JP6833150B1 - Machining program search device and machining program search method - Google Patents

Abstract

The machining program search device (100) has a first position information indicating whether each of the plurality of divided spaces in the three-dimensional space is located inside, outside, or at the boundary portion of the first three-dimensional model. And, in each of the second 3D models, a plurality of second positions indicating whether each of the plurality of divided spaces is located inside, outside, or at the boundary of the second 3D model. A model conversion unit (20) for acquiring information is provided. The machining program search device is a similarity determination unit that determines the degree of similarity in shape between the first three-dimensional model and the second three-dimensional model by comparing the first position information and the second position information. Based on (40) and the determination result in the similarity determination unit, a machining program for machining the second three-dimensional model having a relatively high degree of shape similarity among the plurality of second three-dimensional models is provided. It includes a registration program acquisition unit (50) that searches and acquires from a database.

Description

The present disclosure relates to a machining program search device and a machining program search method for searching a machining program having a shape similar to the shape to be machined.

A machine tool controlled by a control device such as a Numerical Control (NC) device machines a machining object into various finished shapes. In order to perform machining by a machine tool, the control device needs to control the machine tool according to a machining program. However, since it takes a lot of time and experience to create a machining program for controlling a machine tool used by a control device, the creation support function for supporting the creation of the machining program is being enhanced. For example, as a machining program creation support function, a function for creating a machining program by allowing the user to set the dimensions of the machining target on the NC screen while looking at the drawing, and a computer-aided design (CAD) data directly There is a function to read and create a machining program such as an NC program. These creation support functions are based on the premise that a machining program is created from scratch, and even when machining a machining finish shape that is similar to the machining finish shape for which a machining program has been created in the past, the machining program can be created from scratch. Since it was created, the efficiency of creating the machining program was poor.

When the user creates a machining program, the creation efficiency is improved by diverting the machining program corresponding to the above-mentioned machined finish shape of the similar shape, but in order to divert the machining program, the machining finish to be newly machined It is necessary to determine whether or not the shape is similar to the finished machined shape for which a machining program has been created in the past.

The search system described in Patent Document 1 uses a part or the whole of the 3D shape model as a search question for the 3D shape database, and selects a 3D shape similar to the shape of the part in the search question from the 3D database. Techniques related to 3D shape search are disclosed. Specifically, the search system converts a three-dimensional object into a point cloud as a preprocessing, extracts a partial shape by multi-view rendering, and performs posture normalization of the partial shape as a feature extraction, from a multi-viewpoint. The first feature vector is obtained using the normal histogram of. Then, the search system converts the three-dimensional object into a point cloud for the search question, extracts the partial shape by multi-view rendering, performs the posture normalization of the partial shape as the feature amount extraction, and performs the posture normalization of the partial shape from the multi-viewpoint. The second feature vector is obtained using the normal histogram of. When searching for a three-dimensional shape, the search system obtains the degree of difference between the first feature vector and the second feature vector, and uses the value of the difference to obtain a three-dimensional shape similar to the shape of the part in the search question. Select from a dimensional database.

JP-A-2018-136642

However, in the technique of Patent Document 1, a three-dimensional shape model representing a three-dimensional object is converted into a point cloud, converted into an image by multi-view rendering, and then feature quantities are extracted using a normal histogram. Therefore, in the technique of Patent Document 1, a three-dimensional shape including a space formed by a three-dimensional object shape inclusion portion and a hole formed in the three-dimensional object, which does not appear in the image converted by multi-view rendering. There was a problem that it could not be detected.

The present disclosure has been made in view of the above, and obtains a machining program search device that automatically and easily searches a database for a machining program that processes a shape similar to the finished shape of machining in a machine tool. The purpose.

In order to solve the above-mentioned problems and achieve the object, the machining program search device according to the present disclosure searches a database in which a plurality of different machining programs for machining a machining target in a machine tool are registered. .. The machining program search device includes a first three-dimensional model, which is a three-dimensional model indicating the finished shape of the machining target for which a new machining program is created, and a plurality of machining programs machined by a plurality of machining programs registered in the database. A plurality of second three-dimensional models, which are three-dimensional models indicating the processed finished shape of the processing target, are composed of three axes of X-axis, Y-axis, and Z-axis, and a plurality of divided spaces along each of the three axes. Arranged in the virtual three-dimensional space formed by, in the first three-dimensional model, each of the plurality of divided spaces in the three-dimensional space is either inside, outside, or at the boundary of the first three-dimensional model. In the first position information indicating whether or not the portion is located, and in each of the plurality of second 3D models, each of the plurality of divided spaces is either inside, outside, or the boundary portion of the second 3D model. It is provided with a model conversion unit that acquires a plurality of second position information indicating whether or not it is located in the portion of. The machining program search device is a similarity determination unit that determines the degree of similarity in shape between the first three-dimensional model and the second three-dimensional model by comparing the first position information and the second position information. And, based on the judgment result in the similarity determination unit, a machining program for machining the second 3D model having a relatively high shape similarity among the plurality of 2nd 3D models is searched from the database. It is provided with a registration program acquisition unit to be acquired.

The machining program search device according to the present disclosure has an effect that a machining program for machining a shape similar to the machined finished shape of machining in a machine tool can be automatically and easily searched from a database.

The figure which shows the structure of the machining program search apparatus which concerns on Embodiment 1. The figure which shows the structure of the model transformation part which includes the machining program search apparatus which concerns on Embodiment 1. The figure which shows the structure of the similarity determination part provided in the processing program search apparatus which concerns on Embodiment 1. A flowchart showing a processing procedure of the machining program search apparatus according to the first embodiment. A flowchart showing a procedure of processing of the grid data generation unit of the model transformation unit according to the first embodiment. The figure which shows an example in which the 3D model shown by the 3D model data acquired by the grid data generation part of the model conversion part which concerns on Embodiment 1 was drawn. The figure which shows an example which the 3D model shown by the 3D model data acquired by the lattice data generation part of the model conversion part which concerns on Embodiment 1 was arranged and drawn on the lattice space. A flowchart showing a procedure of processing of the data compression unit of the model transformation unit according to the first embodiment. A flowchart showing a procedure of processing of the similarity determination unit according to the first embodiment. The figure which shows the example of the novel 3D model in Embodiment 1. The figure which shows the example of the registration 3D model in Embodiment 1. The figure which shows the registered 3D model searched as the registered 3D model similar to the new 3D model shown in FIG. 10 from the 20 registered 3D models shown in FIG. The figure which shows the structure of the similarity determination part using machine learning in Embodiment 1. A flowchart showing a procedure of processing of the similarity determination unit according to the first embodiment. The figure which shows the 3D model classified into the 1st cluster in the example of clustering 20 3D models into 4 clusters by the k-means method in Embodiment 1. FIG. The figure which shows the 3D model classified into the 2nd cluster in the example of clustering 20 3D models into 4 clusters by the k-means method in Embodiment 1. FIG. The figure which shows the 3D model classified into the 3rd cluster in the example of clustering 20 3D models into 4 clusters by the k-means method in Embodiment 1. FIG. The figure which shows the 3D model classified into the 4th cluster in the example of clustering 20 3D models into 4 clusters by the k-means method in Embodiment 1. FIG. The figure which shows the structure of the machine learning part which concerns on Embodiment 1. The figure which shows the structure of the neural network used by the machine learning apparatus which concerns on Embodiment 1. The figure which shows the registered 3D model sorted in descending order of similarity with the new 3D model shown in FIG. The figure which shows the structure of the model transformation part which executes the data expansion in the machining program search apparatus shown in FIG. A flowchart showing data expansion processing executed by the model transformation unit according to the first embodiment. The figure which shows the structure of the machining program search apparatus which concerns on Embodiment 2. The figure which shows the structure of the difference recognition part provided in the processing program search apparatus which concerns on Embodiment 2. A flowchart showing a procedure of processing of the difference recognition unit according to the second embodiment.

The machining program search apparatus and the machining program search method according to the embodiment will be described in detail below with reference to the drawings. The present disclosure is not limited to this embodiment.

Embodiment 1.
FIG. 1 is a diagram showing a configuration of a machining program search device 100 according to the first embodiment. The machining program search device 100 is a computer that searches for machining programs having a similar shape after machining of the workpiece to be machined.

The machining program search device 100 is connected to the CAD system 110 and acquires three-dimensional model data from the CAD system 110. The three-dimensional model data is processed shape data showing a three-dimensional processed finished shape which is a shape of the work to be processed after processing. That is, the three-dimensional model can be rephrased as the processed finished shape to be processed. The three-dimensional model data includes information on the machined finished shape of the work and geometric information on the machined finished shape of the work. Geometric information is information on a three-dimensional shape in a coordinate system. The machining program search device 100 searches for a 3D model having a shape similar to the shape shown in the 3D model data based on the acquired 3D model data, and corresponds to the searched 3D model and this 3D model. Output the machining program to be performed. Hereinafter, the three-dimensional model may be simply referred to as a model.

The processing program search device 100 includes an input unit 11 for inputting information, a display unit 12 for displaying information, an output unit 13 for outputting information, a processor 14 for executing various processes, and a memory 15 for storing information. And have.

The CAD system 110 creates machined shape data indicating a three-dimensional machined finished shape of the work to be machined by three-dimensional (3D) CAD.

The input unit 11 receives the three-dimensional model data composed of the CAD data created by the CAD system 110 from the CAD system 110, which is an external device external to the machining program search device 100, and transmits the three-dimensional model data to the model conversion unit 20 described later. .. That is, the input unit 11 functions as a processing shape data input unit into which processing shape data is input.

However, the three-dimensional model data is not limited to CAD data. The three-dimensional model data may be any data that can be interpreted by the machining program search device 100. The display unit 12 displays information on the screen and presents it to the user. The output unit 13 outputs information to the outside of the machining program search device 100 and presents it to the user.

The processor 14 is a CPU (Central Processing Unit), a processing unit, an arithmetic unit, a microprocessor, a microcomputer, or a DSP (Digital Signal Processor). The memory 15 includes RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory) or EEPROM (registered trademark) (Electrically Erasable Programmable Read Only Memory), HDD (Hard Disk Drive). Or SSD (Solid State Drive) is included. The processing program of the processing program search device 100 is stored in the memory 15. The processor 14 executes a program stored in the memory 15.

The memory 15 stores a database 60, a processing program (not shown) of the processing program search device 100, and the like.

The database 60 stores three-dimensional model data. In the first embodiment, the case where the database 60 stores the model data DN (N is a natural number) from the model data D1 will be described. Each of the model data D1 to the model data DN includes a three-dimensional model data and a processing program for processing the processed finished shape indicated by the three-dimensional model data. In this way, the database 60 stores a plurality of sets of combinations of the three-dimensional model data and the processing program. The processing program for the three-dimensional model data is a program used when processing the processed finished shape corresponding to the three-dimensional model data. Each of the model data D1 to the model data DN is input from the input unit 11 and stored in the database 60.

In the following description, when it is not necessary to identify the model data DN from the model data D1, any one of the model data D1 to the model data DN may be referred to as model data Dx. The model data Dx of the xth (x is a natural number of any one of 1 to N) includes the 3D model data of the x and the processing program corresponding to the 3D model data of the xth. In this way, in the model data Dx, the xth three-dimensional model data and the xth machining program are associated with each other. That is, the database 60 stores the three-dimensional model and the machining program used when machining the work corresponding to the three-dimensional model in association with each other.

The database 60 may be arranged outside the machining program search device 100. That is, the database 60 may be an external device of the machining program search device 100.

FIG. 1 shows the functional configuration of the machining program search device 100 realized by using the processor 14. The machining program search device 100 includes a model transformation unit 20, a registration model acquisition unit 30, a similarity determination unit 40, and a registration program acquisition unit 50.

The registration model acquisition unit 30 acquires three-dimensional model data from the model data Dx registered in the database 60. The registered model acquisition unit 30 sends the acquired three-dimensional model data to the model conversion unit 20. The registration model acquisition unit 30 copies and acquires the three-dimensional model data of the database 60. That is, the registration model acquisition unit 30 acquires the three-dimensional model data while leaving the three-dimensional model data in the database 60.

The model conversion unit 20 performs data conversion of the three-dimensional model represented by the three-dimensional model data. The model conversion unit 20 converts the 3D model represented by the 3D model data acquired from the input unit 11 and the 3D model represented by the 3D model data acquired from the registered model acquisition unit 30 into grid data. To do. The model transformation unit 20 transmits the data-converted grid data to the similarity determination unit 40.

In the following description, the 3D model represented by the 3D model data acquired from the input unit 11 by the model conversion unit 20 is called a new 3D model, and the 3D model acquired by the model conversion unit 20 from the registered model acquisition unit 30. The 3D model indicated by the data may be called a registered 3D model. Further, the grid data corresponding to the new 3D model may be referred to as new grid data, and the grid data corresponding to the registered 3D model may be referred to as registered grid data. Further, if the new 3D model is the first 3D model, the registered 3D model can be said to be the second 3D model.

FIG. 2 is a diagram showing a configuration of a model transformation unit 20 included in the machining program search device 100 according to the first embodiment. The model transformation unit 20 includes a grid data generation unit 21 and a data compression unit 22.

The grid data generation unit 21 generates grid data based on the three-dimensional model represented by the three-dimensional model data. The grid data is obtained by dividing the shape of the three-dimensional model into data for each grid, which is a virtual three-dimensional space in three-dimensional coordinates. The grid data generation unit 21 generates new grid data of the new three-dimensional model and converts the new three-dimensional model into new grid data. Further, the grid data generation unit 21 generates the registered grid data of the registered 3D model and converts the registered 3D model into the registered grid data. The virtual three-dimensional space used when creating the grid data is composed of three axes of X-axis, Y-axis, and Z-axis. The virtual three-dimensional space is formed by a plurality of divided spaces that divide the virtual three-dimensional space along each of these three axes. The grid data is created by arranging a 3D model in this virtual 3D space. That is, the new lattice data indicates whether each of the plurality of divided spaces in the virtual three-dimensional space is located inside, outside, or at the boundary of the first three-dimensional model. It can be said to be the first position information. In addition, the registered lattice data indicates whether each of the plurality of divided spaces in the virtual three-dimensional space is located inside, outside, or at the boundary of the second three-dimensional model. It can be said to be the second position information.

The data compression unit 22 selects the grid data of a plurality of predetermined grid blocks from the registered grid data and the new grid data generated by the grid data generation unit 21, and selects the grid data of a plurality of selections. Compress into one compression grid data.

The similarity determination unit 40 determines the degree of similarity in shape between the three-dimensional model indicated by the three-dimensional model data acquired from the input unit 11 and the three-dimensional model indicated by the three-dimensional model data acquired from the database 60. .. Similarity determination unit 40, by determining the similarity between the grid data of grid data and the registered three-dimensional model of the new three-dimensional model, determining the similarity of the shape of the new three-dimensional model with the registered three-dimensional model ..

FIG. 3 is a diagram showing a configuration of a similarity determination unit 40 included in the machining program search device 100 according to the first embodiment. The similarity determination unit 40 includes a hash value calculation unit 41, a Hamming distance calculation unit 42, and a Hamming distance sorting unit 43.

The hash value calculation unit 41 calculates the hash value from the grid data obtained by converting the three-dimensional model in the model conversion unit 20. The hash value calculation unit 41 transmits the calculated hash value to the Hamming distance calculation unit 42.

The hash value is a non-regular fixed-length value obtained from the original data by a predetermined calculation procedure. The calculation procedure for obtaining the hash value is called a hash function, a summary function, a message digest function, or the like. The hash value has a predetermined length regardless of the length of the original data, and the same hash value can always be obtained from the same data. On the other hand, the hash values are slightly different data. A completely different hash value can be obtained from a plurality of data. The hash value is calculated through a calculation process that is irreversible and includes a loss of information. Therefore, the original data cannot be restored from the hash value.

The Hamming distance calculation unit 42 calculates the Hamming distance from the hash value of the new 3D model and the hash value of one registered 3D model. The Hamming distance is the number of digits of different values at the corresponding positions in the two values when comparing two values having the same number of digits. When two values with the same number of digits are compared, it is determined that the smaller the Hamming distance, the higher the similarity. The Hamming distance calculation unit 42 transmits the calculated Hamming distance to the Hamming distance sorting unit 43.

The Hamming distance sorting unit 43 determines the similarity of all registered 3D models to the new 3D model based on the Hamming distances of all registered 3D models calculated by the Hamming distance calculating unit 42.

The registration program acquisition unit 50 acquires a processing program of the registered 3D model having a high degree of similarity to the shape shown by the new 3D model from the database 60 based on the similarity determined by the similarity determination unit 40. That is, the registration program acquisition unit 50 machines the second three-dimensional model having a relatively high degree of shape similarity among the plurality of second three-dimensional models based on the determination result in the similarity determination unit 40. It can be said that the machining program for this purpose is searched from the database 60 and obtained. The registration program acquisition unit 50 outputs the registered three-dimensional model having a high degree of similarity to the output unit 13 in association with the processing program associated with the registered three-dimensional model.

Next, the entire processing procedure of the search process executed by the machining program search device 100 will be described. FIG. 4 is a flowchart showing a processing procedure of the machining program search apparatus 100 according to the first embodiment.

In step S1, the model transformation unit 20 acquires the three-dimensional model data from the CAD system 110 via the input unit 11.

In step S2, the model transformation unit 20 acquires all the three-dimensional model data registered in the database 60. Specifically, the registered model acquisition unit 30 acquires all the three-dimensional model data from the database 60 and transmits it to the model conversion unit 20.

In step S3, the model conversion unit 20 converts the 3D model shown in the 3D model data into grid data. That is, the model conversion unit 20 data the new 3D model shown in the 3D model data acquired from the CAD system 110 and the registered 3D model shown in the 3D model data acquired from the database 60 into grid data. Convert.

The process executed by the model transformation unit 20 will be described. First, the process executed by the grid data generation unit 21 will be described. FIG. 5 is a flowchart showing a processing procedure of the grid data generation unit 21 of the model transformation unit 20 according to the first embodiment. The model conversion unit 20 converts the new 3D model acquired from the input unit 11 and all the registered 3D models acquired from the database 60 by the registration model acquisition unit 30 into grid data.

As described above, the database 60 stores the three-dimensional model data and the machining program for machining the machined finished shape indicated by the three-dimensional model data in association with each other. The registration model acquisition unit 30 connects to the database 60 and acquires all the three-dimensional model data registered from the database 60.

In step S10, the lattice data generation unit 21 arranges the three-dimensional model represented by the three-dimensional model data on the lattice space which is a predetermined virtual three-dimensional space. Specifically, the lattice data generation unit 21 generates the shape of the three-dimensional model shown by the three-dimensional model data acquired from the input unit 11, and the three axes of the X-axis, the Y-axis, and the Z-axis are predetermined. Arrange in a lattice space consisting of lengths. The reciprocal space is a virtual three-dimensional space, and is a space in which lattices, which are a plurality of divided spaces having predetermined lengths, are arranged three-dimensionally along each of the three axes. That is, the lattice data generation unit 21 arranges the three-dimensional model on the coordinate system of the lattice space in which the three axes have predetermined lengths and a plurality of lattices are arranged.

FIG. 6 is a diagram showing an example in which a three-dimensional model shown by the three-dimensional model data acquired by the grid data generation unit 21 of the model conversion unit 20 according to the first embodiment is drawn. FIG. 7 is a diagram showing an example in which a three-dimensional model represented by the three-dimensional model data acquired by the lattice data generation unit 21 of the model conversion unit 20 according to the first embodiment is arranged and drawn on the lattice space.

In the lattice space, for example, a 1 mm square cube is regarded as one lattice block, and 200 lattice blocks are arranged in each of the X-axis direction, the Y-axis direction, and the Z-axis direction on three-dimensional coordinates. An example is a virtual three-dimensional space composed of × 200 × 200 = 8000000 lattice blocks. The size and number of grid blocks may be appropriately changed depending on the three-dimensional model to be handled.

In step S20, the grid data generation unit 21 determines the inside and outside of the shape of the three-dimensional model at the vertices of the grid block. Specifically, the lattice data generation unit 21 has the three-dimensional coordinate values of the respective vertices of all the lattice blocks in the lattice space inside the three-dimensional model, outside the three-dimensional model, or three-dimensional. Determine if it is at the boundary of the model. The determination of where the three-dimensional coordinate values of the vertices of the lattice block are inside, outside, or the boundary portion of the three-dimensional model can be performed by performing an inside / outside determination process of the shape of a known solid model.

The lattice data generation unit 21 assigns a value of 1 to the vertices of the lattice block determined that the three-dimensional coordinate values of the vertices of the lattice block are inside the three-dimensional model. The lattice data generation unit 21 assigns a value: 0 to the vertices of the lattice block determined that the three-dimensional coordinate values of the vertices of the lattice block are outside the three-dimensional model. Grid data generation unit 21, the three-dimensional coordinate values of the vertices of the lattice block, the apex of the determined lattice block and the boundary portion of the three-dimensional model, a value: Assign 0.5.

In step S30, the grid data generation unit 21 divides the sum of the values assigned to all the vertices of one grid block by the number of vertices, and assigns the calculated value to the grid block. Hereinafter, the value assigned to the grid block is referred to as grid data. The grid data generation unit 21 generates new grid data of the new 3D model acquired from the input unit 11 by performing the above processing on all the grid blocks, and converts the new 3D model into new grid data. Convert. As a result, in the first 3D model, new lattice data is used as the first position information indicating whether each of the divided spaces in the 3D space is located inside, outside, or the boundary portion of the 3D model. Can be obtained.

In the allocation of the grid data to the grid blocks performed in step S30, for example, when all eight vertices of one grid block are inside the three-dimensional model, the value: 1 = 8/8 is assigned to the grid blocks. Be done. When all eight vertices of one grid block are outside the three-dimensional model, the grid block is assigned a value: 0 = 0/8. If seven vertices in one grid data are inside the 3D model and one vertex is outside the 3D model, the grid block is assigned 0.875 = 7/8. If the 5 vertices of one grid block are inside the 3D model, the 2 vertices are outside the 3D model, and the 1 vertex is at the boundary of the 3D model, then 0.6875 = 5 in the grid block. .5 / 8 is assigned.

Then, the lattice data generation unit 21 performs the above processing on each of the registered 3D models for all the registered 3D models acquired from the registered model acquisition unit 30 to generate lattice data, and the registered 3D model is generated. Convert the model to registered grid data. As a result, in each of the second three-dimensional models, as a plurality of second position information indicating whether each of the divided spaces in the three-dimensional space is located inside, outside, or the boundary portion of the three-dimensional model. , Multiple registered grid data can be acquired. The grid data generation unit 21 transmits the generated new grid data and the registered grid data to the data compression unit 22.

The grid data generation unit 21 transmits the new three-dimensional model and the new grid data corresponding to the new three-dimensional model to the data compression unit 22 in association with each other. Further, the grid data generation unit 21 transmits the registered three-dimensional model and the registered grid data corresponding to the registered three-dimensional model to the data compression unit 22 in association with each other.

Next, the process executed by the data compression unit 22 will be described. FIG. 8 is a flowchart showing a procedure of processing of the data compression unit 22 of the model transformation unit 20 according to the first embodiment.

In step S110, the data compression unit 22 selects the grid data of a plurality of predetermined grid blocks from the new grid data generated by the grid data generation unit 21. The number of selections is the number of a plurality of grid blocks selected by the data compression unit 22 to compress the grid data, and corresponds to the number of grid data to be compressed. Then, the data compression unit 22 calculates the total sum of the grid data assigned to the selected plurality of grid blocks.

In step S120, the data compression unit 22 divides the sum of the obtained grid data by the number of selections, and divides the calculated value as the compression grid data which is the grid data of one compression grid block composed of the selections of grid blocks. Allocate data. The compressed grid data is grid data in which the grid data of the selected grid blocks is compressed, which is assigned to one compressed grid block composed of the selected grid blocks. As a result, the data compression unit 22 can compress a plurality of predetermined number of selected grid data into one compressed grid data.

When compressing the grid data in half, the number of selections may be 2 to the 3rd power = 8. When compressing the grid data to 1/4, the number of selections may be 4 to the 3rd power = 64. For example, when 8 grid blocks are selected and the grid data is compressed in half, if the grid data of the selected 8 grid blocks is all 1, the value of the compressed grid data of the compressed grid block is It becomes 1. When 8 grid blocks are selected and the grid data is compressed in half, the grid data of 7 grid blocks out of the 8 selected grid blocks is 1, and the grid data of 1 grid block When the grid data is 0.75, the value of the compressed grid data of the compressed grid block is 0.96875 = 7.75 / 8.

The data compression unit 22 compresses the grid data of the three-dimensional model by performing the above processing on all the grid data of the three-dimensional model. It is not always necessary to compress the grid data, and the compression rate may be determined according to various conditions such as the capacity of the computer, the usage environment of the computer, and the calculation cost of the computer.

The data compression unit 22 performs the above processing on all of the new grid data and the registered grid data acquired from the grid data generation unit 21. The data compression unit 22 transmits the compressed grid data, which is the compressed grid data of the new grid data and the registered grid data, to the similarity determination unit 40.

The data compression unit 22 transmits the new three-dimensional model and the compressed new grid data corresponding to the new three-dimensional model to the similarity determination unit 40 in association with each other. The compressed new grid data is the compressed new grid data. Further, the data compression unit 22 transmits the registered three-dimensional model and the compressed registration grid data corresponding to the registered three-dimensional model to the similarity determination unit 40 in association with each other. The compressed registration grid data is compressed registration grid data.

Returning to FIG. 4, in step S4, the similarity determination unit 40 determines the degree of similarity in shape between the new three-dimensional model and the registered three-dimensional model. Specifically, the similarity determination unit 40 determines the similarity between the lattice data of the new 3D model and the lattice data of the registered 3D model , and determines the degree of similarity in shape between the new 3D model and the registered 3D model. To judge. That is, in the first embodiment, the similarity determination unit 40 compares the compressed new grid data corresponding to the new three-dimensional model with the compressed registered grid data corresponding to the registered three-dimensional model, and compares the compressed new grid data with the compressed new grid data. Determine the degree of similarity with the compressed registration grid data.

The process executed by the similarity determination unit 40 will be described. FIG. 9 is a flowchart showing a procedure of processing of the similarity determination unit 40 according to the first embodiment.

In step S210, the hash value calculation unit 41 calculates the hash value from the grid data obtained by converting the three-dimensional model in the model conversion unit 20. The hash value calculation unit 41 calculates the hash value of the new 3D model from the compressed new grid data corresponding to the new 3D model. The hash value calculation unit 41 calculates the hash value of the registered 3D model from the compressed registration grid data corresponding to the registered 3D model. When the new grid data and the registered grid data are not compressed, the hash value may be calculated from the new grid data and the registered grid data.

In step S220, the Hamming distance calculation unit 42 calculates the Hamming distance from the hash value of the new three-dimensional model and the hash value of one registered three-dimensional model. The Hamming distance calculation unit 42 calculates the Hamming distance for the hash values of all the registered three-dimensional models. The Hamming distance calculation unit 42 transmits the calculated Hamming distance to the Hamming distance sorting unit 43.

In step S230, the Hamming distance sorting unit 43 sorts the Hamming distance calculated by the Hamming distance calculating unit 42. Specifically, the Hamming distance sorting unit 43 arranges the Hamming distances of all the registered three-dimensional models calculated by the Hamming distance calculation unit 42 in ascending order from the smallest numerical value to the largest numerical value.

Then, the Hamming distance sorting unit 43 determines the ascending order of the Hamming distances of the arranged registered 3D models as the order of the degree of similarity in shape between the new 3D model and the registered 3D model. That is, when the Hamming distance sorting unit 43 compares a plurality of Hamming values calculated by the Hamming distance calculation unit 42, the smaller the Hamming distance value, the higher the degree of similarity in shape with the new 3D model. Is determined. The information in ascending order of the Hamming distances of the registered 3D models arranged in ascending order is the similarity order information which is the information in the order of the degree of similarity of the shapes between the new 3D model and the registered 3D model.

Then, the Hamming distance sorting unit 43 transmits the registered 3D model, the Hamming distance corresponding to the registered 3D model, and the similarity order information corresponding to the registered 3D model in association with each other to the registration program acquisition unit 50. To do.

The degree of similarity is determined not by the method using the hash value, but by the grid data of the new 3D model at the same position when the new 3D model and the registered 3D model are arranged on the same coordinate system. It is also possible to use a method of calculating the number of digits of different values by comparing with the grid data of the registered 3D model. In this case, it is determined that the smaller the number of digits of different values, the higher the similarity. In this case, the compression grid data of the new 3D model and the compression grid data of the registered 3D model may be compared to calculate the number of digits having different values.

Further, the similarity may be determined by using a method of extracting the feature amount of the grid data, searching for the approximate nearest neighbor, vectorizing the grid data to obtain the difference, or the like.

Returning to FIG. 4, in step S5, the registration program acquisition unit 50 registers the registered three-dimensional models having the highest degree of similarity shown in the similarity order information acquired from the Hamming distance sorting unit 43 of the similarity determination unit 40 in this order. The machining program corresponding to the dimensional model is acquired from the database 60. The registration program acquisition unit 50 transmits the registered three-dimensional model, the machining program, the similarity order information, and the similarity information acquired from the similarity determination unit 40 and the database 60 to the output unit 13 in association with each other. .. The similarity information is information on the order of similarity of the registered three-dimensional model in the similarity order information.

In step S6, the output unit 13 outputs the machining program, the registered three-dimensional model, the similarity order information, and the similarity information acquired from the registration program acquisition unit 50 to a display device or the like and presents them to the user. To do. As a result, the machining program search device 100 can search the machining program of the registered three-dimensional model similar to the new three-dimensional model from the database 60 and present it to the user.

FIG. 10 is a diagram showing an example of a new three-dimensional model according to the first embodiment. FIG. 11 is a diagram showing an example of a registered three-dimensional model according to the first embodiment. FIG. 12 is a diagram showing a registered 3D model searched as a registered 3D model similar to the new 3D model shown in FIG. 10 from the 20 registered 3D models shown in FIG. By performing the processes of steps S1 to S6 described above in the machining program search device 100, for example, from the 20 registered 3D models shown in FIG. 11, the registration 3 similar to the new 3D model shown in FIG. 10 As the dimensional model, the registered three-dimensional model shown in FIG. 12 is searched from the database 60. As a result, the machining program search device 100 searches the database 60 for the machining program corresponding to the registered 3D model shown in FIG. 12 as a machining program corresponding to the registered 3D model similar to the new 3D model shown in FIG. Will be done.

In the processing program search device 100 according to the first embodiment as described above, when the three-dimensional model shown in the three-dimensional model data is displayed as an image, the shape-encapsulating portion of the three-dimensional object and 3 which are not displayed in the image are displayed. Similar 3D models can be searched automatically and with high accuracy, including the details of the shape including the hole shape formed in the dimensional object. This makes it possible to easily search the database 60 for a machining program to be diverted when creating a new machining program for machining a new new 3D model, and a new machining program for machining a new 3D model. It is possible to save the user's trouble in creating the machining program of.

Further, in the machining program search device 100 according to the first embodiment, the amount of data to be processed can be reduced by calculating the similarity from the compressed lattice data, and the new three-dimensional model can be registered at high speed and with high accuracy. The degree of similarity of the shape with the three-dimensional model can be calculated.

Next, a mode using machine learning in the similarity determination unit 40 of the machining program search device 100 will be described. FIG. 13 is a diagram showing the configuration of the similarity determination unit 40a using machine learning in the first embodiment. The similarity determination unit 40a includes a teacher data generation unit 71, a machine learning unit 72, an inference unit 73, a similarity calculation unit 74, and a similarity sorting unit 75.

The teacher data generation unit 71 performs clustering by inputting grid data of a plurality of three-dimensional models acquired from the model conversion unit 20, and generates teacher data using the clustered result as a label.

The machine learning unit 72 performs supervised machine learning based on a plurality of teacher data prepared in advance and grid data of a three-dimensional model corresponding to the teacher data, and generates a trained model. The teacher data here is teacher data in which the grid data of the three-dimensional model generated by the teacher data generation unit 71 is input and the label which is the classification result is output.

The inference unit 73 uses the trained model learned by the machine learning unit 72, and uses the lattice data of the registered 3D model shown in the 3D model data and the new 3D model data input to the input unit 11. Inference is performed using the grid data of the 3D model as input, and output data is acquired.

The similarity calculation unit 74 uses the output data for the new 3D model output from the inference unit 73 as the multidimensional feature quantity vector of the new 3D model, and the inference unit for the registered 3D model registered in the database 60. The output data output from 73 is used as the multidimensional feature vector of the registered 3D model, and the distance between the multidimensional feature vector of the new 3D model and the multidimensional feature vector of the registered 3D model is used as the degree of similarity. calculate. That is, the similarity calculation unit 74 can be rephrased as a feature amount acquisition unit that acquires the feature amount of the first three-dimensional model and the feature amount of the second three-dimensional model.

The similarity sorting unit 75 arranges the similarity of all the registered three-dimensional models calculated by the similarity calculation unit 74 in ascending order from the smallest numerical value to the largest numerical value. It is sufficient that the similarity sorting unit 75 can sort in descending order of similarity, and even if the similarity value is divided by 1 and the divided values are arranged in descending order from the largest value to the smallest value. Good.

Next, the process executed by the similarity determination unit 40a will be described. FIG. 14 is a flowchart showing a procedure of processing of the similarity determination unit 40a according to the first embodiment.

In step S310, the teacher data generation unit 71 acquires grid data of a plurality of three-dimensional models from the model transformation unit 20.

In step S320, the teacher data generation unit 71 clusters the grid data of the plurality of 3D models with the grid data of the plurality of 3D models as input data, and the clustered result is the label which is the classification result of the 3D model. To generate teacher data. That is, the teacher data generation unit 71 generates teacher data by inputting the grid data of the three-dimensional model and outputting the label which is the classification result of the three-dimensional model. The case where the k-means method is used for clustering will be described below.

The k-means method is a method of classifying input data into clusters having the closest Euclidean distance of features. Specifically, first, the input data is randomly classified into k clusters determined in advance. Then, the average value of the feature amount of the input data in each of the k clusters is obtained, and the image data that visualizes the average value of the feature amount is created for each cluster. Next, for each image data, the input data is reclassified for the cluster having the average value of the features having the closest Euclidean distance from the features of the input data. The clustering process is completed when this reclassification is no longer performed. The clustering result when the clustering process is completed is determined as a label which is the classification result of the three-dimensional model. When the label of the clustering result is biased, a method may be performed in which the data obtained by projecting the grid data on the XY plane, the YZ plane, and the ZX plane is used as the input data.

The teacher data may be generated by clustering that automatically classifies the given input data without an external standard, and algorithms such as Ward's method, group average method, shortest distance method, and longest distance method may be used. In addition, the user may manually set each three-dimensional model.

FIG. 15 is a diagram showing a three-dimensional model classified into the first cluster C1 in an example in which 20 three-dimensional models are clustered into four clusters by the k-means method in the first embodiment. FIG. 16 is a diagram showing a three-dimensional model classified into a second cluster C2 in an example in which 20 three-dimensional models are clustered into four clusters by the k-means method in the first embodiment. FIG. 17 is a diagram showing a three-dimensional model classified into the third cluster C3 in the example in which 20 three-dimensional models are clustered into four clusters by the k-means method in the first embodiment. FIG. 18 is a diagram showing a three-dimensional model classified into the fourth cluster C4 in the example in which 20 three-dimensional models are clustered into four clusters by the k-means method in the first embodiment.

In step S330, the machine learning unit 72 receives the grid data of the three-dimensional model generated by the teacher data generation unit 71 as input, and outputs a label which is the classification result of the three-dimensional model. Based on the above and the grid data of the 3D model corresponding to the teacher data, supervised machine learning is performed to generate a trained model. The trained model is a trained model for inferring an index using the classification result which is the result of clustering the three-dimensional model from the grid data of the three-dimensional model. Machine learning may be repeated until the desired convergence result is obtained. The machine learning unit 72 stores the classification conditions of the generated three-dimensional model, which is the learned model, in the memory 15. Therefore, the machine learning unit 72 clusters the grid data acquired from the plurality of three-dimensional models to acquire the teacher data, and combines the grid data and the teacher data acquired from the plurality of three-dimensional models into a data set. Generate a trained model based on the set.

FIG. 19 is a diagram showing a configuration of a machine learning unit according to the first embodiment. FIG. 20 is a diagram showing a configuration of a neural network used by the machine learning device according to the first embodiment.

The machine learning unit 72 is a classification condition learning unit having a data acquisition unit 721 and a learning unit 722. The data acquisition unit 721 is used for a plurality of prepared teacher data and teacher data in which the grid data of the three-dimensional model output from the teacher data generation unit 71 is input and the label which is the classification result is output. The grid data of the corresponding three-dimensional model is acquired and sent to the learning unit 722 as a data set. The grid data of the three-dimensional model corresponding to the teacher data is the grid data input when the teacher data is created. That is, in the machine learning unit 72, a three-dimensional model showing the processed finished shape of the processing target is arranged in the three-dimensional space, and in the three-dimensional model, each of the plurality of divided spaces in the three-dimensional space is inside and outside the three-dimensional model. Alternatively, the data set includes position information indicating which part of the boundary portion is located, and teacher data in which a plurality of three-dimensional models are classified into a plurality of classifications. The data acquisition unit 721 acquires a data set.

The learning unit 722 has a plurality of prepared teacher data in which the grid data of the three-dimensional model is input and the label which is the classification result is output, and the three-dimensional model data corresponding to the teacher data. Based on the data set created based on the combination with the grid data of the model, the classification conditions of the 3D model represented by the 3D model data are learned. The learning unit 722 stores the classification conditions of the learned three-dimensional model in the memory 15. That is, the learning unit 722 generates a learned model for the inference unit 73 to infer an index using the classification result of the three-dimensional model based on the data set.

The machine learning unit 72 learns the classification conditions for classifying the lattice data into labels, which are the classification results, that is, the classification conditions of the three-dimensional model, for example, by so-called supervised learning according to the neural network model. Then, the machine learning unit 72 learns the classification conditions of the three-dimensional model to generate a trained model for inferring an index using the classification result of the three-dimensional model. By having the machine learning unit 72 learn what kind of label the grid data is classified into, the inference unit 73 can determine which label the newly input grid data is close to. Here, supervised learning refers to a model in which a large number of sets of data of a certain input and a result (label) are given to a learning device to learn the features in those data sets and estimate the result from the input. ..

A neural network is composed of an input layer composed of a plurality of neurons, an intermediate layer (hidden layer) composed of a plurality of neurons, and an output layer composed of a plurality of neurons. The intermediate layer may be one layer or two or more layers. Further, other methods such as a support vector machine may be used for machine learning.

For example, the neural network shown in FIG. 20 is a three-layer neural network. The input layer contains neurons X1, X2, X3. The middle layer contains neurons Y1 and Y2. The output layer contains neurons Z1, Z2, Z3. The number of neurons in each layer is arbitrary. The plurality of values input to the input layer are multiplied by the weights W1, w11, w12, w13, w14, w15, and w16, and input to the intermediate layer. The plurality of values input to the intermediate layer are multiplied by the weights W2, w21, w22, w23, w24, w25, and w26, and output from the output layer. The output result output from the output layer changes according to the values of the weights W1 and W2.

Subsequently, in step S340, the reasoning unit 73 uses the trained trained model generated by the machine learning unit 72 to grid the registered 3D model represented by the 3D model data registered in the database 60. Inference is performed using the data and the grid data of the new 3D model indicated by the 3D model data input to the input unit 11 as input, and the output data is acquired. That is, the acquisition of the output data by inference is performed by the grid data of the registered 3D model represented by all the 3D model data registered in the database 60 and the new 3D model data input to the input unit 11. Output data is acquired for the grid data of the 3D model.

The inference unit 73 applies a certain input data to a trained model generated by machine learning, and outputs a probability of becoming a label which is a classification result at the time of supervised learning. The inference unit 73 outputs the probability of becoming a label for all the labels that are the classification results at the time of supervised learning. Therefore, the inference unit 73 outputs numerical values for the number of labels as output data. Here, the inference unit 73 applies the grid data of the new 3D model represented by the 3D model data registered in the database 60 to the trained model, and the probability of becoming a label which is the classification result at the time of supervised learning. Is output. The inference unit 73 may be configured to output a multidimensional feature vector described later. That is, the inference unit 73 infers an index using the classification result of the three-dimensional model using the trained model.

Subsequently, in step S350, the similarity calculation unit 74 uses the output data output from the inference unit 73 for the new 3D model input from the input unit 11 as the multidimensional feature quantity vector of the new 3D model, and also The output data output from the inference unit 73 for the registered 3D model registered in the database 60 is used as the multidimensional feature quantity vector of the registered 3D model. Then, the similarity calculation unit 74 compares the multidimensional feature amount vector of the new three-dimensional model with the multidimensional feature amount vector of all the registered three-dimensional models, performs the nearest neighbor search, and performs a new three-dimensional model. The distance between the multidimensional feature amount vector of and the multidimensional feature amount vector of the registered 3D model is calculated as the degree of similarity. The similarity is the similarity between the shape of the new 3D model and the shape of the registered 3D model. The similarity calculation unit 74 transmits the calculated similarity to the similarity sorting unit 75.

To calculate the similarity, the distance scale may be obtained by comparing the multidimensional feature vector of the new 3D model with the multidimensional feature vector of the registered 3D model, and an algorithm such as approximate nearest neighbor search is used. You may.

Subsequently, in step S360, the similarity sorting unit 75 sorts the similarity calculated by the similarity calculation unit 74. Specifically, the similarity sorting unit 75 arranges the similarity of all the registered three-dimensional models calculated by the similarity calculation unit 74 in ascending order from the smallest numerical value to the largest numerical value. Here, the similarity sorting unit 75 arranges the similarity in ascending order from the smallest numerical value to the largest numerical value, but the similarity sorting unit 75 only needs to be able to arrange the similarity in descending order of the similarity numerical value. 1 may be divided by, and the divided numerical values may be arranged in descending order from the largest numerical value to the smallest numerical value.

By performing the above processing, the similarity determination unit 40a can determine the similarity of all the registered three-dimensional models in the same manner as the similarity determination unit 40.

In the machining program search device 100 according to the present embodiment as described above, machine learning of the classification conditions of the three-dimensional model is performed by the similarity determination unit 40a, and the new 3 is based on the learned classification conditions of the three-dimensional model. The degree of similarity between the shapes of the dimensional model and the registered 3D model is calculated. As a result, the shape features of the 3D model can be efficiently recognized, the degree of shape similarity between the new 3D model and the registered 3D model can be calculated, and a new 3D model for processing a new new 3D model can be processed. It is possible to easily search the database 60 for the machining program to be diverted when creating the machining program of.

FIG. 21 is a diagram showing registered 3D models sorted in descending order of similarity to the new 3D model shown in FIG. In FIG. 21, the registered 3D model shown on the left side has a higher degree of similarity to the new 3D model shown in FIG. 10, and the registered 3D model shown on the right side has a higher degree of similarity to the new 3D model shown in FIG. The similarity to the 3D model is low.

Next, another form of the model conversion unit 20 will be described. First, the data expansion executed by the model transformation unit 20 will be described. FIG. 22 is a diagram showing a configuration of a model transformation unit 20a that executes data expansion in the machining program search device 100 shown in FIG. The model transformation unit 20a includes a model shape transformation unit 23, the above-mentioned grid data generation unit 21, and the above-mentioned data compression unit 22. The model shape conversion unit 23 converts the shape of the new three-dimensional model input from the input unit 11 and arranged on the lattice space. The shape transformation means at least one of rotational movement of the shape, enlargement of the shape, and reduction of the shape.

Next, the processing executed by the model transformation unit 20a will be described. FIG. 23 is a flowchart showing the data expansion process executed by the model transformation unit 20a according to the first embodiment.

First, the grid data generation unit 21 carries out the above-mentioned step S10.

Next, in step S410, the model shape conversion unit 23 transforms the shape of the new three-dimensional model arranged on the lattice space.

After that, the above-mentioned steps S20, S30, step S110 and step S120 are carried out on the new three-dimensional model on which the shape conversion is carried out, and the grid data is carried on on the new three-dimensional model on which the shape conversion is carried out. The grid data is compressed from the generation of.

In the machining program search device 100 according to the present embodiment as described above, the new three-dimensional model arranged on the lattice space is shape-converted into various directions or various sizes, and then the lattice of the three-dimensional model is converted. Can generate data. Further, one of the new three-dimensional model and the registered three-dimensional model may be shape-transformed so that the directions and sizes on the lattice space are the same. As a result, the shape features of the three-dimensional model can be accurately recognized and the degree of similarity can be calculated. In addition, there is no need to restrict the orientation of the new 3D model and the registered 3D model in the 3D model data in advance.

Further, the machining program search device 100 uses the model conversion unit 20a and the similarity determination unit 40a to machine-learn the three-dimensional model by expanding the number of data in various directions or in various sizes to make it similar. Since the degree can be calculated, the shape feature of the three-dimensional model can be accurately recognized and the degree of similarity can be calculated.

As described above, in the machining program search device 100 according to the first embodiment, when the three-dimensional model shown in the three-dimensional model data is displayed as an image, the shape inclusion portion of the three-dimensional object and the shape inclusion portion of the three-dimensional object that are not displayed in the image are displayed. Similar 3D models can be searched automatically and with high accuracy, including the details of the shape including the hole shape formed in the 3D object. As a result, the machining program search device 100 can easily search the database 60 for the machining program to be diverted when creating a new machining program for machining a new new 3D model, and the new 3D model can be searched. It is possible to reduce the load on the user in creating a new machining program for machining the model.

Further, in the machining program search device 100 according to the first embodiment, the amount of data to be processed can be reduced by calculating the similarity from the compressed lattice data, and the new three-dimensional model can be registered at high speed and with high accuracy. The degree of similarity of the shape with the three-dimensional model can be calculated.

Further, in the machining program search device 100, machine learning of the classification conditions of the three-dimensional model is performed in the similarity determination unit 40a, and a new three-dimensional model and a registered three-dimensional model are performed based on the learned classification conditions of the three-dimensional model. Calculate the similarity of the shape with. As a result, the shape features of the 3D model can be efficiently recognized, the degree of shape similarity between the new 3D model and the registered 3D model can be calculated, and a new 3D model for processing a new new 3D model can be processed. It is possible to easily search the database 60 for the machining program to be diverted when creating the machining program of.

Therefore, the machining program search device 100 according to the first embodiment has an effect that a machining program for machining a shape similar to the machined finished shape after machining of the work can be automatically and easily searched from the database. ..

Embodiment 2.
FIG. 24 is a diagram showing the configuration of the machining program search device 101 according to the second embodiment. Of the components of FIG. 24, the components that achieve the same function as the machining program search device 100 of the first embodiment shown in FIG. 1 are designated by the same reference numerals, and redundant description will be omitted.

The machining program search device 101 according to the second embodiment recognizes a difference between the shape of the three-dimensional model in which the machining program is newly created and the shape of the three-dimensional model in which the machining program is created in the past, and recognizes the difference. Output the machining program for the part. That is, the machining program search device 101 recognizes the difference portion between the shape of the new three-dimensional model and the shape of the registered three-dimensional model, and outputs the machining program related to the difference portion.

The machining program search device 101 includes a difference recognition unit 80 in addition to the components included in the machining program search device 100. The difference recognition unit 80 is connected to the input unit 11, the registration model acquisition unit 30, the registration program acquisition unit 50, and the output unit 13.

The difference recognition unit 80 recognizes the difference between the shape of the three-dimensional model in which the machining program is newly created and the shape of the three-dimensional model in which the machining program is created in the past. That is, the machining program search device 101 recognizes the difference between the shape of the new three-dimensional model and the shape of the registered three-dimensional model.

FIG. 25 is a diagram showing a configuration of a difference recognition unit 80 included in the machining program search device 101 according to the second embodiment. The difference recognition unit 80 includes a model difference acquisition unit 81 and a machining program difference acquisition unit 82.

The model difference acquisition unit 81 acquires a new 3D model represented by the 3D model data input from the input unit 11 and a registered 3D model represented by the 3D model data acquired by the registration model acquisition unit 30. .. Then, the model difference acquisition unit 81 acquires information on the shape difference between the shape of the new three-dimensional model and the shape of the registered three-dimensional model.

The machining program difference acquisition unit 82 acquires the machining program portion related to the difference portion based on the information of the shape difference portion between the shape of the new 3D model and the shape of the registered 3D model acquired from the model difference acquisition unit 81. To do. That is, the machining program difference acquisition unit 82 acquires the difference between the program for machining the machined finished shape of the new 3D model and the program for machining the machined finished shape of the registered 3D model.

Next, the process executed by the difference recognition unit 80 will be described. FIG. 26 is a flowchart showing a processing procedure of the difference recognition unit 80 according to the second embodiment.

In step S510, the model difference acquisition unit 81 acquires the 3D model data of the new 3D model input from the input unit 11, and also acquires the 3D model data of the registered 3D model from the registered model acquisition unit 30. , Acquires the shape difference between the new 3D model and the registered 3D model. Specifically, the model difference acquisition unit 81 calculates the first difference shape, which is the shape obtained by subtracting the shape of the registered three-dimensional model from the shape of the new three-dimensional model. Further, the model difference acquisition unit 81 calculates the second difference shape, which is the shape obtained by subtracting the shape of the new three-dimensional model from the shape of the registered three-dimensional model. The shape can be subtracted by using the set operation of the solid model. The model difference acquisition unit 81 transmits the calculated information on the first difference shape and the information on the second difference shape to the machining program difference acquisition unit 82.

Next, in step S520, the machining program difference acquisition unit 82 sets the first difference shape and the second difference shape based on the information of the first difference shape and the information of the second difference shape acquired from the model difference acquisition unit 81. Acquire the machining program part corresponding to. The machining program corresponding to the first difference shape needs to be edited or additionally machined with respect to the machining program of the registered 3D model. The machining program corresponding to the second difference shape needs to be edited or reduced compared to the machining program of the registered 3D model. When the second difference shape exists, the difference part machining program corresponding to the difference part of the second difference shape has the processing part of the processing program of the registered three-dimensional model and the shape of the difference part on the three-dimensional coordinates. If they interfere with each other, the machining program portion of the interfering portion becomes the difference portion machining program corresponding to the difference portion. The machining program difference acquisition unit 82 extracts the difference section machining program corresponding to the difference section from the machining program of the registered three-dimensional model.

The machining program difference acquisition unit 82 outputs information on the difference between the shape of the new 3D model and the shape of the registered 3D model, that is, at least one of the information on the first difference shape and the second difference shape. Send to. Further, the machining program difference acquisition unit 82 transmits the information of the machining program portion related to the difference section to the output unit 13.

The output unit 13 outputs the information of the difference portion between the shape of the new 3D model and the shape of the registered 3D model and the difference portion processing program to the display device or the like together with the machining program acquired from the registration program acquisition unit 50. And present it to the user. As a result, the machining program search device 101 can easily present to the user a machining program portion corresponding to the difference portion and the difference portion between the shapes of the registered 3D model similar to the new 3D model and the new 3D model. Can be done.

As described above, in the machining program search device 101 according to the first embodiment, the information of the difference portion between the shape of the new three-dimensional model and the shape of the registered three-dimensional model and the information of the machining program portion related to the difference portion are obtained. To the user. This improves the efficiency of the user's machining program editing work.

The configuration shown in the above-described embodiment shows an example, and can be combined with another known technique, can be combined with each other, and does not deviate from the gist. It is also possible to omit or change a part of the configuration.

11 Input unit, 12 Display unit, 13 Output unit, 14 Processor, 15 Memory, 20 Model conversion unit, 21 Grid data generation unit, 22 Data compression unit, 23 Model shape conversion unit, 30 Registered model acquisition unit, 40, 40a similar Degree judgment unit, 41 hash value calculation unit, 42 humming distance calculation unit, 43 humming distance sorting unit, 50 registration program acquisition unit, 60 database, 71 teacher data generation unit, 72 machine learning unit, 73 inference unit, 74 similarity calculation Unit, 75 similarity sorting unit, 80 difference recognition unit, 81 model difference acquisition unit, 82 machining program difference acquisition section, 100, 101 machining program search device, 110 CAD system, 721 data acquisition section, 722 learning section, C1 1st Cluster, C2 2nd cluster, C3 3rd cluster, C4 4th cluster, D1, DN model data.

Claims (8)

The search target is a database in which multiple different machining programs for machining the machining target in a machine tool are registered.
A first three-dimensional model, which is a three-dimensional model showing the finished shape of the machining target for which a new machining program is created, and a plurality of machining targets machined by the plurality of machining programs registered in the database. A plurality of second three-dimensional models, which are three-dimensional models showing a processed finished shape, are composed of three axes of X-axis, Y-axis, and Z-axis, and are formed by a plurality of divided spaces along each of the three axes. In the first three-dimensional model, each of the plurality of divided spaces in the three-dimensional space is either inside, outside, or at the boundary of the first three-dimensional model. In the first position information indicating whether or not it is located in the portion of, and in each of the plurality of the second three-dimensional models, each of the plurality of divided spaces is inside, outside, or a boundary portion of the second three-dimensional model. A model conversion unit that acquires a plurality of second position information indicating which part of the two is located, and
A similarity determination unit that determines the degree of similarity in shape between the first three-dimensional model and the second three-dimensional model by comparing the first position information with the second position information.
Based on the determination result in the similarity determination unit, the machining program for machining the second three-dimensional model having a relatively high similarity in shape among the plurality of second three-dimensional models is provided. The registration program acquisition unit that searches and acquires from the database,
A machining program search device characterized by being equipped with. Provided with an output unit that outputs the machining program acquired by the registration program acquisition unit.
The machining program search apparatus according to claim 1. The model conversion unit rotates and moves the shape, enlarges the shape, and reduces the shape with respect to at least one of the first three-dimensional model and the second three-dimensional model arranged in the three-dimensional space. Performing at least one of these processes to acquire the first position information and the plurality of the second position information.
The machining program search apparatus according to claim 1 or 2. The similarity determination unit compares the first position information and the plurality of the second position information at the same position in the divided space.
The machining program search device according to any one of claims 1 to 3, wherein the machining program search device is characterized. Provided with a model difference acquisition unit for acquiring information on the shape difference portion between the first three-dimensional model and the plurality of second three-dimensional models.
The machining program search apparatus according to any one of claims 2 to 4, wherein the machining program search device is characterized. Obtaining the difference portion machining program, which is a machining program for machining the second three-dimensional model corresponding to the difference portion of the shape, from the machining program for machining the second three-dimensional model.
The machining program search apparatus according to claim 5. The search target is a database in which multiple different machining programs for machining the machining target in a machine tool are registered.
A first three-dimensional model, which is a three-dimensional model showing the finished shape of the machining target for which a new machining program is created, and a plurality of machining targets machined by the plurality of machining programs registered in the database. A plurality of second three-dimensional models, which are three-dimensional models showing a processed finished shape, are composed of three axes of X-axis, Y-axis, and Z-axis, and are formed by a plurality of divided spaces along each of the three axes. In the first three-dimensional model, each of the plurality of divided spaces in the three-dimensional space is either inside, outside, or at the boundary of the first three-dimensional model. In the first position information indicating whether or not it is located in the portion of, and in each of the plurality of the second three-dimensional models, each of the plurality of divided spaces is inside, outside, or a boundary portion of the second three-dimensional model. A model conversion unit that acquires a plurality of second position information indicating which part of the two is located, and
A three-dimensional model showing the processed finished shape to be processed is arranged in the three-dimensional space, and in the three-dimensional model, each of the plurality of divided spaces in the three-dimensional space is inside, outside, or a boundary portion of the three-dimensional model. A data acquisition unit that acquires position information indicating which part of the three-dimensional model is located, and teacher data that is a classification result obtained by classifying the plurality of three-dimensional models into a plurality of classifications, and the data set. A machine learning unit including a learning unit that generates a trained model for inferring an index using the classification result of the three-dimensional model based on the above.
An index using the classification result of the three-dimensional model in the first position information and the second position output from the first position information and the second position information using the trained model. Based on the index using the classification result of the three-dimensional model in the position information, the feature amount acquisition unit for acquiring the feature amount of the first three-dimensional model and the feature amount of the second three-dimensional model, and the feature amount acquisition unit.
By comparing the feature amount of the first three-dimensional model with the feature amount of the second three-dimensional model, the degree of similarity in shape between the first three-dimensional model and the second three-dimensional model can be determined. Similarity judgment unit to judge and
Based on the determination result in the similarity determination unit, the machining program for machining the second three-dimensional model having a relatively high degree of similarity in shape among the plurality of second three-dimensional models is provided. The registration program acquisition unit that searches and acquires from the database,
A machining program search device characterized by being equipped with. It is a machining program search method in a machining program search device that searches a database in which a plurality of different machining programs for machining a machining target in a machine tool are registered.
A first three-dimensional model, which is a three-dimensional model showing the finished shape of the machining target for which a new machining program is created, and a plurality of machining targets machined by the plurality of machining programs registered in the database. A plurality of second three-dimensional models, which are three-dimensional models showing a processed finished shape, are composed of three axes of X-axis, Y-axis, and Z-axis, and are formed by a plurality of divided spaces along each of the three axes. In the first three-dimensional model, each of the plurality of divided spaces in the three-dimensional space is either inside, outside, or at the boundary of the first three-dimensional model. In the first position information indicating whether or not it is located in the portion of, and in each of the plurality of the second three-dimensional models, each of the plurality of divided spaces is inside, outside, or a boundary portion of the second three-dimensional model. A step of acquiring a plurality of second position information indicating which part of the two is located, and
A step of determining the degree of similarity in shape between the first three-dimensional model and the second three-dimensional model by comparing the first position information with the second position information.
For machining the second three-dimensional model having a relatively high degree of similarity of the shapes among the plurality of the second three-dimensional models based on the determination result in the step of determining the similarity of the shapes. Steps to search and acquire the machining program from the database, and
A machining program search method characterized by including.

Images