, for example, Patent Literature 1 discloses a case in which machine learning is utilized in a process of preparing a test cross-sectional image of a weldment and a test weldment, and determining suitability of weldment specifications such as a strength, a ductility, a hardness, a toughness, and a grain structure based on the test cross-sectional image of the weldment and the test weldment.
Patent Literature 1: JP2019-5809A
However, it is considered difficult to predict, based on material viewpoints, the properties of the built object manufactured by additive manufacturing because an additive manufacturing process is more complicated than a simple welding process. In addition, in a manufacturing method based on additive manufacturing, a degree of freedom in manufacturing conditions is extremely high, and there are various combinations of properties of a built object. Property prediction requires an enormous amount of arithmetic process.
Accordingly, an object of the present invention is to provide a building plan assistance method and a building plan assistance device capable of efficiently predicting properties of a built object with little effort and assisting creation of a more appropriate building plan for the built object.
The present invention includes the following configurations.
According to the present invention, it is possible to efficiently predict properties of a built object with little effort and assist creation of a more appropriate building plan for the built object.
(A) of
Embodiments of the present invention will be described in detail below by referring to the drawings.
Here, although a case in which weld beads each formed by melting and solidifying a filler metal fed from a welding head are additively manufactured into a desired shape by a building device is described as an example, configurations of a building method and a building device are not limited thereto. For example, other building methods such as a powder sintering and additive manufacturing method may be used.
<Configuration of Building System>
The building device 11 includes a welding robot 17 provided with a welding head having a welding torch 15 on a tip shaft, a robot control device 21 that drives the welding robot 17, a filler metal feeding unit 23 that feeds a filler metal (welding wire) M to the welding torch 15, and a welding power source 25 that supplies a welding current.
(Building Device)
The welding robot 17 is a multi-joint robot, and a continuously fed filler metal M is supported at a tip of the welding torch 15 attached to a tip shaft of a robot arm. A position and a posture of the welding torch 15 can be set three-dimensionally desirably within a range of the degree of freedom of the robot arm according to a command from the robot control device 21.
A shape sensor 32 and a temperature sensor 30 that move integrally with the welding torch 15 are provided on the tip shaft of the welding robot 17.
The shape sensor 32 is a non-contact-type sensor that measures a shape of a weld bead 28 to be formed and, if necessary, a shape around a bead forming position. Measurement by the shape sensor 32 may be performed at the same time when a weld bead is formed, or may be performed at different timings before and after the bead is formed. As the shape sensor 32, a laser sensor that detects a three-dimensional shape based on a position of a reflected light of an irradiated laser light or a time from an irradiation timing to a time at which the reflected light is detected can be used. A detection method of the shape sensor 32 is not limited to laser, and the shape sensor 32 may be a sensor using another detection method.
The temperature sensor 30 is a contact-type sensor such as a radiation thermometer or thermography, and detects a temperature (temperature distribution) at any position of a built object.
The welding torch 15 is a gas metal arc welding torch that has a shield nozzle (not shown) and is supplied with a shield gas from the shield nozzle. An arc welding method may be either a consumable electrode type such as shielded metal arc welding or carbon dioxide gas arc welding, or a non-consumable electrode type such as TIG welding or plasma arc welding, and is appropriately selected depending on an additively-manufactured object to be produced.
For example, in the case of the consumable electrode type, a contact tip is disposed inside the shield nozzle, and the filler metal M to which a melting current is to be supplied is held on the contact tip. The welding torch 15 generates an arc from a tip of the filler metal M in a shield gas atmosphere while holding the filler metal M.
The filler metal feeding unit 23 includes a reel 29 around which the filler metal M is wound, and a wire feed sensor 31 that measures a feed amount of the filler metal M fed from the reel 29 to a delivery mechanism and the welding torch 15. The filler metal M is fed from the filler metal feeding unit 23 to a delivery mechanism (not shown) attached to the robot arm or the like, and fed to the welding torch 15 while being fed forward and backward by the delivery mechanism as necessary.
Any commercially available welding wire can be used as the filler metal M. For example, welding wires provided as MAG welding and MIG welding solid wires (JIS Z 3312) for mild steel, high tensile steel and cryogenic steel, and arc welding flux-cored wires (JIS Z 3313) for mild steel, high tensile steel and cryogenic steel can be used. In addition, filler metals M such as aluminum, aluminum alloys, nickel, nickel-based alloys, etc. can be used depending on desired properties.
Then, when the continuously fed filler metal M is melted and solidified by an arc as described above, the weld bead 28 which is a melt-solidified body of the filler metal M is formed on a base plate 27. The base plate 27 is a metal plate such as a steel plate, but is not limited to such a plate-shaped object, and may be in other shapes such as a block shape, a rod shape, or a columnar shape.
(Robot Control Device)
The robot control device 21 drives the welding robot 17 to move the welding torch 15 and melt the continuously fed filler metal M by a welding current and a welding voltage from the welding power source 25.
The robot control device 21 is a computer device including an input and output interface 33, a storage unit 35 and an operation panel 37.
The input and output interface 33 is connected to the welding robot 17, the welding power source 25 and the building control device 13. The storage unit 35 stores various types of information including a drive program, which will be described later. The storage unit 35 includes a storage exemplified by a memory such as a ROM and a RAM, a drive device such as a hard disk and a solid state drive (SSD), a storage medium such as a CD, a DVD, and various memory cards, and various information can be input and output. The operation panel 37 may be an information input unit such as an input operation panel, or may be an input terminal for teaching the welding robot 17 (a teaching pendant).
A building program corresponding to a built object to be produced is transmitted from the building control device 13 to the robot control device 21. The building program includes a large number of instruction codes, and is created based on an appropriate algorithm according to various conditions such as shape data (CAD data, etc.), a material, and a heat input of the built object.
The robot control device 21 executes the building program stored in the storage unit 35 to drive the welding robot 17, the filler metal feeding unit 23, the welding power source 25, etc., and forms the weld bead 28 according to the building program. That is, the robot control device 21 drives the welding robot 19 to move the welding torch 15 along a track (welding track) of the welding torch 15 set in the building program, and drives the filler metal feeding unit 23 and the welding power source 25 according to a set welding condition to melt and solidify the filler metal M at the tip of the welding torch 15 by arc. Accordingly, the weld bead 28 is formed on the base plate 27. The weld beads 28 are adjacent to each other to form a weld bead layer, and a next weld bead layer is deposited on this weld bead layer, which is repeated to form a built object having a desired three-dimensional shape.
It should be noted that the building control device 13 may be disposed apart from the building device 11 and connected to the building device 11 from a remote location via a network, a communication unit, a storage medium, etc. The building program may be created by another device other than the building control device 13 and may be transmitted by communication.
(Generation of Building Program)
Next, a configuration of the building control device 13 and a specific procedure until the building control device 13 generates the building program will be described.
The building control device 13 is a computer device similar to the robot control device 21 and includes a CPU 41, a storage unit 43, an input and output interface 45, an input unit 47, and an output unit 49.
The storage unit 43 includes a ROM, which is a nonvolatile storage area, and a RAM, which is a volatile storage area. The input and output interface 45 is connected to the shape sensor 32, the temperature sensor 30, the filler metal feeding unit 23 including the wire feed sensor 31, the welding power source 25, the robot control device 21, the input unit 47, and the output unit 49, which are described above.
The input unit 47 is an input device such as a keyboard and mouse, and the output unit 49 includes a display device such as a monitor or an output terminal to which an output signal is transmitted.
In addition, the building control device 13 further includes a basic information table 51, a mathematical model generation unit 53, a database creation unit 55, a building plan unit 57, and a search unit 59, each of which will be described in detail below. Each of the components described above is operated according to a command from the CPU 41, and exhibits a function thereof.
First, an operator inputs, by the input unit 47 of the building control device 13 shown in
Next, property values of the built object (metal structure, average grain size, Vickers hardness, tensile strength, toughness, etc.) when the built object is manufactured according to the created building plan are predicted with reference to a database 61 prepared in advance, which indicates correspondences between various manufacturing conditions and properties of the built object manufactured under the conditions. It is preferable to use the parameters described above as the property values. Accordingly, since each property value can be easily measured by using a general-purpose measurement device such as a metallurgical microscope, an electron microscope (for example, SEM), and a Vickers tester, collection of data is facilitated.
When the predicted properties of the built object do not satisfy the desired properties, the building plan is created again by adjusting the various manufacturing conditions described above. Then, when the properties of the built object according to the created building plan satisfy the desired properties, a building program is created according to the building plan. The building program thus created is sent to the robot control device 21 shown in
In a building plan assistance method and device according to the present invention, the database 61 used for predicting and judging whether the built object can obtain the desired properties by the created building program is efficiently constructed with little effort. Accordingly, accurate and quick determination of the building plan can be performed, and thus assistance can be provided for smoothly creating a more appropriate building plan.
<First Database Configuration Example>
Next, a method for constructing the database 61 described above will be described.
Specifically, the building control device 13 creates a building plan according to the input data such as the material, the shape, and the welding condition of the built object. Property values of the built object when the built object is produced according to this building plan are obtained by the following first procedure and second procedure.
In the first procedure, the building control device 13 predicts, with reference to an initial database 63 in which relations between the building plan and the property values are registered in advance, properties of the built object to be additively-manufactured according to the created building plan.
In the second procedure, the building control device 13 drives the robot control device 21 in accordance with the created building plan, and causes the building device 11 to additively manufacture the built object. A test sample is cut out from the additively-manufactured built object, and a mechanical strength, a metal structure, etc. are actually measured by testing (observation).
A mathematical model 62 is generated by comparing a prediction result and a test result of the properties of the built object according to the same building plan obtained in this way such that a difference between the two results is reduced, and the database 61 is created using this mathematical model 62. The initial database 63 and the database 61 are created by the database creation unit 55 shown in
Here, a flow of a series of processes including generating the mathematical model 62 by machine-learning the prediction result of the properties according to the initial database 63 in the first procedure and the test result in the second procedure, and creating the database 61 using this mathematical model 62 will be described.
First, a built object to be manufactured is determined and shape data (shape data by 3D-CAD) is created (S11). A building plan is created based on the shape data of this built object (S12). The building plan includes a plurality of slice data obtained by dividing a model of the built object into layers by defining a predetermined depositing direction axis, a shape of a weld bead in each slice data, a welding condition for forming the weld bead, and the like.
Next, properties of the built object are predicted according to the first procedure based on the created building plan (S13). The properties of the built object are predicted by using the initial database 63. The initial database 63 is created based on the basic information table 51 (
First, parameter information (for example, a pass forming the weld bead, the number of passes, an order of forming the weld bead (welding track), and a cross-sectional shape of the weld bead) to be used in the database is extracted from the basic information table 51 prepared in advance, and is prepared as learning data (S21).
Next, the prepared learning data and property values of the built object corresponding to the learning data are related by an initial mathematical model (S22). That is, by repeatedly performing machine learning on a plurality of pieces of learning data and property values of the built object corresponding the learning data, an initial mathematical model expressing relations between the learning data and the property values of the built object are generated. The “mathematical model” as used herein means a model capable of formulating a quantitative behavior of properties of a built object and simulating nature of the properties of the built object by calculation. That is, the mathematical model is a calculation model created based on a group of experimental data collected in experiments and related by a predetermined algorithm, and this calculation model may be optimized to match well with the experimental data by assuming a predetermined function, or may be created by providing input information and output information by machine learning. Examples of a specific algorithm include a support vector machine, a neural network, and a random forest.
Then, the property values of the built object corresponding to the plurality of pieces of learning data are predicted by using the generated initial mathematical model, and these predicted values are made to correspond to the learning data and are registered as table components of the initial database 63 (S23). In this way, the initial database 63 is created.
Meanwhile, in the second procedure (S14 to S16 in
Then, the prediction result of the properties of the built object obtained in the first procedure is compared with the test result obtained in the second procedure (S17). When the difference between the prediction result and the test result is large, the mathematical model 62 shown in
Then, by using the mathematical model 62 obtained by causing the initial mathematical model to further machine-learn, property values (output information) of the built object corresponding to a plurality of any conditions (input information) are predicted, and the set conditions and the predicted property values are associated with each other to form table components of the database 61. In this way, the initial database 63 is corrected by using the mathematical model 62 to construct the database 61 in which a prediction result and a test result for a specific condition accurately match (S19).
Thus, a part where a test result does not exist can be complemented by predicting output information from a plurality of input information by using the mathematical model 62, thereby easily increasing an amount of information in the database 61 and improving accuracy of prediction.
Next, a specific method of constructing the database 61 by using the mathematical model 62 will be described in more detail.
(A) of
Here, a filler metal, which is a material of the built object, will be described as an example of the input information. As shown in (A) of
In that case, each type of filler metal is related to a property value by using a separate mathematical model such as the property value A of the built object for the filler metal A using a mathematical model A, the property value B of the built object for the filler metal B using a mathematical model B, and the property value C of the built object for the filler metal C using a mathematical model C.
Therefore, as shown in (B) of
The type of filler metal may be specified by a trade name such as MG-51T and MG-S63B (solid wire manufactured by Kobe Steel, Ltd.), or may be distinguished by a component composition (for example, carbon content) of the filler metal.
In the above-mentioned example, each property value is related to each type of filler metal, but actual input information includes more various kinds of items.
The input information at least includes a material of a built object, a welding condition, and a partial welding track. In addition to the filler metal described above, examples of a material of a weldment include members such as the base plate 27 (
Examples of the welding condition include at least one of a welding current, a welding voltage, a travel speed, a width of a pitch between welding tracks, an interpass time, a target position of the welding head, a welding position of the welding head, and a speed of feeding the filler metal when the weld bead is formed, or a combination thereof. Here, the target position of the welding head is a torch tip position for arranging a torch tip at a welding location, and the welding position of the welding head is an inclination angle between a vertical axis and a torch axis and a circumferential angle in a torch inclination direction around the vertical axis. In addition, the width of a pitch between welding tracks is a distance between adjacent welding tracks, and the interpass time represents a time moving from a welding pass of one welding track to a welding pass of a next welding track in a plurality of welding tracks.
The above-mentioned interpass time affects a metal structure of the weld bead to be formed.
During formation of the weld bead, when a filler metal made of a molten mild steel is quenched, the filler metal becomes a mixed structure mainly containing bainite. In addition, when the filler metal made of the molten mild steel solidifies naturally, the filler metal becomes a structure containing coarse ferrite, pearlite, and bainite. In a case of depositing weld beads, the structure becomes a structure in which when the weld beads are heated above a transformation point of ferrite by depositing weld beads of layers subsequent to the next layer, pearlite and bainite transform into ferrite, and coarse ferrite is refined.
In a case of adjusting the interpass time, for example, depositing the weld beads of the next layer while controlling an interlayer time and a heat input, and similarly depositing weld beads of the layers subsequent to the next layer, so that an interpass temperature falls within a range of 200° C. to 550° C., the weld beads are heated above the transformation point of ferrite. In that case, a homogenized structure made of a fine ferrite phase with an average grain size of 10 μm or less is obtained. Such a weld bead has a high hardness (for example, about 130 to 180 Hv in Vickers hardness), a good mechanical strength, and a substantially uniform hardness with little variation.
Meanwhile, when the interpass temperature is less than 200° C. in a case of depositing the weld beads of the next layer, even when the weld beads are heated by depositing the weld beads of the layers subsequent to the next layer, the transformation point of ferrite is not exceeded, and a homogenized structure made of a fine ferrite phase cannot be obtained. For example, at an initial stage of building, the interpass temperature in the case of depositing the weld beads of the next layer is less than 200° C. due to heat removal by the base plate 27. In that case, the weld beads at the initial stage of building become a mixed structure mainly containing bainite. In addition, when the interpass temperature exceeds 550° C., the weld beads are heated by depositing the weld beads of the next layer, and the weld beads are flattened and drip, making it impossible to deposit the weld beads in a predetermined shape. Further, since weld beads at a later stage of building (the uppermost layer of the built object) are not deposited with weld beads of a next layer and are not heated again, the molten filler metal remains in a naturally solidified state, that is, a structure containing coarse ferrite, pearlite and bainite.
Thus, the metal structure of the weld bead to be formed during the interpass time changes, and accordingly properties of the built object also change. The above is about the effect of the interpass time on the properties of the built object, but it has been found that other parameters similarly affect the properties of the built object.
The partial welding track is a welding track for an element shape obtained by cutting out a part of a shape of the built object, and means a welding track for, when a complex shape is decomposed into simple shapes (element shapes), building the simple shapes. Information regarding each welding track includes information regarding a pass forming the weld bead, the number of passes, an order of forming the weld bead, and a cross-sectional shape of the weld bead.
Here, a material of a building material, the welding condition, and the partial welding track described above are each referred to as an “item”, and the filler metals A, B, C, . . . , the welding current, the welding voltage, the travel speed, . . . , the element shape, the pass, the number of passes, . . . for items are each referred to as an “input subitem”.
By dividing each item of the input information into a plurality of input subitems, a range that can be input can be restricted. That is, by preventing a content other than the input subitems from being set as input data, for example, it is possible not to deviate from a recommended range of the welding robot 17 or the like of the building device 11, a recommended condition for using the filler metal, etc. Accordingly, it is possible to prevent a trouble due to a failure in a device and a material in advance, and to avoid presentation of an inappropriate condition.
As shown in
In addition, when contents of the items are represented by numerical values, regarding input data for each item, input subitems that divide a range of the input data into a plurality of sections may be defined, and a representative value corresponding to each input subitem may be defined as the input data. The representative value for each input subitem may be a value that represents the input subitem, such as a value of a median value, or an upper limit value, or a lower limit value within the input subitem.
In addition, a range of the input data does not need to be the same as the input information, which is performance data. The database creation unit 55 inputs the input data for each input subitem determined in this way to a mathematical model created by the mathematical model generation unit 53 to obtain output data for each input subitem.
Thus, the input subitems of each item are related to the property values of the built object by the mathematical model. Although it is possible to cause a plurality of mathematical models to learn in all combinations as described above, it is preferable to aggregate the plurality of mathematical models into approximately one mathematical model based on a specific welding condition, a welding track pattern, etc., and tune for each parameter based on the mathematical model. The “tune” as used here includes transfer learning and the like, in which one (learned model) learned in one area serves and is caused to efficiently learn in another area. Accordingly, it is possible to reduce an amount of calculation by reducing the learning data.
Next, an element shape in a case of determining a partial welding track and a welding track for each element shape will be described together with a specific example of a built object.
Here, a built object including a main body 65A, a first protrusion 65B connected to one surface of the main body 65A, and a second protrusion 65C connected to the other surface of the main body 65A is exemplified as the built object 65. When the built object 65 is divided into simple element shapes, the cylindrical first protrusion 65B, the cubic main body 65A, and the U-shaped second protrusion 65C are obtained. The division into the element shapes may be performed manually or by pattern matching with pre-registered simple shapes or the like.
For each of the divided element shapes, a welding track indicating an order of forming the weld bead is determined. That is, the welding track is determined for each divided element shape. The welding track for each element shape may be determined by designing each time the main body is divided into element shapes, but since the element shape is a simple shape, a plurality of types of welding tracks (reference welding tracks) each having a simple shape may be registered in advance in an element database, and a welding track having a shape corresponding to the element shape may be determined with reference to this element database.
For example, in a case of a cylindrical element shape, the cylindrical body is divided into a plurality of layers, and for each of the divided layers, a pass (torch track) for forming the weld bead become a determined reference welding track. By applying this reference welding track to the first protrusion 65B, a welding track B, which is a building procedure in a case of building the first protrusion 65B with the weld bead, can be easily determined.
For the main body 65A and the second protrusion 65C, similarly, reference welding tracks each having a similar shape can be determined by searching from the element database, and a welding track A of the main body 65A and a welding track C of the second protrusion 65C can be easily determined from the determined reference welding tracks. Thus, even for a built object having a complicated shape, by dividing the built object into element shapes, the built object can be regarded as an aggregate of simple shapes, and thus a building plan can be simplified.
When the shape data of the built object to be produced is input to the building plan unit 57 of the building control device 13 shown in
Each welding track is determined by applying the extracted reference welding track to the corresponding element shape (S34), and a building plan for the entire built object is created by combining the welding track and the welding condition (S35).
The created building plan is the building plan of S12 shown in
It should be noted that regarding the welding condition, information regarding the welding condition can be easily collected from drive signals and the like of the building device 11, the wire feed sensor 31, the shape sensor 32, and the welding robot 17. Those values can also be used to feedback control the shape of the model as necessary.
<Second Database Configuration Example>
Next, a case in which intermediate output information is provided in addition to the input information and the output information of the database 61 described above will be described.
Items of the input information including a material of a built object, a welding condition, and a partial welding track each include a plurality of input subitems. The intermediate output information is related to each combination of the input subitems by a separate first mathematical model. In addition, each input subitem of the intermediate output information is related to each input subitem of the output information by a second mathematical model. Here, information regarding a temperature history of the built object will be described as an example of the intermediate output information.
When the material of the built object such as a filler metal is heated according to the welding condition and melted and solidified along a predetermined welding track, the temperature history of the built object (weld bead) to be formed differs depending on the conditions in the items described above. Therefore, properties such as mechanical strength and metal structure of the built object to be formed also differ depending on the conditions.
In a case of estimating the property values of the built object, even when it is difficult to directly estimate the properties of the built object based on each item (each condition) of the input information, if the temperature history can be understood for each item, it may be easier to estimate the properties of the built object based on the temperature history. Therefore, in a case of relating the input information to the properties of the built object which are the output information, a two-step relation is performed including first relating each item of the input information to the temperature history, which is the intermediate output information, and then relating the temperature history to each property value of the built object, which is the output information. Accordingly, compared with the case in which the input information and the output information are directly related, it is possible to relate and estimate with higher precision.
Assuming that a melting point Tw of the weld bead is 1534° C., which is the melting point of iron (carbon steel), and a transformation point Tt of the weld bead (the A1 transformation point of carbon steel) is 723° C., a material of the weld bead after solidification is substantially determined by the temperature history in a range from the transformation point Tt to Tw equal to or lower than the melting point. That is, although heating and cooling are repeated in additive manufacturing, a factor that affects a structure of the built object is the temperature history in a range Aw described above. Therefore, by extracting a feature amount of the temperature history in the range (inspection temperature range) Aw from the transformation point Tt to Tw equal to or lower than the melting point, the properties of the built object can be predicted.
For example, among a plurality of peaks shown in
As shown in (A) of
Further, by predicting the structure by combining the temperature of the low-temperature-side local maximum point Pk2 and the temperature of the high-temperature-side local maximum point Pk1, prediction accuracy can be improved compared with a case in which prediction is made based on only one of the temperatures.
Thus, when the temperature history, which is a factor determining the material of the weld bead, can be specified based on the feature amount described above, the material of the weld bead formed in the temperature history can be predicted with relatively high accuracy. Therefore, items of intermediate processing information are set as determinants of the material of the building material, and the input information and the intermediate output information are related by the first mathematical model and the intermediate output information and the output information are related by the second mathematical model. Accordingly, it is possible to expect an effect that the input information and the output information can be related more accurately than when the input information and the output information are directly related.
Regarding this temperature history, temperature data at a predetermined position may be acquired by monitoring the temperature of the built object by the temperature sensor (
In addition, a temperature simulation calculation may be performed based on the type of filler metal or the welding condition.
An example of a basic equation used for temperature simulation will be shown below.
(Equation 1)
t+Δt
{H}=
1
{H}−Δt[C][K]
t
{T}−Δt[C]
t
{F}+Δt
t
{Q} (1)
The basic equation (1) is an equation for heat transfer analysis by a so-called explicit finite element method (FEM). Each parameter in the basic equation (1) is as follows.
Accordingly, a nonlinear phenomenon such as latent heat release can be calculated with high accuracy by using the enthalpy as an unknown quantity. It should be noted that a heat input during welding is input as a parameter for the volumetric heat generation or the heat flux.
In the above-mentioned basic equation (1), which is a three-dimensional heat conduction equation, the heat input during building (welding) may be applied to a welding region in accordance with the travel speed.
In addition, when the weld bead is short, heat input may be applied to the entire one bead.
<Other Database Configuration Examples>
In the first database configuration example described above, the input information and the output information are related by using a mathematical model I, and a database DB1 (database 61 described above) is constructed by the mathematical model I.
In addition, in the second database configuration example, the input information and the intermediate output information are related by using a mathematical model IIa and the intermediate output information and the output information are related by using a mathematical model IIb, and a database DB2 (database 61 described above) is constructed by the mathematical model IIa and the mathematical model IIb.
Then, the constructed databases DB1 and DB2 are compared, and a database whose output information is more accurate with respect to the input information is used as the database 61 shown in
Accordingly, by constructing a plurality of databases and selectively using a more accurate database, accuracy of predicting a property value of a built object is improved, and creation of a more appropriate building plan can be assisted.
Thus, the present invention is not limited to the embodiments described above, and the combination of configurations of the embodiments with each other and the modification or application by a person skilled in the art based on the statements in the description and common techniques are also expected in the present invention and are included in the claimed range.
As described above, the present description discloses the following items.
According to the building plan assistance method, even in additive manufacturing in which welding is complicated and the number of passes tends to be enormous, creation of an appropriate building plan can be assisted by preparing a database in advance.
According to the building plan assistance method, treating the temperature history, which is a representative process feature of the built object, as the intermediate output information makes it easier to correlate the input information with the properties of the built object such as a hardness. In addition, regarding the temperature history, it is possible not only to use a value measured by actually building, but also to calculate the value by temperature simulation. Therefore, data can be easily supplemented, and database construction is facilitated.
According to the building plan assistance method, since a viscosity of the weld bead during melting varies depending on the type of filler metal and the cross-sectional shape of the weld bead tends to vary accordingly, a welding condition and a track plan suitable for each filler metal can be set by creating a mathematical model for each type of filler metal.
According to the building plan assistance method, since all the information can be monitored during building, data can be easily collected.
According to the building plan assistance method, since all the information can be monitored during building, data can be easily collected.
According to the building plan assistance method, by cutting out the shape of the built object into several patterns of element shapes and planning a welding condition and a welding track for each element shape, a building plan can be easily created. By creating, in advance, a partial welding track corresponding to each element shape in various variations, even for a built object having a complicated shape, a building plan can be created without requiring complicated processing.
According to the building plan assistance method, the construction of a database is facilitated by using a state of a metal structure, a hardness (Vickers hardness, etc.), and a mechanical strength, which can be tested relatively easily and in a short period of time.
According to the building plan assistance method, by constructing a mathematical model by machine learning, a part without test data can be complemented, and the prediction accuracy is improved as the data is complemented. In addition, since data corresponding to input and output can be collected from basic built objects such as wall building or block building, machine learning data can be easily prepared.
According to the building plan assistance method, by setting a limit on an input range so as not to deviate from a recommended range for driving a building device, a recommended condition for using a filler metal, etc., it is possible to avoid inputting a condition that is likely to cause a trouble due to a failure in a device or a material.
According to the building plan assistance device, even in additive manufacturing in which welding is complicated and the number of passes tends to be enormous, creation of an appropriate building plan can be assisted by preparing a database in advance.
According to the building plan assistance device, treating the temperature history, which is a representative process feature of the built object, as the intermediate output information makes it easier to correlate the input information with the properties of the built object such as a hardness. In addition, regarding the temperature history, it is possible not only to use a value measured by actually building, but also to calculate the value by temperature simulation. Therefore, data can be easily supplemented, and database construction is facilitated.
It should be noted that the present application is based on a Japanese patent application (Japanese Patent Application No. 2020-123860) filed on Jul. 20, 2020, contents of which are incorporated by reference in the present application.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2020-123860 | Jul 2020 | JP | national |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/JP2021/025684 | 7/7/2021 | WO |
