JP7428350B2 - Processing condition recommendation device, processing condition recommendation method, program, metal structure manufacturing system, and metal structure manufacturing method - Google Patents
Processing condition recommendation device, processing condition recommendation method, program, metal structure manufacturing system, and metal structure manufacturing method Download PDFInfo
- Publication number
- JP7428350B2 JP7428350B2 JP2021090839A JP2021090839A JP7428350B2 JP 7428350 B2 JP7428350 B2 JP 7428350B2 JP 2021090839 A JP2021090839 A JP 2021090839A JP 2021090839 A JP2021090839 A JP 2021090839A JP 7428350 B2 JP7428350 B2 JP 7428350B2
- Authority
- JP
- Japan
- Prior art keywords
- laminate
- amount
- heat input
- processing
- heat source
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000012545 processing Methods 0.000 title claims description 106
- 238000000034 method Methods 0.000 title claims description 67
- 239000002184 metal Substances 0.000 title claims description 28
- 229910052751 metal Inorganic materials 0.000 title claims description 28
- 238000004519 manufacturing process Methods 0.000 title claims description 24
- 239000000463 material Substances 0.000 claims description 126
- 239000000203 mixture Substances 0.000 claims description 52
- 239000000654 additive Substances 0.000 claims description 50
- 230000000996 additive effect Effects 0.000 claims description 49
- 230000008569 process Effects 0.000 claims description 42
- 230000007547 defect Effects 0.000 claims description 25
- 238000003754 machining Methods 0.000 claims description 16
- 238000001816 cooling Methods 0.000 claims description 14
- 230000014759 maintenance of location Effects 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 8
- 238000010030 laminating Methods 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 5
- 238000004088 simulation Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 description 24
- 238000010801 machine learning Methods 0.000 description 16
- 238000003475 lamination Methods 0.000 description 8
- 230000000704 physical effect Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 239000000843 powder Substances 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 229910000831 Steel Inorganic materials 0.000 description 5
- 229910001315 Tool steel Inorganic materials 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 5
- 239000010959 steel Substances 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 238000007637 random forest analysis Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 239000008186 active pharmaceutical agent Substances 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 239000011812 mixed powder Substances 0.000 description 3
- 238000010845 search algorithm Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 239000000155 melt Substances 0.000 description 2
- 150000002739 metals Chemical class 0.000 description 2
- 238000010791 quenching Methods 0.000 description 2
- 230000000171 quenching effect Effects 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 238000005496 tempering Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 229910000967 As alloy Inorganic materials 0.000 description 1
- 229910001347 Stellite Inorganic materials 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- AHICWQREWHDHHF-UHFFFAOYSA-N chromium;cobalt;iron;manganese;methane;molybdenum;nickel;silicon;tungsten Chemical compound C.[Si].[Cr].[Mn].[Fe].[Co].[Ni].[Mo].[W] AHICWQREWHDHHF-UHFFFAOYSA-N 0.000 description 1
- 238000013329 compounding Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010894 electron beam technology Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 239000002648 laminated material Substances 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000007769 metal material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000003647 oxidation Effects 0.000 description 1
- 238000007254 oxidation reaction Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 229910001220 stainless steel Inorganic materials 0.000 description 1
- UONOETXJSWQNOL-UHFFFAOYSA-N tungsten carbide Chemical compound [W+]#[C-] UONOETXJSWQNOL-UHFFFAOYSA-N 0.000 description 1
Description
本発明は、金属の付加材を用いた積層造形における、付加材の組成及び加工条件の設定に関する。 The present invention relates to setting the composition of an additive material and processing conditions in additive manufacturing using a metal additive material.
金属材料による構造体の全部または一部を形成する場合に、付加材（例えば金属ワイヤ、金属粉末など）を熱源（例えばレーザ、電子ビーム、アーク、プラズマなど）により溶融し、構造体上に所定の形状に積層する技術が用いられる。溶融された付加材は、母材及び先行して積層された部位を一部溶融し、強固な金属結合となって積層される。 When forming all or part of a structure using a metal material, an additional material (e.g., metal wire, metal powder, etc.) is melted by a heat source (e.g., laser, electron beam, arc, plasma, etc.) and placed on the structure in a predetermined position. A technique is used to stack them in the shape of . The molten additional material partially melts the base material and the previously laminated parts, and forms a strong metal bond and is laminated.
このようにして積層された金属の物性（例えば硬度、耐摩耗性、耐腐食性、靭性など）は、付加材の組成、熱源の加工条件及び熱履歴（予熱温度、熱容量、熱伝導率などで決まる加熱・冷却の状態）に影響を受ける。 The physical properties of the metals laminated in this way (e.g. hardness, wear resistance, corrosion resistance, toughness, etc.) are determined by the composition of the additive material, processing conditions of the heat source, and thermal history (preheating temperature, heat capacity, thermal conductivity, etc.). (heating and cooling conditions).
また、積層造形された部位には、要求される寸法精度に対しての誤差、熱による変形、不適切な加工条件による穴明き欠陥、割れ、表面状態の過剰な酸化など、多くの品質欠陥が発生する場合がある。 In addition, additively manufactured parts have many quality defects such as errors in the required dimensional accuracy, deformation due to heat, hole defects due to inappropriate processing conditions, cracks, and excessive oxidation of the surface condition. may occur.
ところで、積層造形部を要求範囲の寸法精度で欠陥なく加工するための付加材の組成及び加工条件には、多数の複雑なパラメータが存在する。適切なパラメータを設定するためには、従来、熟練技術者が経験に基づいて試作を行うことが必要であり、試作品の品質を評価して何度かの試行錯誤を行っていた。熟練技術者が減少する昨今、このことは製品の効率的な作製に大きな影響を及ぼす。 By the way, there are many complicated parameters for the composition of the additional material and the processing conditions for processing the additively manufactured part with the required dimensional accuracy and without defects. In the past, in order to set appropriate parameters, it was necessary for skilled engineers to make a prototype based on their experience, and the quality of the prototype was evaluated and several trials and errors were performed. With the number of skilled engineers decreasing these days, this has a major impact on the efficient production of products.
よって、本発明は、金属の付加材を用いた積層造形において、製品に要求される精度・品質レベルに応じた好適な付加材の組成及び加工条件を簡便に得ることを目的とする。 Therefore, an object of the present invention is to easily obtain a suitable additive material composition and processing conditions according to the accuracy and quality level required for a product in additive manufacturing using a metal additive material.
上記の課題を解決すべく、本発明の第１の態様は、
母材に積層される付加材と、積層体の厚さ及び許容される欠陥を含む要求性能と、前記積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む要求品質と、を示すデータを取得する取得部と、
付加材の供給量及び熱源からの入熱量を、積層体の厚さ及び欠陥レベルのそれぞれと関連付ける第１予測モデルを用いて、付加材ごとに、前記要求性能を満たし得る当該付加材の供給量及び熱源からの入熱量の領域を算出する第１推定部と、
付加材の供給量及び熱源からの入熱量に応じて積層体内部の熱履歴を算出する第２予測モデルを用いて、前記第１推定部において算出された供給量及び入熱量の領域に属する加工条件の下で、積層体内部の保持時間及び冷却速度を予測する第２推定部と、
積層体の熱履歴を、積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む加工品質に関連付ける第３予測モデルを用いて、前記加工条件の下で、付加材の成分組成ごとに加工品質の予測値を算出する第３推定部と、
前記第３推定部において算出された品質の予測値を前記要求品質と比較し、比較結果に基づいて付加材の成分組成及び熱源からの入熱量の推奨値を決定する決定部と、
を具備することを特徴とする加工条件推奨装置を提供する。
ここで、第１推定部において成分組成を仮決めし、第３推定部で推奨の成分組成を決めたら、第１推定部に戻ってやり直す、というように、第１～第３推定部における処理を繰り返し実施してもよいものとする。
In order to solve the above problems, a first aspect of the present invention is to
Indicates an additional material to be laminated to the base material, required performance including the thickness and allowable defects of the laminate, and required quality including at least one of hardness, wear resistance, and toughness of the laminate. an acquisition unit that acquires data;
Using a first prediction model that associates the supply amount of the additive material and the amount of heat input from the heat source with the thickness and defect level of the laminate, calculate the supply amount of the additive material that can satisfy the required performance for each additive material. and a first estimator that calculates an area of heat input from the heat source;
Using a second prediction model that calculates the thermal history inside the laminate according to the supply amount of the additional material and the heat input amount from the heat source, processing that belongs to the region of the supply amount and heat input amount calculated in the first estimator. a second estimator that predicts the retention time and cooling rate inside the laminate under the conditions;
Using a third prediction model that relates the thermal history of the laminate to the processing quality including at least one of the hardness, wear resistance, and toughness of the laminate, under the processing conditions, for each component composition of the additive material, a third estimation unit that calculates a predicted value of processing quality;
a determining unit that compares the predicted quality value calculated in the third estimating unit with the required quality and determines a recommended value for the component composition of the additional material and the amount of heat input from the heat source based on the comparison result;
Provided is a processing condition recommendation device characterized by comprising:
Here, the first estimation section tentatively determines the component composition, the third estimation section determines the recommended component composition, and then returns to the first estimation section and starts over, and so on. may be carried out repeatedly.
本発明の加工条件推奨装置においては、前記第１予測モデルが、付加材の供給量及び熱源からの入熱量と、当該供給量及び当該入熱量の下で作製された積層体の厚さ及び欠陥の計測結果と、を教師データとして機械学習された学習済みモデルであること、が好ましい。 In the processing condition recommendation device of the present invention, the first prediction model includes the supply amount of the additional material and the heat input amount from the heat source, and the thickness and defects of the laminate produced under the supply amount and the heat input amount. It is preferable that the model is a trained model that has been machine-learned using the measurement results of and as training data.
また、本発明の加工条件推奨装置においては、前記第２予測モデルが、付加材の供給量及び熱源からの入熱量と、当該供給量及び当該入熱量の下における熱伝導シミュレーションの結果と、を教師データとして機械学習された学習済みモデルであること、が好ましい。 Furthermore, in the processing condition recommendation device of the present invention, the second prediction model calculates the supply amount of the additional material and the heat input amount from the heat source, and the results of a heat conduction simulation under the supply amount and the heat input amount. Preferably, the model is a trained model that has been machine learned as training data.
また、本発明の加工条件推奨装置においては、前記第３予測モデルが、付加材の成分組成及び積層体の熱履歴と、当該成分組成及び当該熱履歴の下における積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む加工品質の計測結果と、を教師データとして機械学習された学習済みモデルであること、が好ましい。 Further, in the processing condition recommendation device of the present invention, the third prediction model is based on the component composition of the additional material and the thermal history of the laminate, and the hardness and wear resistance of the laminate under the component composition and the thermal history. It is preferable that the model is a trained model that has been machine-learned using, as training data, measurement results of machining quality including at least one of the following: and toughness.
本発明の第２の態様は、
コンピュータに、
母材に積層される付加材と、積層体の厚さ及び許容される欠陥を含む要求性能と、前記積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む要求品質と、を示すデータを取得する取得処理と、
付加材の供給量及び熱源からの入熱量を、積層体の厚さ及び欠陥レベルのそれぞれと関連付ける第１予測モデルを用いて、付加材ごとに、前記要求性能を満たし得る当該付加材の供給量及び熱源からの入熱量の領域を算出する第１推定処理と、
付加材の供給量及び熱源からの入熱量に応じて積層体内部の熱履歴を算出する第２予測モデルを用いて、前記第１推定処理において算出された供給量及び入熱量の領域に属する加工条件の下で、積層体内部の保持時間及び冷却速度を予測する第２推定処理と、
積層体の熱履歴を、積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む加工品質に関連付ける第３予測モデルを用いて、前記加工条件の下で、付加材の成分組成ごとに加工品質の予測値を算出する第３推定処理と、
前記第３推定処理において算出された品質の予測値を前記要求品質と比較し、比較結果に基づいて付加材の成分組成及び熱源からの入熱量の推奨値を決定する決定処理と、
を実行させることを特徴とする加工条件推奨方法を提供する。
ここで、第１推定処理において成分組成を仮決めし、第３推定処理で推奨の成分組成を決めたら、第１推定処理に戻る、というように、第１～第３推定処理を繰り返し実施してもよいものとする。
The second aspect of the invention is
to the computer,
Indicates an additional material to be laminated to the base material, required performance including the thickness and allowable defects of the laminate, and required quality including at least one of hardness, wear resistance, and toughness of the laminate. an acquisition process for acquiring data;
Using a first prediction model that associates the supply amount of the additive material and the amount of heat input from the heat source with the thickness and defect level of the laminate, calculate the supply amount of the additive material that can satisfy the required performance for each additive material. and a first estimation process for calculating a region of heat input from the heat source;
Processing that belongs to the region of the supply amount and heat input amount calculated in the first estimation process using a second prediction model that calculates the thermal history inside the laminate according to the supply amount of the additional material and the heat input amount from the heat source. a second estimation process that predicts the retention time and cooling rate inside the laminate under the conditions;
Using a third prediction model that relates the thermal history of the laminate to the processing quality including at least one of the hardness, wear resistance, and toughness of the laminate, under the processing conditions, for each component composition of the additive material, a third estimation process that calculates a predicted value of processing quality;
A determination process that compares the predicted quality value calculated in the third estimation process with the required quality and determines a recommended value for the component composition of the additional material and the amount of heat input from the heat source based on the comparison result;
A method for recommending machining conditions is provided.
Here, after tentatively determining the component composition in the first estimation process, determining the recommended component composition in the third estimation process, and returning to the first estimation process, the first to third estimation processes are repeatedly performed. It shall be permitted to do so.
本発明の第３の態様は、
コンピュータに、
母材に積層される付加材と、積層体の厚さ及び許容される欠陥を含む要求性能と、前記積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む要求品質と、を示すデータを取得する取得処理と、
付加材の供給量及び熱源からの入熱量を、積層体の厚さ及び欠陥レベルのそれぞれと関連付ける第１予測モデルを用いて、付加材ごとに、前記要求性能を満たし得る当該付加材の供給量及び熱源からの入熱量の領域を算出する第１推定処理と、
付加材の供給量及び熱源からの入熱量に応じて積層体内部の熱履歴を算出する第２予測モデルを用いて、前記第１推定処理において算出された供給量及び入熱量の領域に属する加工条件の下で、積層体内部の保持時間及び冷却速度を予測する第２推定処理と、
積層体の熱履歴を、積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む加工品質に関連付ける第３予測モデルを用いて、前記加工条件の下で、付加材の成分組成ごとに加工品質の予測値を算出する第３推定処理と、
前記第３推定処理において算出された品質の予測値を前記要求品質と比較し、比較結果に基づいて付加材の成分組成及び熱源からの入熱量の推奨値を決定する決定処理と、
を実行させるためのプログラムを提供する。
ここで、第１推定処理において成分組成を仮決めし、第３推定処理で推奨の成分組成を決めたら、第１推定処理に戻る、というように、第１～第３推定処理を繰り返し実施してもよいものとする。
The third aspect of the present invention is
to the computer,
Indicates an additional material to be laminated to the base material, required performance including the thickness and allowable defects of the laminate, and required quality including at least one of hardness, wear resistance, and toughness of the laminate. an acquisition process for acquiring data;
Using a first prediction model that associates the supply amount of the additive material and the amount of heat input from the heat source with the thickness and defect level of the laminate, calculate the supply amount of the additive material that can satisfy the required performance for each additive material. and a first estimation process for calculating a region of heat input from the heat source;
Processing that belongs to the region of the supply amount and heat input amount calculated in the first estimation process using a second prediction model that calculates the thermal history inside the laminate according to the supply amount of the additional material and the heat input amount from the heat source. a second estimation process that predicts the retention time and cooling rate inside the laminate under the conditions;
Using a third prediction model that relates the thermal history of the laminate to the processing quality including at least one of the hardness, wear resistance, and toughness of the laminate, under the processing conditions, for each component composition of the additive material, a third estimation process that calculates a predicted value of processing quality;
A determination process that compares the predicted quality value calculated in the third estimation process with the required quality and determines a recommended value for the component composition of the additional material and the amount of heat input from the heat source based on the comparison result;
Provide a program to run.
Here, after tentatively determining the component composition in the first estimation process, determining the recommended component composition in the third estimation process, and returning to the first estimation process, the first to third estimation processes are repeatedly performed. It shall be permitted to do so.
本発明の第４の態様は、
上記のいずれかに記載の加工条件推奨装置と、
母材に付加材を積層するための熱源と、
を含む金属構造体製造システムを提供する。
The fourth aspect of the present invention is
A processing condition recommendation device described in any of the above,
a heat source for laminating additional material on the base material;
The present invention provides a metal structure manufacturing system including:
本発明の第５の態様は、
上記に記載の加工条件推奨方法に基づいて付加材の成分組成及び熱源からの入熱量の推奨値を取得する工程と、
前記推奨値に基づいて母材に付加材を積層する工程と、
を含む金属構造体製造方法を提供する。
The fifth aspect of the present invention is
a step of obtaining recommended values for the component composition of the additive material and the amount of heat input from the heat source based on the processing condition recommendation method described above;
Laminating an additional material on the base material based on the recommended value;
A method for manufacturing a metal structure is provided.
上記本発明により、金属の付加材を用いた積層造形において、製品に要求される精度・品質レベルに応じた好適な付加材の組成及び加工条件を簡便に得ることができる。また、得られた付加材の組成及び加工条件に基づいて効率的に金属構造体を製造することができる。 According to the present invention, in additive manufacturing using a metal additive material, it is possible to easily obtain the composition and processing conditions of the additive material suitable for the precision and quality level required for the product. Moreover, a metal structure can be efficiently manufactured based on the composition and processing conditions of the obtained additional material.
以下、図面を参照しながら、本発明の代表的な実施形態に係る加工条件推奨装置、加工条件推奨方法、プログラム、金属構造体製造システム及び金属構造体の製造方法を説明する。また、本発明は図示されるものに限られるものではない。また、図面は本発明を概念的に説明するためのものであるから、理解容易のために比や数を誇張又は簡略化して表している場合もある。以下の説明では、同一又は相当部分には同一符号を付し、重複する説明は省略することもある。 DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a processing condition recommendation device, a processing condition recommendation method, a program, a metal structure manufacturing system, and a metal structure manufacturing method according to typical embodiments of the present invention will be described with reference to the drawings. Moreover, the present invention is not limited to what is illustrated. Further, since the drawings are for conceptually explaining the present invention, ratios and numbers may be exaggerated or simplified for ease of understanding. In the following description, the same or corresponding parts will be denoted by the same reference numerals, and redundant descriptions may be omitted.
本実施形態では、積層加工の一例としてレーザによる肉盛加工を挙げるが、本発明はこれに限られない。ここでは、レーザによる肉盛加工の一例として、２種類の粉末状の付加材（例えば熱間工具鋼と高速度工具鋼）を混合した混合粉末を母材に積層することを挙げるが、付加材はワイヤ等の他の形態でもよい。 In this embodiment, build-up processing using a laser is used as an example of lamination processing, but the present invention is not limited to this. Here, as an example of laser overlay processing, we will cite laminating a mixed powder of two types of powdered additive materials (for example, hot work tool steel and high speed tool steel) on a base material. may be in other forms such as a wire.
１．加工条件推奨の枠組み
図１を参照して、本実施形態における加工条件の推奨又は提示の全体的な枠組みを説明する。
1. Framework for Recommending Processing Conditions The overall framework for recommending or presenting processing conditions in this embodiment will be described with reference to FIG.
本実施形態において、加工条件推奨装置１０及び加工条件推奨方法は、次の機能を実行することで、要求される積層体の厚さ（肉盛り高さ）及び品質（硬度等）などの入力から、出力として、付加材の成分組成（混合比又は配合比）及び加工条件等の推奨値を得ることができる。
（１） 条件仮決め機能： 予め選定した付加材ごとに、要求性能（積層厚さ、欠陥レベルなど）に見合う付加材の供給量及び熱源からの入熱量の領域を算出し、基本条件を仮決めする。
（２） 熱履歴推定機能： 上記（１）で仮決めした基本条件の下で付加材の組成と加工条件から積層体の熱履歴を予測する。
（３） 物性推定機能： 上記（２）で算出した熱履歴から、付加材の成分組成に対応する加工品質（硬度、耐摩耗性、靭性など）を予測する。
必要に応じて、上記機能のうち少なくとも１つを複数回利用して処理することができ、これにより更に高い精度の推奨条件を得ることができる。例えば、条件仮決め機能において成分組成を仮決めし、物性推定機能で推奨の成分組成を決めたら、第１推定処理に戻る、というように、条件仮決め機能から物性推定機能までの各機能を繰り返し実施するとよい。
In this embodiment, the machining condition recommendation device 10 and the machining condition recommendation method perform the following functions to obtain input information such as the required thickness (height of build-up) and quality (hardness, etc.) of the laminate. , Recommended values for the component composition (mixing ratio or compounding ratio) of the additive material, processing conditions, etc. can be obtained as output.
(1) Temporary condition determination function: For each additional material selected in advance, calculate the range of the amount of additional material supplied and the amount of heat input from the heat source that meet the required performance (laminated thickness, defect level, etc.), and tentatively determine the basic conditions. Decide.
(2) Thermal history estimation function: The thermal history of the laminate is predicted from the composition of the additional material and processing conditions under the basic conditions tentatively determined in (1) above.
(3) Physical property estimation function: From the thermal history calculated in (2) above, the processing quality (hardness, wear resistance, toughness, etc.) corresponding to the component composition of the additive material is predicted.
If necessary, at least one of the above functions can be used multiple times for processing, thereby making it possible to obtain recommended conditions with even higher accuracy. For example, after tentatively determining the component composition using the condition tentative determination function and determining the recommended composition using the physical property estimation function, each function from the condition tentative determination function to the physical property estimation function can be used, such as returning to the first estimation process. It is recommended to perform this repeatedly.
これら機能は、機械学習により学習済みの予測モデルを用いて実行される。機械学習としては、ニューラルネットワークやランダムフォレストなどのアルゴリズムを用いることができる。上記（１）（３）については実験データを、上記（２）については熱伝導シミュレーションの結果を、それぞれ教師データとすることができる。教師データの要素としては、付加材の組成、比重、比熱、熱伝導度、母材の材質、熱容量、熱源の出力、エネルギ密度、加工速度、積層パスの順序が挙げられる。
以下、上記機能を実行する加工条件推奨装置を詳細に説明する。
These functions are executed using predictive models trained by machine learning. As machine learning, algorithms such as neural networks and random forests can be used. For the above (1) and (3), experimental data can be used as the teacher data, and for the above (2), the results of the heat conduction simulation can be used as the teacher data. The elements of the training data include the composition of the additional material, specific gravity, specific heat, thermal conductivity, material of the base material, heat capacity, output of the heat source, energy density, processing speed, and order of lamination passes.
The processing condition recommendation device that performs the above functions will be described in detail below.
２．加工条件推奨装置について
まず物理構成を説明すると、加工条件推奨装置１０は、ＣＰＵ、ＲＡＭ、ＲＯＭ、入力装置、出力装置及び必要な通信インターフェイスを含むコンピュータとして構成される。加工条件推奨装置１０は、１台のコンピュータでもよいし、複数台のコンピュータで構成されてもよい。また、加工条件推奨装置１０は、外部に接続あるいは依存せずに単独で機能してもよいし、ネットワークを通じて外部と通信可能に接続されてもよい。更に、加工条件推奨装置１０は、クラウドコンピューティング環境下に置かれてもよい。
2. Regarding the machining condition recommendation device First, the physical configuration will be described. The machining condition recommendation device 10 is configured as a computer including a CPU, a RAM, a ROM, an input device, an output device, and a necessary communication interface. The processing condition recommendation device 10 may be composed of one computer or a plurality of computers. Further, the processing condition recommendation device 10 may function independently without being connected to or dependent on the outside, or may be communicably connected to the outside through a network. Furthermore, the processing condition recommendation device 10 may be placed in a cloud computing environment.
各構成部品については、ＣＰＵは、各種プログラム及び各種データをＲＡＭに読み出したうえで、各種演算を実行し、演算結果をＲＯＭに記憶することで、後述する各機能を実行する。ＲＯＭは、各種プログラム及び各種データを記憶するとともに、演算結果を記憶する。入力装置は、キーボード、マウス、タッチパネルなどであり、出力装置は、ディスプレイ、プリンタなどである。 Regarding each component, the CPU reads various programs and data into the RAM, executes various calculations, and stores the calculation results in the ROM, thereby executing each function described below. The ROM stores various programs and data, as well as calculation results. The input device is a keyboard, mouse, touch panel, etc., and the output device is a display, printer, etc.
次いで機能構成を説明すると、図７に示すように、加工条件推奨装置１０は、取得部１１、条件仮決め部１２、熱履歴推定部１３、品質推定部１４及び決定部１５の各機能部を含んで構成される。 Next, to explain the functional configuration, as shown in FIG. It consists of:
取得部１１は、ユーザ操作等に応じて各種データを取得する。具体的には、取得部１１は、母材情報としての材質及び熱容量、付加材情報としての付加材名及び組成、要求性能としての積層体厚さ、欠陥レベル（例えばブローホールの数・大きさ、割れの有無）、要求品質としての硬度、耐摩耗性及び靭性、加工パス情報、加工機スペックとしての最大出力及びスポット面積、などを示すデータを取得できる。かかるデータは、ユーザ入力によるものでもよいし、加工機等の外部装置から出力されたものでもよい。 The acquisition unit 11 acquires various data according to user operations and the like. Specifically, the acquisition unit 11 acquires material and heat capacity as base material information, additional material name and composition as additional material information, laminate thickness as required performance, and defect level (for example, number and size of blowholes). , presence or absence of cracks), hardness, wear resistance and toughness as required quality, machining path information, maximum output and spot area as machining machine specifications, etc. can be obtained. Such data may be input by a user or may be output from an external device such as a processing machine.
条件仮決め部１２は、第１推定部に相当する機能部であり、予測モデル（第１予測モデル）を用いて上述した条件仮決め機能を実行する。この予測モデルは、加工条件推奨装置１０のＲＯＭに記憶されていてもよいし、例えばＡＰＩとしてネットワーク経由で提供されてもよい。予測モデルは、例えば、次の手順で生成される。 The condition provisional determination unit 12 is a functional unit equivalent to the first estimation unit, and executes the above-described condition provisional determination function using a prediction model (first prediction model). This prediction model may be stored in the ROM of the processing condition recommendation device 10, or may be provided as an API via a network, for example. The prediction model is generated, for example, by the following procedure.
まず、付加材の供給量と熱源からの入熱量との種々の組合せについてレーザ肉盛り加工を実施し、積層厚さ及び性能・品質（例えばブローホールの大きさ・数、耐摩耗試験の数値、割れの有無など）を含む実験データを得る。ここで、熱源からの入熱量は、例えば、温度センサ又は光センサによる溶融した付加材の計測結果、熱源へ投入した電力量などに基づいて算出することができる。また、性能・品質は、例えば断面観察及びＸ線透過試験を通じて求めることもできる。なお、機械学習のために、性能は、ブローホールの数に、その大きさで重みをつけた欠陥スコアで表されてもよい（図３参照）。 First, laser build-up processing is performed for various combinations of the amount of additional material supplied and the amount of heat input from the heat source, and the laminated thickness, performance, and quality (e.g., size and number of blowholes, wear resistance test values, Obtain experimental data including the presence or absence of cracks, etc. Here, the amount of heat input from the heat source can be calculated based on, for example, the measurement result of the melted additional material by a temperature sensor or an optical sensor, the amount of electric power input to the heat source, and the like. Furthermore, performance and quality can also be determined, for example, through cross-sectional observation and X-ray transmission tests. Note that for the purpose of machine learning, performance may be expressed as a defect score obtained by weighting the number of blowholes by their size (see FIG. 3).
この実験データを教師データとして教師あり学習を行い、学習済みモデルを得る。例えば、付加材の供給量及びレーザ加工条件によって求められる入熱量を説明変数（ｘ）とし、積層厚さ及び数値化された性能を目的変数（ｙ）とする。機械学習のアルゴリズムとしては、図２に例示するニューラルネットワークのほか、ランダムフォレストが挙げられる。 Supervised learning is performed using this experimental data as training data to obtain a trained model. For example, the amount of heat input determined by the supply amount of the additional material and the laser processing conditions is used as an explanatory variable (x), and the laminated thickness and the numerically expressed performance are used as objective variables (y). Examples of machine learning algorithms include the neural network illustrated in FIG. 2, as well as random forest.
機械学習のアルゴリズムとしてニューラルネットワークが用いられる場合、入力層のユニットｘｉ及び出力層のユニットｙｉの個数は適宜設定されてよく、中間層（隠れ層）の層数及びその各ユニットｚｉの個数もまた適宜設定されてよい。 When a neural network is used as a machine learning algorithm, the number of input layer units x i and output layer units y i may be set as appropriate, and the number of intermediate layers (hidden layers) and each unit z i may be set as appropriate. The number may also be set appropriately.
このとき、第ｌ層に属するｊ番目のユニットの入力ｕｊ
（ｌ）は次式で表される。
このユニットの出力ｚｊ
（ｌ）は、入力ｕｊ
（ｌ）にユニット固有のバイアスｂｊ
（ｌ）を加え、更に活性化関数ｆ（ｌ）を作用させて
The output zz (l) of this unit is obtained by adding the unit-specific bias b j (l) to the input u j ( l ) and further applying the activation function f (l).
そして、例えば次式で表される誤差関数（損失関数）Ｅを最小化するようにパラメータｗ＊を決定する。
このようにして機械学習させて得た学習済みモデルを用いて、所定範囲における付加材の供給量及び熱源からの入熱量に対して、積層厚さ及び性能レベルの推定値（又は予測値）を算出する。ここで、所定範囲は、例えば利用可能な加工設備のスペック（最大レーザ出力、付加材の供給量など）に応じて決定されてよい。また、算出された推定値は、例えば図３に示す等高線図のように可視化して出力されてもよい。 Using the trained model obtained through machine learning in this way, estimated values (or predicted values) of the lamination thickness and performance level are calculated for the amount of additional material supplied and the amount of heat input from the heat source within a predetermined range. calculate. Here, the predetermined range may be determined, for example, according to the specifications of available processing equipment (maximum laser output, supply amount of additional material, etc.). Further, the calculated estimated value may be visualized and output as, for example, a contour map shown in FIG. 3.
算出された積層厚さ及び品質レベルの推定値のうち、予め設定された積層厚さ及び性能レベルを満足する推定値に対応する供給量及び入熱量の範囲を、仮の又は暫定的な加工条件として選択する。積層厚さ及び品質レベルが可視化される場合には、例えば図３のように、両方の等高線図を重ね合わせ、求められる積層厚さ及び品質レベルに該当する供給量及び入熱量の範囲（例えば図３の領域Ｓ）を暫定的な推奨条件として仮決めする。 Among the estimated values of the calculated lamination thickness and quality level, the range of supply amount and heat input corresponding to the estimated value that satisfies the preset lamination thickness and performance level is set under temporary or provisional processing conditions. Select as. When the lamination thickness and quality level are visualized, for example, as shown in Fig. 3, both contour maps are superimposed and the range of supply amount and heat input corresponding to the required lamination thickness and quality level (for example, the figure Region S) of 3 is tentatively determined as a provisional recommended condition.
なお、第１予測モデルは、良好領域の推定のために、ガウス過程を用いたベイズ推論を更に用いることができる。より具体的には、多数のガウス分布を線形回帰し、観測データによる分布の更新後もガウス分布になっているガウス過程に、測定したデータが従っていると仮定し、観測データによりモデルの確率分布を更新することによって、期待値と分散の分布を得られる手法である。この手法を用いることで、比較的少ない（実用的には数百点程度）データ点数でも、例えばブローホールの発生確率を予測して、良好な加工条件範囲を決定できる。 Note that the first prediction model can further use Bayesian inference using a Gaussian process to estimate the good region. More specifically, we linearly regress a large number of Gaussian distributions, assume that the measured data follows a Gaussian process that remains Gaussian even after updating the distribution using observed data, and calculate the probability distribution of the model using the observed data. This is a method that allows you to obtain the distribution of expected value and variance by updating . By using this method, even with a relatively small number of data points (in practice, about several hundred points), it is possible to predict the probability of blowhole occurrence, for example, and determine a favorable processing condition range.
又、図４のように、目的変数毎に機械学習させて得た学習済みモデルを用いて、探索アルゴリズムによって要求性能を満たし得る加工条件や粉末組成を抽出してもよい。具体的には、各目的変数の要求性能値と学習済みモデルの予測値の誤差を標準化し、標準化した値それぞれに重みＷＮを掛けて、それらの和を評価指標ＥＩとし、評価指標が最小となるような説明変数（粉末供給量、レーザ出力、加工速度など）を抽出する。探索アルゴリズム（例えばベイズ最適化）を用いて、次の条件探索、モデルから予測値を出力、評価指標の算出を繰り返し実施する。
熱履歴推定部１３の説明に移ると、熱履歴推定部１３は、第２推定部に相当する機能部であり、予測モデル（第２予測モデル）を用いて、上述した熱履歴推定機能を実行する。すなわち、熱履歴推定部１３は、条件仮決め部１２において仮決めした加工条件と、加工対象たる製品の熱容量及び粉末の熱的な性質（熱伝導率など）とに基づき、各加工パスの中での焼き入れ・焼き戻し特性を評価し、積層体内部の変態点位以上での保持時間や冷却速度を予測する。予測モデルは、加工条件推奨装置１０のＲＯＭに記憶されていてもよいし、例えばＡＰＩとしてネットワーク経由で提供されてもよい。この予測モデルは、例えば、次の手順で生成される。 Moving on to the explanation of the thermal history estimating unit 13, the thermal history estimating unit 13 is a functional unit corresponding to a second estimating unit, and executes the above-mentioned thermal history estimating function using a prediction model (second prediction model). do. In other words, the thermal history estimating unit 13 calculates the processing conditions during each processing pass based on the processing conditions temporarily determined by the condition temporary determination unit 12, the heat capacity of the product to be processed, and the thermal properties of the powder (thermal conductivity, etc.). Evaluate the quenching and tempering characteristics of the laminate and predict the retention time and cooling rate above the transformation point inside the laminate. The prediction model may be stored in the ROM of the processing condition recommendation device 10, or may be provided as an API via a network, for example. This prediction model is generated, for example, by the following procedure.
具体的には、付加材の種類によっては、後続パスの入熱による焼き入れや焼き戻しの効果により、積層体の物性（硬度等）が変化する。この入熱の影響（熱履歴）は後続パスとの境界面からの距離、予熱温度（直下層温度）、レーザ加工条件、製品の熱容量、レーザ出力、加工速度など多くのパラメータにより変化する。 Specifically, depending on the type of additional material, the physical properties (hardness, etc.) of the laminate change due to the effects of quenching and tempering due to heat input in subsequent passes. The influence of this heat input (thermal history) changes depending on many parameters such as the distance from the interface with the subsequent pass, preheating temperature (temperature directly below), laser processing conditions, heat capacity of the product, laser output, and processing speed.
そこで、本実施形態では、実現可能な範囲（図３に領域Ｓで示される条件範囲よりも広範囲）の加工条件に対応する保持時間及び冷却速度を予め熱伝導シミュレータで計算しておく。例えば、図５では、パスＷ１～Ｗ１２が３層に形成された積層体において、パスＷ１１，Ｗ１２のパス境界から一定距離だけ離れたＡ点及びＢ点の間（Ｗ１１とＷ１２の境界から－０．４、０、０．５、１．０、１．５、２．０、２．５、３．０だけ離れた地点）における温度変化、パス境界からの距離と保持時間及び温度低下時間（冷却速度）との関係が例示的に示されている。 Therefore, in this embodiment, the holding time and cooling rate corresponding to processing conditions within a feasible range (wider than the range of conditions shown by region S in FIG. 3) are calculated in advance using a heat conduction simulator. For example, in FIG. 5, in a stacked body in which paths W1 to W12 are formed in three layers, between points A and B that are a certain distance away from the path boundaries of paths W11 and W12 (-0 from the boundary between W11 and W12) .4, 0, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0 points away), distance from the path boundary, holding time, and temperature drop time ( (cooling rate) is exemplarily shown.
そして、加工条件を説明変数（ｘ）とし、保持時間及び冷却速度を目的変数（ｙ）とした教師あり学習を行い、学習済みモデルを得る。機械学習としては、上述したニューラルネットワークやランダムフォレストなどのアルゴリズムを用いることができる。 Then, supervised learning is performed with the processing conditions as explanatory variables (x) and the holding time and cooling rate as objective variables (y) to obtain a learned model. As machine learning, algorithms such as the above-mentioned neural network and random forest can be used.
機械学習により得られた学習済みモデルを用いて、先に算出した暫定的な加工条件から保持時間及び冷却速度を予測する。機械学習では学習に時間が掛かるものの、予測は非常に短時間で終了するので、オンデマンドの熱履歴推定が可能となる。この点、従来技術では、これらの条件をモデル化して熱伝導解析を行うが、加工条件が変わるごとに、モデルを定義してメッシュを切るなどのプリプロセスを行って、解析を実行し結果をもとに変態点以上の保持時間や冷却速度を求める必要があるため、オンデマンドで、直ちに結果を求めることはできない。 Using the trained model obtained through machine learning, the holding time and cooling rate are predicted from the provisional processing conditions calculated previously. Although machine learning takes time to learn, predictions can be completed in a very short time, making on-demand thermal history estimation possible. In this regard, in conventional technology, these conditions are modeled to perform heat conduction analysis, but each time the machining conditions change, a pre-process such as defining a model and cutting a mesh is performed, and the analysis is executed and the results obtained. Because it is necessary to first determine the holding time and cooling rate above the transformation point, it is not possible to immediately obtain results on demand.
次いで、品質推定部１４は、第３推定部に相当する機能部であり、予測モデル（第３予測モデル）を用いて、上述した品質推定機能を実行する。すなわち、品質推定部１４は、予め選定した付加材を任意の成分組成で、先の手順で得られた加工条件及び熱履歴から加工品質（硬度等の物性）を予測する。予測モデルは、加工条件推奨装置１０のＲＯＭに記憶されていてもよいし、例えばＡＰＩとしてネットワーク経由で提供されてもよい。この予測モデルは、例えば、次の手順で生成される。 Next, the quality estimation unit 14 is a functional unit corresponding to a third estimation unit, and executes the quality estimation function described above using the prediction model (third prediction model). That is, the quality estimation unit 14 predicts the processing quality (physical properties such as hardness) of the preselected additional material with any component composition from the processing conditions and thermal history obtained in the previous procedure. The prediction model may be stored in the ROM of the processing condition recommendation device 10, or may be provided as an API via a network, for example. This prediction model is generated, for example, by the following procedure.
金属の積層造形に適した付加材は、合金工具鋼（例えば高速度工具鋼や熱間工具鋼）、ステンレス系、ステライト系など数種類に限られる。さらに添加剤としてタングステンカーバイドなどが用いられる。積層体の品質（硬度等）は付加材の組成によって大きく影響を受け、割れやブローホール欠陥の発生にも影響を受けるので、要求性能を単一の付加材で満たせる条件は少ない。 Additive materials suitable for additive manufacturing of metals are limited to a few types, such as alloy tool steels (for example, high-speed tool steels and hot work tool steels), stainless steels, and stellite-based materials. Furthermore, tungsten carbide or the like is used as an additive. The quality of the laminate (hardness, etc.) is greatly affected by the composition of the additive material, and is also affected by the occurrence of cracks and blowhole defects, so there are few conditions in which the required performance can be met with a single additive material.
そこで、本実施形態では、付加材の様々な成分組成の下で積層体の様々な部位の品質又は物性を測定し、実験データを得る。併せて、熱伝導シミュレータで熱履歴を求めておく。そして、付加材の成分組成及び各部位の熱履歴を説明変数（ｘ）とし、各部位の品質を目的変数（ｙ）とした教師あり学習を行い、学習済みモデルを得る。機械学習としては、上述したニューラルネットワークやランダムフォレストなどのアルゴリズムを用いることができる。 Therefore, in this embodiment, the quality or physical properties of various parts of the laminate are measured under various component compositions of the additive material to obtain experimental data. At the same time, determine the thermal history using a heat conduction simulator. Supervised learning is then performed using the component composition of the additional material and the thermal history of each part as explanatory variables (x) and the quality of each part as an objective variable (y) to obtain a learned model. As machine learning, algorithms such as the above-mentioned neural network and random forest can be used.
機械学習により得られた学習済みモデルを用いて、付加材の様々な成分組成に対応する積層体の品質を予測する。例えば、図６は、２種類の粉末状の付加材の混合率（熱間工具鋼に対する高速度工具鋼の割合）を０％から１００％まで変化させ、そのときの積層体の部位の硬度（パス境界からの距離に応じた硬度）を予測したものである。硬度を発現するメカニズムが異なるため、２種類の粉末で得られる硬度を混合率で単純に案分したものにはならないが、本実施形態では、機械学習システムで各基本粉末と数種類の混合粉末の実験結果を学習することにより、混合粉末の硬度を高い精度で予測することができる。 Using trained models obtained through machine learning, we predict the quality of laminates corresponding to various component compositions of additive materials. For example, FIG. 6 shows that the mixing ratio of two types of powdered additives (the ratio of high-speed tool steel to hot-work tool steel) is varied from 0% to 100%, and the hardness of the part of the laminate ( This is a prediction of the hardness depending on the distance from the path boundary. Because the mechanisms that produce hardness are different, the hardness obtained with two types of powder cannot be simply divided by the mixing ratio, but in this embodiment, a machine learning system is used to calculate the hardness of each basic powder and several types of mixed powder. By learning the experimental results, the hardness of mixed powder can be predicted with high accuracy.
決定部１５は、品質推定部１４の算出結果を要求品質と比較し、比較結果に基づいて付加材の成分組成及び熱源条件等の推奨値を決定する。例えば、決定部１５は、硬度の予測値のうち、要求硬度に近い予測値に対応する粉末の組成を推奨する。出力される推奨値としては、例えば、付加材情報としての材料名、成分組成及び供給量、加工条件としての予熱温度（直下層温度）、レーザ出力、加工速度及びスポット面積が挙げられるが、これらに限られない。 The determining unit 15 compares the calculation result of the quality estimating unit 14 with the required quality, and determines recommended values for the component composition of the additional material, heat source conditions, etc. based on the comparison result. For example, the determining unit 15 recommends a powder composition corresponding to a predicted value close to the required hardness among the predicted values of hardness. Recommended values to be output include, for example, the material name, component composition, and supply amount as additional material information, preheating temperature (temperature directly below the layer), laser output, processing speed, and spot area as processing conditions. Not limited to.
決定部１５はまた、予想される性能として積層厚さ、硬度等の物性及び欠陥レベルを出力してもよい。そして、決定部１５は、推奨値等をディスプレイ等の出力装置に表示させたり、通信インターフェイスを介して外部装置（例えば、後述する制御装置２１）に出力したりする。 The determining unit 15 may also output physical properties such as layer thickness and hardness, and defect level as expected performance. Then, the determining unit 15 displays the recommended value and the like on an output device such as a display, or outputs it to an external device (for example, the control device 21 described later) via a communication interface.
３．加工条件推奨方法
したがって、加工条件推奨装置１０では、図８に示す次の処理が実行されて加工条件が推奨される。
ステップＳ１において、取得部１１が、母材に積層される付加材と、積層体の厚さ及び許容される欠陥を含む要求性能と、積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む要求品質と、を示すデータを取得する。
ステップＳ２において、条件仮決め部１２が、付加材の供給量及び熱源からの入熱量を、積層体の厚さ及び欠陥レベルのそれぞれと関連付ける第１予測モデルを用いて、付加材ごとに、要求性能を満たし得る当該付加材の供給量及び熱源からの入熱量の領域を算出する。第１予測モデルは、更に、ガウス過程を用いたベイズ推論を用いてもよい。
ステップＳ３において、熱履歴推定部１３が、付加材の供給量及び熱源からの入熱量に応じて積層体内部の熱履歴を算出する第２予測モデルを用いて、ステップＳ２において算出された供給量及び入熱量の領域に属する加工条件の下で、積層体内部の保持時間及び冷却速度を予測する。
ステップＳ４において、品質推定部１４が、積層体の熱履歴を、積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む加工品質に関連付ける第３予測モデルを用いて、加工条件の下で、付加材の成分組成ごとに加工品質の予測値を算出する。
ステップＳ５において、決定部１５が、ステップＳ４において算出された品質の予測値を要求品質と比較し、比較結果に基づいて付加材の成分組成及び熱源からの入熱量の推奨値を決定する。
3. Machining Condition Recommendation Method Therefore, the machining condition recommendation device 10 executes the following process shown in FIG. 8 to recommend machining conditions.
In step S1, the acquisition unit 11 acquires at least one of the additional material to be laminated on the base material, the required performance including the thickness and allowable defects of the laminate, and the hardness, wear resistance, and toughness of the laminate. Obtain data indicating the required quality, including the required quality.
In step S2, the condition provisional determination unit 12 uses a first prediction model that associates the supply amount of the additional material and the amount of heat input from the heat source with the thickness and defect level of the laminate to determine the requirements for each additional material. Calculate the range of the supply amount of the additional material and the amount of heat input from the heat source that can satisfy the performance. The first prediction model may further use Bayesian inference using a Gaussian process.
In step S3, the thermal history estimating unit 13 calculates the supply amount calculated in step S2 using a second prediction model that calculates the thermal history inside the laminate according to the supply amount of the additional material and the amount of heat input from the heat source. The retention time and cooling rate inside the laminate are predicted under processing conditions that fall within the range of heat input and heat input.
In step S4, the quality estimation unit 14 uses a third prediction model that associates the thermal history of the laminate with the processing quality including at least one of the hardness, wear resistance, and toughness of the laminate under the processing conditions. Then, a predicted value of processing quality is calculated for each component composition of the additive material.
In step S5, the determining unit 15 compares the predicted quality value calculated in step S4 with the required quality, and determines recommended values for the component composition of the additional material and the amount of heat input from the heat source based on the comparison result.
このように、機械学習によって得られる汎化性能により、実験を行っていない加工条件や付加材組成に対しても、積層厚さや加工欠陥の程度、積層物の硬度、耐摩耗性、靭性等を予測することができ、加工条件及び付加材組成の推奨値を得ることができる。これにより非熟練技術者であっても、試作品を製作して評価する試行回数を大幅に減少させることができる。 In this way, with the generalization performance obtained through machine learning, it is possible to determine the laminated thickness, degree of machining defects, hardness, wear resistance, toughness, etc. of the laminated material, even for machining conditions and additive material compositions that have not been tested. It is possible to make predictions and obtain recommended values for processing conditions and additive material composition. As a result, even unskilled engineers can significantly reduce the number of attempts to manufacture and evaluate prototypes.
つまり、金属積層加工のように相互に影響を及ぼす多くのパラメータが存在する場合、一つの機械学習モデルでは精度の高い解が得られないのに対して、本実施形態では、パラメータの相互依存関係を考慮し、３つの学習ブロックに分割して、学習結果を遷移させていくことにより、精度の高い予測結果に収束させることができる。 In other words, when there are many parameters that influence each other, such as in metal lamination processing, a single machine learning model cannot provide a highly accurate solution.In contrast, in this embodiment, the mutual dependence of parameters By taking this into account, dividing the learning blocks into three learning blocks, and transitioning the learning results, it is possible to converge to a highly accurate prediction result.
４．金属構造体の製造装置及び製造方法
加工条件推奨装置１０及び加工条件推奨方法の応用例として、この装置及び方法を組み込んだ金属構造体製造装置及び方法を説明する。
4. Metal Structure Manufacturing Apparatus and Manufacturing Method As an application example of the processing condition recommendation apparatus 10 and the processing condition recommendation method, a metal structure manufacturing apparatus and method incorporating this apparatus and method will be described.
図９に示すように、金属構造体製造システム２０は、加工条件推奨装置１０、制御装置２１及び熱源２３を含み、熱源２３で付加材Ｐを溶融して母材３１に積層体３３を積層させる。母材３１としては、例えば鉄鋼圧延用ロール、各種工具、金型類が挙げられ、金属構造体製造システム２０により、例えば、高速度工具鋼を溶接した鉄鋼圧延用ロール等の金属構造体が得られる。 As shown in FIG. 9, the metal structure manufacturing system 20 includes a processing condition recommendation device 10, a control device 21, and a heat source 23, and the heat source 23 melts an additional material P to laminate a laminate 33 on a base material 31. . Examples of the base material 31 include steel rolling rolls, various tools, and molds, and the metal structure manufacturing system 20 produces metal structures such as steel rolling rolls welded with high-speed tool steel. It will be done.
熱源２３は例えばレーザ光Ｌの光源である。制御装置２１は、加工条件推奨装置１０から出力された推奨値に沿って熱源２３を制御する。なお、加工条件推奨装置１０と制御装置２１とは一つの装置でもよいし、内部又は外部のネットワークを介して通信可能に接続されていてもよい。 The heat source 23 is, for example, a light source of laser light L. The control device 21 controls the heat source 23 in accordance with the recommended value output from the processing condition recommendation device 10. Note that the processing condition recommendation device 10 and the control device 21 may be one device, or may be communicably connected via an internal or external network.
かかる金属構造体製造システム２０では、ユーザ操作に応じて加工条件推奨装置１０に付加材の種類等が入力されると、加工条件推奨装置１０は付加材の成分組成等の加工条件を制御装置２１に出力する。制御装置２１は、加工条件を受信すると、その加工条件に基づいて熱源２３を制御し、これにより金属構造体が製造される。 In such a metal structure manufacturing system 20, when the type of additional material, etc. is input to the processing condition recommendation device 10 in response to a user operation, the processing condition recommendation device 10 sets processing conditions such as the component composition of the additional material to the control device 21. Output to. When the control device 21 receives the processing conditions, it controls the heat source 23 based on the processing conditions, thereby manufacturing the metal structure.
したがって、ユーザは、加工条件の設定のために試行錯誤をする必要がないか、そうでなくとも試行錯誤を大幅に減少させることができる。また、金属構造体の製造効率が大幅に向上する。 Therefore, the user does not need to perform trial and error to set processing conditions, or even if not, the trial and error can be significantly reduced. Moreover, the manufacturing efficiency of the metal structure is greatly improved.
以上、本発明の代表的な実施形態について図面を参照しつつ説明してきたが、本発明は、これらの実施形態に限定されるものではなく、特許請求の範囲の記載の精神及び教示を逸脱しない範囲でその他の改良例や変形例が存在する。そして、かかる改良例や変形例は全て本発明の技術的範囲に含まれる。 Although typical embodiments of the present invention have been described above with reference to the drawings, the present invention is not limited to these embodiments and does not depart from the spirit and teachings of the claims. There are other improvements and variations within the scope. All such improvements and modifications are included within the technical scope of the present invention.
１０ 加工条件推奨装置
１１ 取得部
１２ 条件仮決め部
１３ 熱履歴推定部
１４ 品質推定部
１５ 決定部
２０ 金属構造体製造システム
２３ 熱源
３１ 母材
３３ 積層体
Ｌ レーザ光
Ｐ 付加材
10 Processing condition recommendation device 11 Acquisition unit 12 Condition temporary determination unit 13 Thermal history estimation unit 14 Quality estimation unit 15 Determination unit 20 Metal structure manufacturing system 23 Heat source 31 Base material 33 Laminated body L Laser beam P Additional material
Claims (8)
付加材の供給量及び熱源からの入熱量を、積層体の厚さ及び欠陥レベルのそれぞれと関連付ける第１予測モデルを用いて、付加材ごとに、前記要求性能を満たし得る当該付加材の供給量及び熱源からの入熱量の領域を算出する第１推定部と、
付加材の供給量及び熱源からの入熱量に応じて積層体内部の熱履歴を算出する第２予測モデルを用いて、前記第１推定部において算出された供給量及び入熱量の領域に属する加工条件の下で、積層体内部の保持時間及び冷却速度を予測する第２推定部と、
積層体の熱履歴を、積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む加工品質に関連付ける第３予測モデルを用いて、前記加工条件の下で、付加材の成分組成ごとに加工品質の予測値を算出する第３推定部と、
前記第３推定部において算出された品質の予測値を前記要求品質と比較し、比較結果に基づいて付加材の成分組成及び熱源からの入熱量の推奨値を決定する決定部と、
を具備することを特徴とする加工条件推奨装置。 Indicates an additional material to be laminated to the base material, required performance including the thickness and allowable defects of the laminate, and required quality including at least one of hardness, wear resistance, and toughness of the laminate. an acquisition unit that acquires data;
Using a first prediction model that associates the supply amount of the additive material and the amount of heat input from the heat source with the thickness and defect level of the laminate, calculate the supply amount of the additive material that can satisfy the required performance for each additive material. and a first estimator that calculates an area of heat input from the heat source;
Using a second prediction model that calculates the thermal history inside the laminate according to the supply amount of the additional material and the heat input amount from the heat source, processing that belongs to the region of the supply amount and heat input amount calculated in the first estimator. a second estimator that predicts the retention time and cooling rate inside the laminate under the conditions;
Using a third prediction model that relates the thermal history of the laminate to the processing quality including at least one of the hardness, wear resistance, and toughness of the laminate, under the processing conditions, for each component composition of the additive material, a third estimation unit that calculates a predicted value of processing quality;
a determining unit that compares the predicted quality value calculated in the third estimating unit with the required quality and determines a recommended value for the component composition of the additional material and the amount of heat input from the heat source based on the comparison result;
A processing condition recommendation device characterized by comprising:
を特徴とする請求項１に記載の加工条件推奨装置。 The first prediction model uses the amount of supply of the additional material and the amount of heat input from the heat source, and the measurement results of the thickness and defects of the laminate produced under the amount of supply and the amount of heat input as training data. be a trained model,
The processing condition recommendation device according to claim 1, characterized in that:
を特徴とする請求項１又は２に記載の加工条件推奨装置。 The second prediction model is a trained model that has been machine learned using the supply amount of the additional material, the amount of heat input from the heat source, and the results of a heat conduction simulation under the supply amount and the heat input amount as training data. thing,
The processing condition recommendation device according to claim 1 or 2, characterized in that:
を特徴とする請求項１～３のいずれかに記載の加工条件推奨装置。 The third prediction model is based on the component composition of the additional material and the thermal history of the laminate, and the processing quality including at least one of the hardness, wear resistance, and toughness of the laminate under the component composition and the thermal history. It is a trained model that has been machine learned using the measurement results and training data,
The processing condition recommendation device according to any one of claims 1 to 3, characterized in that:
母材に積層される付加材と、積層体の厚さ及び許容される欠陥を含む要求性能と、前記積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む要求品質と、を示すデータを取得する取得処理と、
付加材の供給量及び熱源からの入熱量を、積層体の厚さ及び欠陥レベルのそれぞれと関連付ける第１予測モデルを用いて、付加材ごとに、前記要求性能を満たし得る当該付加材の供給量及び熱源からの入熱量の領域を算出する第１推定処理と、
付加材の供給量及び熱源からの入熱量に応じて積層体内部の熱履歴を算出する第２予測モデルを用いて、前記第１推定処理において算出された供給量及び入熱量の領域に属する加工条件の下で、積層体内部の保持時間及び冷却速度を予測する第２推定処理と、
積層体の熱履歴を、積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む加工品質に関連付ける第３予測モデルを用いて、前記加工条件の下で、付加材の成分組成ごとに加工品質の予測値を算出する第３推定処理と、
前記第３推定処理において算出された品質の予測値を前記要求品質と比較し、比較結果に基づいて付加材の成分組成及び熱源からの入熱量の推奨値を決定する決定処理と、
を実行させることを特徴とする加工条件推奨方法。 to the computer,
Indicates an additional material to be laminated to the base material, required performance including the thickness and allowable defects of the laminate, and required quality including at least one of hardness, wear resistance, and toughness of the laminate. an acquisition process for acquiring data;
Using a first prediction model that associates the supply amount of the additive material and the amount of heat input from the heat source with the thickness and defect level of the laminate, calculate the supply amount of the additive material that can satisfy the required performance for each additive material. and a first estimation process for calculating a region of heat input from the heat source;
Processing that belongs to the region of the supply amount and heat input amount calculated in the first estimation process using a second prediction model that calculates the thermal history inside the laminate according to the supply amount of the additional material and the heat input amount from the heat source. a second estimation process that predicts the retention time and cooling rate inside the laminate under the conditions;
Using a third prediction model that relates the thermal history of the laminate to the processing quality including at least one of the hardness, wear resistance, and toughness of the laminate, under the processing conditions, for each component composition of the additive material, a third estimation process that calculates a predicted value of processing quality;
A determination process that compares the predicted quality value calculated in the third estimation process with the required quality and determines a recommended value for the component composition of the additional material and the amount of heat input from the heat source based on the comparison result;
A method for recommending machining conditions characterized by executing the following.
母材に積層される付加材と、積層体の厚さ及び許容される欠陥を含む要求性能と、前記積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む要求品質と、を示すデータを取得する取得処理と、
付加材の供給量及び熱源からの入熱量を、積層体の厚さ及び欠陥レベルのそれぞれと関連付ける第１予測モデルを用いて、付加材ごとに、前記要求性能を満たし得る当該付加材の供給量及び熱源からの入熱量の領域を算出する第１推定処理と、
付加材の供給量及び熱源からの入熱量に応じて積層体内部の熱履歴を算出する第２予測モデルを用いて、前記第１推定処理において算出された供給量及び入熱量の領域に属する加工条件の下で、積層体内部の保持時間及び冷却速度を予測する第２推定処理と、
積層体の熱履歴を、積層体の硬度、耐摩耗性及び靭性のうち少なくとも１つを含む加工品質に関連付ける第３予測モデルを用いて、前記加工条件の下で、付加材の成分組成ごとに加工品質の予測値を算出する第３推定処理と、
前記第３推定処理において算出された品質の予測値を前記要求品質と比較し、比較結果に基づいて付加材の成分組成及び熱源からの入熱量の推奨値を決定する決定処理と、
を実行させるためのプログラム。 to the computer,
Indicates an additional material to be laminated to the base material, required performance including the thickness and allowable defects of the laminate, and required quality including at least one of hardness, wear resistance, and toughness of the laminate. an acquisition process for acquiring data;
Using a first prediction model that associates the supply amount of the additive material and the amount of heat input from the heat source with the thickness and defect level of the laminate, calculate the supply amount of the additive material that can satisfy the required performance for each additive material. and a first estimation process for calculating a region of heat input from the heat source;
Processing that belongs to the region of the supply amount and heat input amount calculated in the first estimation process using a second prediction model that calculates the thermal history inside the laminate according to the supply amount of the additional material and the heat input amount from the heat source. a second estimation process that predicts the retention time and cooling rate inside the laminate under the conditions;
Using a third prediction model that relates the thermal history of the laminate to the processing quality including at least one of the hardness, wear resistance, and toughness of the laminate, under the processing conditions, for each component composition of the additive material, a third estimation process that calculates a predicted value of processing quality;
A determination process that compares the predicted quality value calculated in the third estimation process with the required quality and determines a recommended value for the component composition of the additional material and the amount of heat input from the heat source based on the comparison result;
A program to run.
母材に付加材を積層するための熱源と、
を含む金属構造体製造システム。 A processing condition recommendation device according to any one of claims 1 to 4,
a heat source for laminating additional material on the base material;
Metal structure manufacturing system including.
前記推奨値に基づいて母材に付加材を積層する工程と、
を含む金属構造体製造方法。
a step of obtaining recommended values for the component composition of the additive material and the amount of heat input from the heat source based on the processing condition recommendation method according to claim 5;
Laminating an additional material on the base material based on the recommended value;
A metal structure manufacturing method including.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2020095193 | 2020-06-01 | ||
| JP2020095193 | 2020-06-01 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JP2021186883A JP2021186883A (en) | 2021-12-13 |
| JP7428350B2 true JP7428350B2 (en) | 2024-02-06 |
Family
ID=78847927
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP2021090839A Active JP7428350B2 (en) | 2020-06-01 | 2021-05-31 | Processing condition recommendation device, processing condition recommendation method, program, metal structure manufacturing system, and metal structure manufacturing method |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JP7428350B2 (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP7353413B1 (en) | 2022-03-30 | 2023-09-29 | 株式会社アマダ | Processing system and processability judgment system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190188346A1 (en) | 2016-09-19 | 2019-06-20 | Siemens Product Lifecycle Management Software Inc. | System and method for modeling characteristics of a melt pool that forms during an additive manufacturing process |
| JP2020527475A (en) | 2017-05-24 | 2020-09-10 | リラティビティ スペース，インク． | Real-time adaptive control of additive manufacturing processes using machine learning |
-
2021
- 2021-05-31 JP JP2021090839A patent/JP7428350B2/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190188346A1 (en) | 2016-09-19 | 2019-06-20 | Siemens Product Lifecycle Management Software Inc. | System and method for modeling characteristics of a melt pool that forms during an additive manufacturing process |
| JP2020527475A (en) | 2017-05-24 | 2020-09-10 | リラティビティ スペース，インク． | Real-time adaptive control of additive manufacturing processes using machine learning |
Also Published As
| Publication number | Publication date |
|---|---|
| JP2021186883A (en) | 2021-12-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Majumder et al. | Prediction and optimization of surface roughness and micro-hardness using grnn and MOORA-fuzzy-a MCDM approach for nitinol in WEDM | |
| Zhang et al. | Predictive models for fatigue property of laser powder bed fusion stainless steel 316L | |
| Acayaba et al. | Prediction of surface roughness in low speed turning of AISI316 austenitic stainless steel | |
| Pandey et al. | Fuzzy expert system for prediction of kerf qualities in pulsed laser cutting of titanium alloy sheet | |
| Park et al. | Process modeling and parameter optimization using neural network and genetic algorithms for aluminum laser welding automation | |
| Tsai et al. | Optimal laser-cutting parameters for QFN packages by utilizing artificial neural networks and genetic algorithm | |
| Ciurana et al. | Estimating the cost of vertical high-speed machining centres, a comparison between multiple regression analysis and the neural networks approach | |
| Vasanth et al. | A neural network model to predict surface roughness during turning of hardened SS410 steel | |
| Paul et al. | Multi-objective optimization of some correlated process parameters in EDM of Inconel 800 using a hybrid approach | |
| Hossain et al. | Adaptive neuro-fuzzy inference system (ANFIS) based surface roughness prediction model for ball end milling operation | |
| JP7428350B2 (en) | Processing condition recommendation device, processing condition recommendation method, program, metal structure manufacturing system, and metal structure manufacturing method | |
| Chigilipalli et al. | An experimental investigation and neuro-fuzzy modeling to ascertain metal deposition parameters for the wire arc additive manufacturing of Incoloy 825 | |
| Pham et al. | Fast and accurate prediction of temperature evolutions in additive manufacturing process using deep learning | |
| Shahri et al. | Prediction of temperature and HAZ in thermal-based processes with Gaussian heat source by a hybrid GA-ANN model | |
| Sen et al. | GEP-and ANN-based tool wear monitoring: a virtually sensing predictive platform for MQL-assisted milling of Inconel 690 | |
| Chan et al. | An integrated FEM and ANN methodology for metal-formed product design | |
| Ponticelli et al. | An optimal fuzzy decision-making approach for laser powder bed fusion of AlSi10Mg alloy | |
| Era et al. | Prediction of mechanical behaviors of L-DED fabricated SS 316L parts via machine learning | |
| Bikas et al. | A decision support method for knowledge-based Additive Manufacturing process selection | |
| Singh Sivam et al. | An investigation of hybrid models FEA coupled with AHP-ELECTRE, RSM-GA, and ANN-GA into the process parameter optimization of high-quality deep-drawn cylindrical copper cups | |
| Hossain et al. | Surface roughness prediction model for ball end milling operation using artificial intelligence | |
| Haghshenas Gorgani et al. | A hybrid algorithm for adjusting the input parameters of the wirecut EDM machine in order to obtain maximum customer satisfaction | |
| Yang et al. | A study on fiber laser lap welding of thin stainless steel | |
| Sawant et al. | Surface roughness and surface crack length prediction using supervised machine learning–based approach of electrical discharge machining of deep cryogenically treated NiTi, NiCu, and BeCu alloys | |
| Yetilmezsoy et al. | Implementation of fuzzy logic approach to estimate the degree of expulsion and spattering index and weld strength in projection welding |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| A521 | Request for written amendment filed |
Free format text: JAPANESE INTERMEDIATE CODE: A821 Effective date: 20210601 |
|
| A621 | Written request for application examination |
Free format text: JAPANESE INTERMEDIATE CODE: A621 Effective date: 20230405 |
|
| TRDD | Decision of grant or rejection written | ||
| A977 | Report on retrieval |
Free format text: JAPANESE INTERMEDIATE CODE: A971007 Effective date: 20231220 |
|
| A01 | Written decision to grant a patent or to grant a registration (utility model) |
Free format text: JAPANESE INTERMEDIATE CODE: A01 Effective date: 20231226 |
|
| A61 | First payment of annual fees (during grant procedure) |
Free format text: JAPANESE INTERMEDIATE CODE: A61 Effective date: 20240117 |
|
| R150 | Certificate of patent or registration of utility model |
Ref document number: 7428350 Country of ref document: JP Free format text: JAPANESE INTERMEDIATE CODE: R150 |