JP2018045266A - Design support device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a design support device capable of shortening a calculation time and having high prediction accuracy.SOLUTION: A design support device according to the present invention includes: a data analysis section that predicts prediction data corresponding to input data on the basis of learning data stored in a storage section; an additional candidate point calculation section that calculates additional candidate points of the prediction data predicted by the data analysis section, from distribution of the learning data and the prediction data; an additional candidate point correction section that corrects the additional candidate points calculated by the additional candidate point calculation part, by simulation; and an output section that visualizes and displays at least any of the prediction data predicted by the data analysis section, the additional candidate points calculated by the additional candidate point calculation part, and the additional candidate points corrected by the additional candidate point correction part.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a design support apparatus that learns a design space composed of design parameters with respect to design parameters constituting products and services by machine learning and predicts design parameters that realize a design plan according to a customer's request.

  In recent years, in the development of products and services, in order to maximize customer value, it is required to develop products and services that meet customer needs while communicating with customers. In order to facilitate communication with customers, it is effective to estimate and present specifications of products and services that meet customer needs in a short time. However, if the number of design parameters that make up a product or service increases, it takes time to evaluate the specifications proposed by the designer in consideration of the customer's request using the design tool, which increases the number of man-hours for reviewing estimates. It will lead to. For this reason, a technology that can reduce the time required for estimation and present a product specification proposal is effective.

  In order to reduce the time required to examine the specifications of products and services and present a specification proposal, it is possible to reduce the time required for calculation such as performance evaluation with a design tool. Calculation time can be reduced by increasing computational resources, using high performance computers, or simplifying models.

  However, if the design tools already used are improved, there is a problem that it is necessary to re-verify the accuracy of the calculation results. Also, it is difficult to prepare a high-performance computer directly for customer interaction. Therefore, using design tools, the design space consisting of design parameters is learned by machine learning within the expected range corresponding to the customer's request, and the method of predicting the design parameter of the product specification according to the customer's request is a fast estimate It is effective in performing. Here, the design parameters represent the dimensions and performance of the parts that make up the product with numerical values, etc., and the product specification draft is the shape and structure of the parts determined from the dimensions and performance values of the design parameters, for example, between the parts. Corresponds to the provisions of layout, number of required parts, materials to be used, etc.

  Patent Documents 1 and 2 show methods for creating and optimizing a design space using a design tool. Patent Document 1 discloses a configuration in which an experimental design method sample is created on a design space using an experimental design method and a genetic algorithm, and sample points are evaluated.

  Further, Patent Document 2 learns the mapping relationship between a factor group and a characteristic group of experimental data such as design / combination, etc., estimates a characteristic value from the factor condition, and creates a factor for arbitrary characteristic data. A method for obtaining the optimum value of data is disclosed.

JP 2010-009595 A JP 2003-58582 A

  In order to present specifications for customer requirements at high speed, when predicting the optimum values of product design parameters using machine learning, if the learned sample distribution is biased, the prediction accuracy will be reduced.

  Patent Documents 1 and 2 do not consider sample bias. Further, in Patent Document 1, only sample points reflecting the simulation result are evaluated, and the number of data is small.

  Accordingly, an object of the present invention is to provide a design support apparatus that can shorten the calculation time and has high prediction accuracy.

  In order to solve the above problems, a design support apparatus according to the present invention includes an input unit that receives input data, a storage unit that stores learning data, and an input based on the learning data stored in the storage unit. A data analysis unit that predicts prediction data corresponding to the generated data, an additional candidate point calculation unit that calculates additional candidate points from the learning data and the distribution of the prediction data for the prediction data predicted by the data analysis unit, and an addition Additional candidate point correction unit for correcting additional candidate points calculated by the candidate point calculation unit, prediction data predicted by the data analysis unit, additional candidate points calculated by the additional candidate point calculation unit, and additional candidate point correction And an output unit that visualizes and displays at least one of the additional candidate points corrected by the unit.

  ADVANTAGE OF THE INVENTION According to this invention, calculation time can be shortened and the design support apparatus with high prediction accuracy can be provided.

It is a block diagram of the design support apparatus which concerns on one Embodiment of this invention. 3 is an example of a processing flow according to the first embodiment. It is an example of the result of having visualized the design parameter to learn. It is an example of the method of adding the design parameter which should be learned. It is an example of the data for learning. It is an example of the screen which visualized the design parameter to learn. It is an example of the screen which visualized the design parameter to learn. It is an example of the screen which visualized the design parameter to learn. It is an example of the screen which visualized the design parameter to learn.

  In the present invention, a new sample point (additional candidate point) is calculated from the learning data and the distribution of the prediction data so that the prediction error in the variable range of the input design parameter is not more than a predetermined value. We propose a device that simulates additional candidate points calculated by calculation, and corrects it based on past performance data and product operation / environmental conditions based on the simulation results.

  Data predicted by machine learning is a design parameter of specifications such as products and services that match the items requested by the customer. High prediction accuracy of machine learning leads to a reduction in the number of reviews of product specification proposals. On the other hand, when the prediction accuracy is low, the design estimate is revised, and the number of examination steps increases. For this reason, it is important to improve the prediction accuracy of design parameters. However, when there is a bias in the distribution of data to be learned, the prediction accuracy deteriorates if the pattern of the specification proposal presented for a certain customer request item is not learned. Therefore, it is necessary to prepare simulation data by executing simulation so that the distribution is not biased. However, in order to prepare learning data that covers the design space, it leads to increasing the number of times of execution of simulation, and calculation time is required. Therefore, it is effective to prepare learning data at a location where the prediction accuracy should be increased. In particular, when not only the design parameters that make up a product but also the parameters that make up a service are targets of machine learning, it takes more time to create comprehensive learning data. For this reason, after obtaining the location where the prediction accuracy should be increased from the past prediction results, the actual data, and the product operating environment data, it is calculated by simulation whether it can be registered as the data to be learned. The data to be learned is corrected based on the simulation result and presented to the user. Thereby, it is possible to predict the specification design plan in a short time with high accuracy.

  Hereinafter, examples will be described with reference to the drawings.

  FIG. 1 is a configuration diagram of a design support apparatus according to an embodiment of the present invention. This system creates learning data for presenting product specification proposals for customer requirements by machine learning from simulation calculation results, past prediction results, past performance data, and operating environment condition data. . The design support apparatus calculates an additional candidate point, an input unit 101 to which various data is input, a storage unit 104 that stores the various data, a data analysis unit 102 that predicts prediction data corresponding to the input data, and the like. An additional candidate point calculation unit 103, an additional candidate point correction unit 105 that corrects the additional candidate point, an output unit 108 that visualizes and displays at least one of the prediction data, the additional candidate point, and the corrected additional candidate point; Is provided.

  Various parameters (hereinafter referred to as input data) that are customer-specified specifications are input to the input unit 101.

  The data analysis unit 102 predicts prediction data corresponding to the input data from the learning data stored in the storage unit. Here, the learning data is data in which the correspondence relationship between customer requirement specifications and design parameters based on past performance data or calculation results is known. As the learning data, data stored in the storage unit in advance is used. Note that new learning data may be input to the input unit and stored in the storage unit.

  The data analysis unit first analyzes the characteristics of the learning data stored in the storage unit, and visualizes the distribution of the learning data. As a visualization method, a principal component analysis or a self-organizing map method for compressing multidimensional data into two or three dimensions can be used. In the present embodiment, description will be given by taking principal component analysis as an example. After the principal component analysis in the data analysis unit, the distribution of learning data is displayed on a two-dimensional graph composed of x and y axes. The result of the feature analysis of the learning data is learned in the storage unit. Using the learned data, output data (predicted data) corresponding to the input data is predicted. As for the prediction data, the result of dimension compression using principal component analysis can be displayed as a prediction point on a two-dimensional graph.

  The additional candidate point calculation unit 103 calculates additional candidate points from the distribution of learning data and prediction data in order to improve learning accuracy. For example, additional candidate points can be obtained geometrically from the distribution of learning data. The relationship between input and output of additional candidate points is stored in the storage unit 104 as learning data.

  The additional candidate points are preferably obtained so that the prediction accuracy in a certain variable range is not more than a predetermined value. For example, the designer inputs a variable range and a predetermined value into the design support apparatus in advance, and calculates additional candidate points so that the prediction accuracy in the variable range input by the designer is less than or equal to the designer's input value. Thus, it is possible to add sample points that meet the conditions set by the designer in consideration of the operation and environmental conditions of the product.

  Moreover, it is preferable that an additional candidate point calculation part determines the necessity of an additional candidate point. When determining whether or not additional candidate points are necessary, a simulation execution unit executes a simulation of predicted data. Thereafter, an error between the prediction data and the simulation result is calculated by the additional candidate point calculation unit. If the error is within the threshold, it is determined that the additional candidate point is “unnecessary”, and the additional candidate point is not added. On the other hand, when the error exceeds the threshold, it is determined that the additional candidate point is “necessary”, and the additional candidate point is added from the distribution of the learning data and the prediction data.

  The additional candidate point correction unit corrects the additional candidate points by simulation. The additional candidate point correction unit performs inverse transformation from the coordinates of the additional candidate points on the principal component space, and creates additional candidate point learning data. The simulation execution unit 106 executes a simulation for learning data for additional candidate points. The additional candidate point correction unit corrects the additional candidate points so as to match the simulation result.

  Moreover, it is preferable that an additional candidate point correction | amendment part evaluates whether an additional candidate point is materialized as learning data. The additional candidate points are obtained geometrically from the distribution of learning data, and it is necessary to verify whether they are valid as design data. For example, the additional candidate point correcting unit can compare the additional candidate point obtained by the additional candidate point calculating unit with the simulation result and determine whether the additional candidate point calculating unit needs to be corrected. If it is determined that the correction is “necessary”, the additional candidate points may be corrected based on the simulation result.

  The output unit 108 displays the corrected additional candidate points. The additional candidate points are preferably displayed on a two-dimensional graph composed of x and y axes. The output unit may display the prediction data predicted by the data analysis unit and the additional candidate points calculated by the additional candidate point calculation unit.

  FIG. 2 is a flow showing a process when adding learning data. First, in the input unit, learning data such as past performance data and simulation calculation data, and customer requirement specifications (hereinafter referred to as input data) are read as learning data (201). The read learning data is stored in the storage unit.

  Then, the data analysis unit compresses the learning data from multidimensional data to two-dimensional data by a method such as principal component analysis (202). By performing principal component analysis, a principal component axis capturing data characteristics can be obtained, and data can be plotted on a two-dimensional graph composed of a first principal component axis and a second principal component axis. The output unit 108 displays the analysis result of the learning data by the data analysis unit on a two-dimensional graph composed of the first principal component axis and the second principal component axis (203). After learning the result of the feature analysis of the learning data by the data analysis unit (204), prediction data (output data) corresponding to the input data is predicted by machine learning (205). For the combination of input and output, principal component scores (additional candidate points) are calculated from the result of the feature analysis performed at 202 (206). The calculated principal component score is plotted on a two-dimensional graph composed of principal component axes (207).

  In order to evaluate how much the prediction result for the input data is different from the result of the simulation, a simulation is performed by the simulation execution unit (208). The additional candidate point calculation unit evaluates an error between the prediction result and the simulation result (209). If the error between the prediction result and the simulation result is within the threshold, the position of the plotted prediction point is not moved (216). If the error is equal to or greater than the threshold value, it is necessary to improve the learning accuracy, so the learning data is added as a new point. Hereinafter, the added new point is referred to as an additional candidate point. The additional candidate points are added so that a triangle including the prediction result can be formed on the principal component space (210).

  In the additional candidate point correction unit, the coordinates of the additional candidate points in the obtained principal component space are subjected to inverse transformation of the principal component result, and learning data corresponding to the additional candidate points is created (211). A simulation is performed on the created learning data (inversely transformed data) to evaluate whether additional candidate points can be used as learning data. Specifically, the calculation is executed using the simulation input value included in the inversely transformed data (212). The calculation result is converted again into principal component scores and plotted on a two-dimensional graph (213). At this time, if the position of the simulation result and the position of the additional candidate point are the same, the position of the point is not changed (214). If the position of the additional candidate point is shifted from the simulation result, the coordinates of the additional candidate point are corrected to the position of the simulation result (215).

  Machine learning is performed on the corrected additional candidate points, and design parameters can be obtained as output data corresponding to the input customer request items.

  FIG. 3 shows a two-dimensional graph composed of principal component axes on the screen. In the two-dimensional graph, the horizontal axis 302 is the first principal component axis and the vertical axis 303 is the second principal component axis, and the learning data 304, the prediction data 305, and the additional candidate points 306 are displayed. . In FIG. 3, the shape of the plot is changed depending on the type of data, but other highlighting is also possible, for example, the display is divided by color. On the screen, a button 307 for reading learning data and a new sample point, a button 308 for displaying a result of principal component analysis, a button (309) for displaying a prediction result at a new sample point, and an additional candidate point are created. A button 310, a button 311 for correcting the position of the candidate point, and a button 312 for outputting learning data including additional candidate points are provided.

  FIG. 4 shows a method for obtaining additional candidate points from a two-dimensional graph of principal component scores. (A) is a two-dimensional graph composed of principal component axes. (B) is an enlarged view of a region 401 surrounded by a solid line in (a). Points p1 and p2 indicate learning data, and point p3 indicates prediction data. First, it is searched whether or not learning data is included around the prediction data p3 predicted by the data analysis unit. For example, it is assumed that a point p1 and a point p2 are included in a certain radius r. At this time, as shown in (c), a point p3 that is prediction data is included therein, and points that form a triangle with the points p1 and p2 are added as additional candidate points p4. For example, the point p4 (406) is added to the position where the distance from the point p3 is maximum on the circumscribed circle (402) of the points p1 and p2. A circumscribed circle passing through the points p1 and p2 is expressed by the following (formula 1).

  The position where the distance between the points p3 and p4 becomes the maximum can be obtained by the following (Equation 2).

  It is evaluated whether or not the triangle composed of the points p1, p2, and p4 includes the point p3. If not, the position of the point p4 is moved and it is evaluated again whether it is included in the triangle. Thereby, an additional candidate point is newly created. (D) is a two-dimensional graph after adding additional candidate points.

  FIG. 5 is an example of learning data (a) and prediction data (b) when the target product is a compressor. The relationship between input and output at the time of learning can be indicated by, for example, a customer request item 501 and a design parameter 502. As customer requirement items and design parameters, there are parameters such as flow rate, pressure, molecular weight, number of stages, number of rotations, impeller diameter, and the like. Such learning data is compressed into two-dimensional data composed of principal component axes by principal component analysis. As for the prediction data, input data 503 and output data 504 are determined. The output data is a prediction result by machine learning.

  FIG. 6 is an example of a screen for explaining an operation of displaying the result of compressing the learning data by principal component analysis as a two-dimensional graph. When the learning data and the prediction data are read with the read button 602 and the data display button 603 is pressed, the learning data is displayed. Further, when the prediction result display button 604 is pressed, the prediction result of the new sample data is displayed in a form as indicated by 305.

  FIG. 7 is an example of a screen showing a prediction result when the prediction result display button 702 is pressed. When the addition candidate point creation button 703 is pressed after the prediction result is shown, candidate points to be added are created for the learning data and the prediction result. An example of the creation method is a method of obtaining from geometric features as described in FIG.

  FIG. 8 is an example of a screen when additional candidate points are created. When an additional candidate point creation button 801 is pressed, an additional candidate point is created. Subsequently, when the candidate point correction button 803 is pressed, a simulation is performed to correct the position of the additional candidate point in order to evaluate whether or not the additional candidate point obtained from the geometric feature is established as learning data. Display on a dimensional graph.

  FIG. 9 is an example showing a result of correcting the position of the additional candidate point. When the candidate point correction button 902 is pressed, the position of the added candidate point is corrected, and when the data output button 903 is pressed, the principal component score of the added candidate point is inversely converted by principal component analysis, and the additional candidate point is input. The learning data including the output data is output in text or csv format.

  As described above, in the present invention, after obtaining a place where the prediction accuracy should be improved by machine learning from past prediction results, actual data, and product operating environment data, the data created at that place should be simulated and learned. The data is corrected and presented to the user. Accordingly, since appropriate data can be added to the learning data, the accuracy of machine learning can be improved, and the specification plan can be predicted in a short time with high accuracy.

DESCRIPTION OF SYMBOLS 101 ... Input part, 102 ... Data analysis part, 103 ... Additional candidate point calculation part, 104 ... Memory | storage part, 105 ... Additional candidate point correction part, 106 ... Simulation execution part, 107 ... Visualization part, 108 ... Output part, 601, 701, 801, 901 ... Display screen, 307, 602 ... Read button, 308, 603 ... Data display button, 309, 604, 702 ... Prediction result display button, 310, 703, 802 ... Addition candidate point creation button, 311, 803 902 ... Candidate point correction button 312, 903 ... Data output button

Claims (5)

An input section for inputting data;
A storage unit for storing learning data;
A data analysis unit that predicts prediction data corresponding to input data based on the learning data stored in the storage unit;
For the prediction data predicted by the data analysis unit, an additional candidate point calculation unit that calculates additional candidate points from the learning data and the distribution of the prediction data;
An additional candidate point correction unit that corrects the additional candidate points calculated by the additional candidate point calculation unit by simulation; and
Output for visualizing and displaying at least one of predicted data predicted by the data analysis unit, additional candidate points calculated by the additional candidate point calculation unit, and additional candidate points corrected by the additional candidate point correction unit A design support apparatus. The design support apparatus according to claim 1,
The additional candidate point calculation unit calculates an additional candidate point from the learning data and the distribution of the prediction data so that the prediction accuracy in the input variable range is a predetermined value or less. The design support apparatus according to claim 1 or 2,
The additional candidate point calculation unit calculates additional candidate points from the learning data and the distribution of the prediction data so as to cover the prediction data. The design support apparatus according to claim 3,
The design support apparatus, wherein the additional candidate point correction unit evaluates by simulation whether the corrected additional candidate point can be used as learning data. A design support apparatus according to any one of claims 1 to 4,
The additional candidate point calculation unit determines whether or not an additional candidate point needs to be created based on a simulation result and prediction data predicted by the data analysis unit.

Images