KR20200065159A - Machinability Diagnostic Method based on Machine Learning for Optimizing Manufacturing System - Google Patents

Abstract

According to the present invention, a learning-based processability diagnosis method for optimizing a production system comprises the steps of: (a) collecting basic processing information required for processing; (b) processing the basic processing information; (c) deriving a plurality of integrated data sets for each processing experiment condition by reflecting the processing tool information on the basic processing information processed by the step (b); (d) deriving a plurality of parameter sets by combining at least a part of a plurality of cutting condition parameters that affect a cutting force; (e) deriving a plurality of combination data by selecting and combining any one of the plurality of integrated data sets and any one of the plurality of parameter sets; (f) verifying the accuracy of the plurality of combination data; (g) extracting representatively selected combination data from the plurality of combination data of which accuracy is verified by the step (f); and (h) diagnosing processability under the pre-set conditions based on the optimal combination data.

Description

Machinability Diagnostic Method based on Machine Learning for Optimizing Manufacturing System

The present invention relates to a learning-based processability diagnosis method for optimizing a production system, and more specifically, by accurately diagnosing processability from multiple angles using a plurality of learning models and data, so as to accurately predict the processability and processing trend of an object. It relates to a learning-based processability diagnosis method for optimizing production systems.

Under the modern process system, where multi-variety production is the main, a solution that can respond in a timely manner to a customized production request is required.

In particular, the cutting process field is highly influenced by the characteristics of equipment, materials, and tools, so process optimization and proactive response to event occurrence are very important issues.

And one of the top goals in this cutting process is to increase productivity through efficient machining under optimal conditions.

In the conventional case, in order to achieve such a target, a method of finding an optimal condition through a trial-and-error method through data collected over a long period of time is performed by performing an actual processing process through a processing tool.

However, this method had to take time until a sufficiently meaningful data pool was established, and it was difficult to determine whether it was in the actual optimal condition, so productivity improvement was very limited. In addition, when the processing environment and equipment are replaced, there is a problem that data must be newly collected again.

Therefore, a method for solving these problems is required.

Korean Patent Publication No. 10-2014-0083198

The present invention is an invention devised to solve the above-mentioned problems of the prior art, and an object of the present invention is to provide a processability diagnosis method for optimizing a production system by diagnosing processability from multiple angles using a plurality of learning models and data. Have

The problems of the present invention are not limited to the problems mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the following description.

The learning-based processability diagnosis method for optimizing the production system of the present invention for achieving the above object comprises: (a) collecting basic processing information required for processing, (b) processing the basic processing information, Step (c) of deriving a plurality of integrated data sets for each machining test condition by reflecting machining tool information in the basic machining information processed by step (b), at least a part of a plurality of cutting condition parameters affecting cutting force (D) step of deriving a plurality of parameter sets by combining, and (e) step of deriving a plurality of combination data by selecting and combining any one of the plurality of integrated data sets and one of the plurality of parameter sets, and the plurality of Step (f) of verifying the accuracy of the combination data of the dog, step (g) of extracting representatively selected combination data among the plurality of combination data whose accuracy is verified by the step (f), and based on the optimal combination data And (h) diagnosing processability under a preset condition.

And in step (b), performing the machining simulation through the basic processing information to extract the simulation processing data (b-1) and performing the actual processing using the basic processing information, and extracting the actual processing data The step (b-2) may be included.

In addition, step (b) may further include a step (b-3) of synchronizing the simulation processing data and the actual processing data to derive synchronization sample data.

Further, in step (b-3), the period of each of the simulation processing data and the actual processing data may be calculated, and synchronization may be performed according to the calculated cycle to derive the synchronization sample data.

In addition, in step (e), 1:1 matching may be performed on the plurality of integrated data sets and the plurality of parameter sets.

In the step (f), the step (f-1) of defining a plurality of learning models, the step (f-2) of substituting the plurality of combination data into the plurality of learning models, and using the plurality of combination data It may include (f-3) step of performing the actual processing and (f-4) step of verifying the accuracy by comparing the results of the steps (f-2) and (f-3).

In addition, in step (g), the optimal combination data having the highest accuracy among the plurality of combination data may be representatively selected.

In addition, the basic processing information may include at least one or more information among materials to be processed, types of processing equipment, types of processing tools, and processing test conditions.

In addition, the processing tool information may include at least one or more of a type of processing tool for performing processing, the shape of the processing tool, the size of the processing tool, the weight of the processing tool, and the processing angle of the processing tool.

In addition, the cutting condition parameter may include at least one or more information of a supply current, a supply voltage, a rotational speed of the machining tool, and a moving speed of the machining tool.

The learning-based processability diagnosis method for optimizing the production system of the present invention for solving the above problems can meet the needs of the industrial site by improving the added value of the production system by providing a solution for upgrading productivity and quality in the industrial site. There is an advantage.

In addition, the present invention has the advantage of accurately predicting the processability and processing trend of an object by diagnosing the processability from multiple angles using a plurality of learning models and data.

The effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description of the claims.

1 is a view showing each step of a learning-based processability diagnosis method for optimizing a production system according to an embodiment of the present invention;
2 is a diagram showing a process for processing basic processing information in a learning-based processability diagnosis method for optimizing a production system according to an embodiment of the present invention;
3 is a learning-based processability diagnosis method for optimizing a production system according to an embodiment of the present invention, showing a process of deriving a plurality of integrated data sets;
4 is a diagram illustrating a process of deriving a plurality of parameter sets in a learning-based processability diagnosis method for optimizing a production system according to an embodiment of the present invention;
5 is a diagram illustrating a process of deriving a plurality of combination data in a learning-based processability diagnosis method for optimizing a production system according to an embodiment of the present invention; And
6 is a view showing a process for verifying the accuracy of the combination data in a learning-based processability diagnosis method for optimizing a production system according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention, in which the object of the present invention can be specifically realized, will be described with reference to the accompanying drawings. In describing this embodiment, the same name and the same code are used for the same configuration, and additional description will be omitted.

1 is a view showing each step of a learning-based processability diagnosis method for optimizing a production system according to an embodiment of the present invention.

As illustrated in FIG. 1, a learning-based processability diagnosis method for optimizing a production system according to an embodiment of the present invention includes (a) collecting basic processing information required for processing, and processing the basic processing information Steps (b) and (c) deriving a plurality of integrated data sets for each processing experiment condition by reflecting the machining tool information in the basic processing information processed by the step (b), and a plurality of steps affecting the cutting force (D) deriving a plurality of parameter sets by combining at least some of the cutting condition parameters, and selecting and combining any one of the plurality of integrated data sets and one of the plurality of parameter sets to generate a plurality of combination data. Extracting (e) step, (f) step of verifying the accuracy of the plurality of combination data, and extracting representative combination data among the plurality of combination data whose accuracy is verified by the (f) step and (g) and (h) diagnosing processability under a preset condition based on the optimum combination data.

Hereinafter, each of these steps will be described in more detail.

First, in step (a), basic processing information required for a processing process to be performed is collected.

In the case of this embodiment, the basic processing information may include at least one or more information among materials to be processed, types of processing equipment, types of processing tools, and processing test conditions. It is a matter of course that the basic processing information may further include various other information required for a processing process in addition to the various information listed above.

After collecting the basic processing information as described above, step (b) of processing the basic processing information is performed. The step (b) is a process of processing the basic processing information in a form that can be processed by any processor.

And in the case of this embodiment, as shown in Figure 2, the step (b) is a step (b-1) to extract the simulation processing data by performing a processing simulation through the basic processing information, and using the basic processing information And performing the actual processing, and extracting the actual processing data (b-2), and synchronizing the simulation processing data and the actual processing data to derive synchronization sample data (b-3). .

That is, in step (b-1), processing simulation is performed through the collected basic processing information, and in step (b-2), real processing is performed through the collected basic processing information.

Accordingly, the simulation processing data may be extracted in step (b-1), and the actual processing data may be extracted in step (b-2). However, since the simulation processing data and the actual processing data may be incompatible with each other, each data is synchronized through step (b-3).

At this time, step (b-3) may be performed through a process of calculating the periods of each of the simulation processing data and the actual processing data, and performing synchronization according to the calculated cycle to derive the synchronization sample data.

Next, step (c) is performed by deriving a plurality of integrated data sets for each processing test condition by reflecting the processing tool information in the basic processing information processed in the form of synchronized sample data by the step (b).

In the case of this step, as shown in FIG. 3, the machining tool information is reflected in the synchronized sample data derived by step (b-3), wherein the machining tool information is the type of machining tool for performing the machining, machining It may include at least one or more of the shape of the tool, the size of the processing tool, the weight of the processing tool and the processing angle of the processing tool.

Accordingly, n integrated data sets may be generated by a combination of the processing tool information and synchronization sample data. In this embodiment, for convenience of description, serial numbers are sequentially attached to a plurality of integrated data sets to provide first integrated data. The set (D ₁ ), the second integrated data set (D ₂ ), the third integrated data set (D ₃ ), and the fourth integrated data set (D ₄ ) were named in the form of the nth integrated data set.

Next, step (d) of deriving a plurality of parameter sets by combining at least some of the plurality of cutting condition parameters affecting the cutting force is performed.

In this step, as shown in FIG. 4, various parameters are derived by arbitrarily combining each item belonging to a plurality of cutting condition parameters affecting cutting force, wherein the cutting condition parameters are supply current, supply voltage, and machining. It may include at least one or more information of the rotational speed of the tool and the moving speed of the tool.

Accordingly, n parameter sets may be generated by a combination between each item, and in this embodiment, for convenience of description, serial numbers are sequentially assigned to a plurality of parameter sets to set a first parameter (P ₁ ) and a second parameter. Named in the form of the nth parameter set, such as the set P ₂ , the third parameter set P ₃ , and the fourth parameter set P ₄ .

Thereafter, step (e) of deriving a plurality of combination data by selectively combining any one of the plurality of integrated data sets derived by step (c) and one of the plurality of parameter sets derived by step (d) Is performed.

In this step, as illustrated in FIG. 5, 1:1 matching may be performed on n integrated data sets and n parameter sets, respectively, and thus n×n combination data may be derived. In the present invention, it is possible to perform learning-based processability diagnosis using such combination data.

After deriving a plurality of combination data as described above, a step (f) of verifying the accuracy of the plurality of combination data and a plurality of combination data of which the accuracy is verified by the step (f) is typically selected. Step (g) of extracting the combination data and step (h) of diagnosing processability under a preset condition based on the optimum combination data may be sequentially performed.

Also, as shown in FIG. 6, step (f) is a step (f-1) of defining a plurality of learning models and a step (f-2) of substituting the plurality of combination data into the plurality of learning models. And, (f-3) step of performing the actual processing using the plurality of combination data, and (f-2) and (f-3) step to compare the results of verifying the accuracy (f) -4) may be included.

The step (f-1) is a step of designating a plurality of learning models, wherein the learning models are at least one of a self-learning model previously known in the art, a self-learning model in a private state, and a self-learning model to be developed later. The above may be selectively included.

In addition, step (f-2) calculates a result value by substituting a plurality of combination data derived by step (e) into the plurality of learning models, and step (f-3) is different from the plurality of steps. Actual processing is performed through the combination data to calculate the result.

In addition, in step (f-4), accuracy is verified by contrasting the result value calculated in step (f-2) with the result value calculated in step (f-3).

Subsequently, in the step (g), the optimal combination data having the highest accuracy among the plurality of combination data whose accuracy has been verified in the step (f) is typically extracted, and in the step (h), the optimal combination data extracted as described above. Substituting the preset conditions on the basis of, it is possible to diagnose the processability under the conditions.

As described above, the preferred embodiments according to the present invention have been examined, and the fact that the present invention can be embodied in other specific forms without departing from the spirit or scope of the embodiments described above has ordinary knowledge in the art. It is obvious to them. Therefore, the above-described embodiments should be regarded as illustrative rather than restrictive, and accordingly, the present invention is not limited to the above description and may be changed within the scope of the appended claims and their equivalents.

D ₁ ~ D ₄ : Integrated data set
P ₁ ~ P ₄ : Parameter set

Claims (10)

(A) collecting basic processing information required for processing;
(B) processing the basic processing information;
A step (c) of deriving a plurality of integrated data sets for each processing experiment condition by reflecting processing tool information in the basic processing information processed in step (b);
(D) deriving a plurality of parameter sets by combining at least some of a plurality of cutting condition parameters that affect a cutting force;
(E) deriving a plurality of combination data by selectively combining any one of the plurality of integrated data sets and one of the plurality of parameter sets;
(F) verifying accuracy of the plurality of combination data;
(G) extracting representatively selected combination data among the plurality of combination data whose accuracy has been verified by step (f); And
(H) diagnosing processability under a preset condition based on the optimum combination data;
A learning-based processability diagnosis method for optimizing a production system including a. According to claim 1,
Step (b) is,
(B-1) extracting simulation processing data by performing a processing simulation through the basic processing information; And
Step (b-2) of performing actual processing using the basic processing information and extracting actual processing data;
A learning-based processability diagnosis method for optimizing a production system including a. According to claim 2,
Step (b) is,
A learning-based processability diagnosis method for optimizing a production system further comprising the step of (b-3) deriving synchronization sample data by synchronizing the simulation processing data and the actual processing data. According to claim 3,
In step (b-3),
A learning-based processability diagnosis method for optimizing a production system that calculates each of the simulation processing data and the actual processing data and performs synchronization according to the calculated cycle to derive the synchronization sample data. According to claim 1,
Step (e) is,
A learning-based processability diagnosis method for optimizing a production system to perform 1:1 matching on the plurality of integrated data sets and the plurality of parameter sets. According to claim 1,
Step (f) is,
Defining (f-1) a plurality of learning models;
(F-2) substituting the plurality of combination data into the plurality of learning models;
(F-3) performing real machining using the plurality of combination data; And
Step (f-4) of verifying accuracy by comparing the results of steps (f-2) and (f-3);
A learning-based processability diagnosis method for optimizing a production system including a. According to claim 1,
Step (g) is,
A learning-based processability diagnosis method for optimizing a production system to representatively select the optimal combination data having the highest accuracy among the plurality of combination data. According to claim 1,
The basic processing information,
A learning-based processability diagnosis method for optimizing a production system that includes at least one or more of materials to be processed, types of processing facilities, types of processing tools, and processing test conditions. According to claim 1,
The processing tool information,
A learning-based processability diagnosis method for optimizing a production system including at least one of a type of a processing tool for performing processing, a type of processing tool, a size of the processing tool, a weight of the processing tool, and a processing angle of the processing tool. According to claim 1,
The cutting condition parameter,
A learning-based processability diagnosis method for optimizing a production system that includes at least one or more of supply current, supply voltage, rotation speed of a processing tool, and movement speed of a processing tool.

Images