JP2008084306A - Production support method and program of industrial product - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a production support method and program of an industrial product with which the optimal design parameter which meets desired conditions of a producer is obtained in a short time by minimum experiments and simulations. <P>SOLUTION: The production support method of the industrial product has: a process 203 for obtaining characteristic values to a part of combinations selected from among combinations at a plurality of levels set to each of a plurality of control factors; a process 203 for obtaining fluctuation of the characteristic values by an error factor about a part of the combinations; a process 204 for obtaining an average value and standard deviation of the characteristic values about all the combinations at all the levels of the control factors; a process 205 for obtaining frequency distribution of the characteristic values about all the combinations at all the levels of the control factors; a process 206 for inputting order acceptance conditions of the industrial product, production conditions of the industrial product and the desired conditions of the producer and a process 208 for extracting a combination at levels of control factors which meets the desired conditions by the frequency distribution. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  This invention is based on quality engineering and is used to determine the optimum design parameters of industrial products through minimal experiments and computer simulations, and to obtain the optimum design parameters that satisfy the desired conditions of the producer. The present invention relates to a production support method and program.

  In recent years, there has been a strict requirement for shortening the period and reducing the cost for the development and production of industrial products, and an effective method for that purpose is desired. On the other hand, the quality engineering method (also called Taguchi method) is a method for designing highly stable industrial products with little influence on characteristics due to use conditions and other error factors. An overview of this quality engineering approach is shown in FIG. The outline of the quality engineering method will be described with reference to FIG.

  In designing an industrial product, a design parameter that can be arbitrarily set by a designer is used as a control factor, and a parameter that causes an error in the result of operation characteristics or the like is used as an error factor. In the procedure 101, first, it is determined what parameters are assigned to the control factor and the error factor, and a plurality of levels (numerical values, process types, etc.) are set for the plurality of control factors. Multiple levels are set for each error factor. The characteristic value for the combination of these levels is obtained by experiment.

  Each level combination is selected based on an orthogonal table. That is, in step 102, each level of a plurality of control factors is assigned to the orthogonal table. In step 103, an experiment (including computer simulation) is performed based on the orthogonal table, and a characteristic value as a result is obtained.

  Only a very small number of partial combinations are selected for each level combination to be tested. For example, the standard L18 orthogonal table can test and evaluate a combination of a plurality of levels of 8 parameters (1 parameter is 2 levels and 7 parameter is 3 levels) as a control factor. There are 4374 total combinations of each level, but only 18 combinations are selected from the standard L18 orthogonal table. These 18 combinations are selected so that the evaluation for each level of each parameter can be performed equally. Thereby, it is possible to perform an equal and reasonable evaluation for each level of each parameter, and it is possible to greatly reduce the time for that.

  Next, in step 104, the S / N ratio (value indicating variation in characteristic values due to error factors) and sensitivity (value related to the average value of characteristic values) are calculated for the combination of the levels. Specifically, the SN ratio is a value in which (square of the average value of characteristic values) / (variance of the characteristic value) is expressed in decibel [dB], and the sensitivity is the square of the average value of characteristic values in decibel [dB]. This is the notation. In step 105, an optimum combination of the SN ratio and sensitivity is selected. That is, a combination of control factor levels that has high stability (robustness) with respect to an error factor and is close to a target characteristic (target characteristic) is determined.

  Then, in step 106, it is determined whether or not the optimum combination satisfies various conditions as an industrial product. If satisfied, the industrial product may be produced by the design. If various conditions are not satisfied, it is necessary to fundamentally change the design and production process. That is, if the condition is not satisfied in the procedure 106, the procedure 101 is repeated. Further, in this quality engineering technique, as shown as procedure 107, the SN ratio and sensitivity can be estimated for all combinations of all levels by using additiveness. That is, the average value of the characteristic values and the standard deviation of the characteristic values can be estimated for all combinations of all levels.

  If development using a quality engineering technique as shown in FIG. 1 is successful, an industrial product that is easy to use and can achieve a desired result can be developed regardless of who uses it under any use condition. However, there are few combinations of design parameter levels in the vicinity of the desired characteristics, in which variations in characteristic values are extremely small, and there are many limitations in this quality engineering technique.

The following Patent Document 1 is known as a design / manufacturing method using a quality engineering technique. Patent Document 1 uses a variation in temperature distribution of a molded product as an evaluation characteristic, obtains an optimal solution that minimizes the variation in temperature distribution, performs mold design based on the optimal solution, and manufactures a mold. Furthermore, it is described that injection molding is performed. In addition, using the quality engineering technique to extract the main parameters using the variation in the temperature distribution of the molded product as an evaluation characteristic, and select the value with a large SN ratio of the extracted parameter to reduce the variation in the temperature distribution of the molded product. It is described that an optimum solution to be minimized is obtained.
JP 2004-268318 A

  As described above, quality engineering techniques can be used effectively for optimum design in product development, and computer software for using such quality engineering techniques also exists. However, considering the order conditions (product specifications, price, delivery date, quantity, etc.), production conditions (material cost, processing cost, running cost, etc.) and producer's desired conditions (product accuracy, production cost, yield rate, etc.) There has never been a method for determining parameters for optimal production management, and there has been no computer software for that purpose.

  Therefore, the present invention obtains the optimum design parameters of industrial products based on quality engineering by minimum experiments and computer simulations, and obtains the optimum design parameters satisfying the producer's desired conditions in a short time. An object of the present invention is to provide a production support method and program for industrial products that can be used.

  In order to achieve the above object, the industrial product production support method of the present invention uses a design parameter for adjusting a characteristic value of an industrial product as a control factor, and a parameter that causes an error in the characteristic value of the industrial product. As the error factor, a procedure for obtaining the characteristic value for a partial combination selected from a plurality of combinations of levels set for each of the plurality of control factors, and for the partial combination, the error factor From the procedure for obtaining the variation of the characteristic value, and the characteristic value and the variation of the partial combination, the average value and the standard deviation of the characteristic value for all combinations of all levels of the control factor are added by additivity. A procedure for obtaining a frequency distribution of the characteristic values for all combinations of all levels of the control factors, and receiving the industrial product. Conditions, and procedures for inputting the production condition and producers of desired conditions of the industrial products, by the frequency distribution, and has a procedure for extracting a combination of levels of the control factors that satisfy the desired conditions.

  In the industrial product production support method, the procedure for obtaining the average value and the standard deviation of the characteristic values is set for the characteristic values corresponding to a plurality of levels set in the error factor and the levels. It is preferable to obtain them based on the frequency.

  In the production support method for industrial products, the frequencies of the error factors with respect to a plurality of levels may be set based on a normal distribution.

  In the above-described industrial product production support method, the frequencies of the error factors for a plurality of levels are set based on a distribution obtained by combining two normal distributions having different standard deviations with the maximum likelihood value as a boundary. be able to.

  In the industrial product production support method, the partial combination is preferably selected based on an orthogonal table.

  In the industrial product production support method, it is preferable that the procedure for obtaining the characteristic value and the procedure for obtaining the fluctuation of the characteristic value are to obtain the characteristic value and the fluctuation thereof by computer simulation.

  In the above-described industrial product production support method, it is preferable that the procedure for obtaining the frequency distribution of the characteristic values is obtained assuming that the frequency distribution is a normal distribution.

  The industrial product production support program of the present invention uses a design parameter for adjusting the characteristic value of the industrial product as a control factor, and a parameter that causes an error in the characteristic value of the industrial product as an error factor. A procedure for obtaining the characteristic values for some combinations selected from a plurality of combinations of levels set for each of the control factors, and for the some combinations, fluctuations in the characteristic values due to the error factors Obtaining a mean value and standard deviation of the characteristic values for all combinations of all levels of the control factor by additiveness from the characteristic values of the partial combinations and their variations; and the control A procedure for obtaining the frequency distribution of the characteristic values for all combinations of all levels of factors, the order conditions for the industrial product, the process A step of inputting a production condition and producers of desired conditions of the product, by the frequency distribution, in which to execute the steps of extracting a combination of levels of the control factors that satisfy the desired conditions in the computer.

  In the industrial product production support program, the procedure for obtaining the average value and standard deviation of the characteristic values is set for the characteristic values corresponding to a plurality of levels set for the error factor and the levels. It is preferable to obtain them based on the frequency.

  In the industrial product production support program, the frequencies of the error factors with respect to a plurality of levels may be set based on a normal distribution.

  In the industrial product production support program, the frequencies of the error factors for a plurality of levels are set based on a distribution obtained by combining two normal distributions having different standard deviations with the maximum likelihood value as a boundary. be able to.

  Further, in the industrial product production support program, the partial combination is preferably selected based on an orthogonal table.

  In the industrial product production support program described above, the procedure for obtaining the frequency distribution of the characteristic values is preferably obtained by assuming that the frequency distribution is a normal distribution.

  Since this invention is comprised as mentioned above, there exist the following effects.

  With a small number of experiments, each level combination of control factors that satisfies the producer's desired conditions can be selected in a short time, and the development and production period of industrial products can be shortened and the cost can be reduced. It is possible.

  By setting the frequency to each level of the error factor, the average value and the standard deviation of the characteristic values can be obtained with high accuracy in consideration of the statistical distribution of the error factor values.

  By using computer simulation, it is possible to obtain an optimum level of combination in a short time without actually conducting experiments using experimental equipment.

  Embodiments of the present invention will be described with reference to the drawings. FIG. 2 is a flowchart showing the procedure of the production support method according to the first embodiment of the present invention. Each procedure of this production support method is executed on a computer. Similar to the quality engineering method described above, in designing an industrial product, a design parameter that can be arbitrarily set by a designer is used as a control factor, and a parameter that causes an error in the result of operation characteristics or the like is used as an error factor. In step 201, first, it is determined what parameters are assigned to the control factor and the error factor, and a plurality of levels (numerical values, process types, etc.) are set for each of the control factors, and one or more error factors are set. Also set multiple levels for each.

  Next, in step 202, each level of a plurality of control factors is assigned to an orthogonal table (for example, a standard L18 orthogonal table). That is, the combination of each level is selected based on the orthogonal table. In step 203, an experiment (including computer simulation) is performed based on the orthogonal table, and a characteristic value as a result is obtained. That is, the characteristic value by the combination of the levels of the control factors is obtained, and at the same time, the fluctuation of the characteristic value by the error factor is obtained. Only a very small number of partial combinations are selected for each level combination to be tested. As described above, the experimental design based on the orthogonal table can perform an equal and reasonable evaluation for each level of each parameter, and can greatly reduce the time required for the evaluation.

  Note that the reduction in the number of experiments in steps 202 and 203 is not limited to being performed based on the orthogonal table, and the number of experiments may be reduced by other methods. However, it is desirable to have a method that can equally evaluate each level of each control factor. Furthermore, the experiment of the procedure 203 may be actually performed using experimental equipment, or may be a computer simulation. As the simulation, for example, a simulation by a finite element method can be used. Further, with respect to the procedures 201 to 203, computer software for implementing quality engineering can be used as it is.

  Next, in step 204, average values and standard deviations of characteristic values are calculated for all combinations of levels of all control factors. As described above, the quality engineering software can determine the SN ratio and sensitivity of characteristic values using additiveness for all combinations of all control factors, so that data can be used. The average value and standard deviation of characteristic values can be calculated easily.

  Next, in step 205, the frequency distribution of characteristic values is calculated for all combinations of levels of all control factors. In general, the frequency distribution can be calculated from the average value and the standard deviation, assuming that the distribution is a normal distribution. However, when a distribution other than the normal distribution is predicted in advance, the predicted distribution can be used.

  Next, in step 206, data on order conditions, production conditions, and producer's desired conditions are input. These conditions are data necessary for production management. Order conditions include product specifications, allowable range of characteristic values, price, delivery date, order quantity, and the like. Production conditions include material costs, processing costs, production equipment running costs, labor costs, disposal unit costs, depreciation costs, and the like. The conditions desired by the producer include product accuracy, production cost, production speed, and yield rate.

  This desired condition is set on the assumption of the order receiving condition and the production condition, and is set so as to satisfy both the order receiving condition and the production condition. Also, for example, if you want to make the production cost as low as possible, set the condition of production cost: XX or less, and if you want to produce as soon as possible, set the condition of production speed: XX or more. That's fine.

  Next, in step 207, production management factors are calculated for all combinations of all control factors. This is a calculation of a factor for determining whether or not the desired condition input in step 206 is satisfied. Production management factors include the yield rate, total production quantity, processing time, total production time, and production cost. The yield rate is determined from the frequency distribution of characteristic values and the allowable range of characteristic values. The total production quantity is obtained as the order quantity / non-defective rate. The processing time is obtained as the reciprocal of the production speed. The total production time is obtained as total production number × processing time. The production cost is calculated as (material cost + running cost) × total production number + disposal unit price × number of disposal + work cost × total production time.

  Next, in step 208, all the combinations of the levels of the control factors satisfying the desired condition are extracted from the combinations of the levels of all the control factors, and the extracted combinations are output. When a plurality of combinations are applicable, they are output in order from the most preferable one. When outputting combinations of levels satisfying the desired conditions, it is preferable to output not only the combinations of levels but also the production management factors in the combinations at the same time. The producer may perform the production by a combination of levels satisfying the desired conditions. When a plurality of combinations are output, the most appropriate combination may be selected in consideration of other secondary conditions.

  Moreover, if the SN ratio and sensitivity of the characteristic values are set simultaneously as the desired conditions, the desired conditions from the viewpoint of production management factors are satisfied, and at the same time, the conditions regarding the SN ratio and sensitivity from the viewpoint of quality engineering are also satisfied. It is also possible to find an optimal combination of levels.

  Next, a second embodiment of the present invention will be described. FIG. 3 is a flowchart showing the procedure of the production support method according to the second embodiment of the present invention. Each procedure of the production support method of the second embodiment is also executed on the computer as in the first embodiment. In the second embodiment, processing is performed according to substantially the same procedure as in the first embodiment, but the frequency is set for each level of error factors to improve the accuracy of subsequent calculation results. It is what I did. The following description will focus on improvements to the first embodiment.

  As in the first embodiment, a design parameter that can be arbitrarily set by the designer is used as a control factor, and a parameter that causes an error in the result of operation characteristics or the like is used as an error factor. In step 301, it is first determined what parameters should be assigned to control factors and error factors, and a plurality of levels (numbers, process types, etc.) are set for each of the control factors, and one or more error factors are set. Also set multiple levels for each. In the second embodiment, the frequency is set for each of a plurality of levels set as error factors.

  The statistical distribution of error factor values is generally not a flat distribution, but often shows a distribution curve such as a normal distribution. However, in the quality engineering technique and the first embodiment of the present invention, each of a plurality of levels of error factors is treated equally, and the SN ratio and sensitivity of the subsequent characteristic values are obtained. As a result, an error also occurs in the distribution of characteristic values (average value and standard deviation). That is, in quality engineering and the first embodiment, it is equivalent to processing the distribution of error factors as a flat one.

  In quality engineering, qualitative changes in the S / N ratio and sensitivity are often a problem, and the S / N ratio and sensitivity errors caused by the distribution of error factor values are rarely a problem. However, in the present invention, since the production management factor is calculated using the distribution of characteristic values obtained from the SN ratio and sensitivity, it is extremely important to obtain the SN ratio and sensitivity with high accuracy.

  In addition to setting multiple levels for error factors, by setting the frequency for each of these multiple levels, the statistical distribution of error factor values is taken into account, increasing the signal-to-noise ratio and sensitivity. The accuracy can be obtained. Specifically, when the frequency n is set with respect to the level i of the error factor, the characteristic value data for the level i is not n pieces of data when calculating the average value and the standard deviation of the characteristic values. Can be handled as Note that as the procedure 301, computer software for performing general quality engineering cannot be used, so it is necessary to use software unique to the present invention.

  Since the procedure 302 and the procedure 303 are the same as the procedure 202 and the procedure 203 of the first embodiment, a description thereof will be omitted. The procedure 304 is the same as the procedure 204, except that the characteristic value data for the level i of the error factor is handled as n data as described above, and the calculation is performed taking into account the statistical distribution of the error factor values. Since the steps 305 to 308 are the same as the steps 205 to 208 of the first embodiment, description thereof is omitted.

  According to the production support method of the second embodiment, the average value and the standard deviation of the characteristic values can be obtained with high accuracy in consideration of the statistical distribution of error factor values. Compared with the calculation method normally performed so far (the number of error factor levels is set to 2 and 3, and the frequency is not set), in this second embodiment, the calculation accuracy of the average value and the standard deviation of the characteristic values is 4 to 4. It can be improved 5 times or more. Specifically, when the level number of the error factor is 3, the average value of the characteristic value and the calculation error of the standard deviation are about 10%, but when the level number is 5 and an appropriate frequency is set for each level, Those calculation errors could be reduced to 1%.

  Next, an error factor level and frequency setting method in the second embodiment will be described. The number of levels set for the error factor and the numerical range of each level may be arbitrary, and these values can be set individually, and the frequency according to the distribution of error factors can be set for each level. Such a setting method has the highest degree of freedom and can be applied to any error factor distribution, but individual input of set values is troublesome and there is a high possibility of erroneous input of numerical values.

  Since the error factor distribution is often a normal distribution, the setting of the level and frequency of the error factor can be simplified based on this distribution. The distribution of error factors is assumed to be a normal distribution, and an average value and a standard deviation are input for each of a plurality of error factors. For each error factor, enter the entire range to be evaluated and the number of levels to divide it. For each level set by equally dividing the entire range by the number of levels, an appropriate frequency is calculated and set based on the normal distribution. According to such an input method, even if the number of levels is increased, there are few numerical values to be input, and the level and frequency of the error factor can be easily set.

  The distribution of error factors often deviates from the normal distribution and is asymmetric with respect to the maximum likelihood value (peak value). In such a case, two normal distributions having different standard deviations on both sides of the maximum likelihood value can be approximated as distributions synthesized so as to be continuous at the maximum likelihood value. Even in such a case, the setting of the level and frequency of the error factor can be simplified based on two normal distributions. Assuming that the distribution of error factors is a composite normal distribution as described above, the average value and standard deviation of two normal distributions are input for each of a plurality of error factors.

  For each error factor, enter the evaluation range on both sides of the maximum likelihood value and the number of levels to equally divide it. For each level set by equally dividing the evaluation range on both sides by the number of levels, an appropriate frequency is calculated and set based on the normal distribution. According to such an input method, an asymmetric distribution deviating from the normal distribution can be dealt with. Further, even if the number of levels is increased, there are few numerical values to be input, and the level and frequency of the error factor can be easily set.

  As described above, by using the production support method for industrial products of the present invention, it is possible to appropriately select combinations of control factors satisfying the producer's desired conditions in a short time with a small number of experiments. . Moreover, if a computer simulation is used, an optimum combination of levels can be obtained in a short time without actually performing an experiment using experimental equipment.

  It should be noted that the industrial product production support method of the present invention is suitably implemented as a program executed on a computer.

  According to the present invention, by a small number of experiments, combinations of levels of control factors that satisfy the producer's desired conditions can be determined in a short time, and the development and production period of industrial products can be shortened and reduced. Cost reduction is realized.

It is a figure which shows the outline | summary of the technique of quality engineering. It is a flowchart which shows the procedure of the production support method of 1st Embodiment of this invention. It is a flowchart which shows the procedure of the production support method of 2nd Embodiment of this invention.

Explanation of symbols

101-107 Each procedure of the technique of quality engineering 201-208 Each procedure of the production support method of 1st Embodiment of this invention 301-308 Each procedure of the production support method of 2nd Embodiment of this invention

Claims (13)

A design parameter for adjusting the characteristic value of the industrial product is used as a control factor, and a parameter that causes an error in the characteristic value of the industrial product is used as an error factor, and a plurality of levels set for each of the plurality of control factors are set. A procedure (203, 303) for obtaining the characteristic values for some combinations selected from the combinations;
A procedure (203, 303) for obtaining a variation in the characteristic value due to the error factor for the partial combination;
A procedure (204, 304) for obtaining an average value and a standard deviation of the characteristic values for all combinations of all levels of the control factor by additiveity from the characteristic values for the partial combinations and variations thereof;
A procedure (205, 305) for obtaining a frequency distribution of the characteristic values for all combinations of all levels of the control factors;
A procedure (206, 306) for inputting the order condition of the industrial product, the production condition of the industrial product and the desired condition of the producer;
And a procedure (208, 308) for extracting a combination of levels of the control factors satisfying the desired condition from the frequency distribution. A production support method for an industrial product according to claim 1,
The procedure (304) for obtaining the average value and standard deviation of the characteristic values is to obtain them based on the characteristic values corresponding to a plurality of levels set as the error factor and the frequencies set for the levels. A production support method for industrial products. A production support method for an industrial product according to claim 2,
The industrial product production support method, wherein the frequencies of the error factors with respect to a plurality of levels are set based on a normal distribution. A production support method for an industrial product according to claim 2,
The frequency for the plurality of levels of the error factor is set based on a distribution obtained by synthesizing two normal distributions having different standard deviations with the maximum likelihood value as a boundary. A production support method for industrial products according to any one of claims 1 to 4,
The method for supporting production of an industrial product, wherein the partial combination is selected based on an orthogonal table. A production support method for an industrial product according to any one of claims 1 to 5,
The procedure (203, 303) for obtaining the characteristic value and the procedure (203, 303) for obtaining the variation of the characteristic value are for obtaining the characteristic value and its variation by computer simulation. . A production support method for an industrial product according to any one of claims 1 to 6,
The procedure (205, 305) for obtaining the frequency distribution of the characteristic values is an industrial product production support method for obtaining the frequency distribution as a normal distribution. A design parameter for adjusting the characteristic value of the industrial product is used as a control factor, and a parameter that causes an error in the characteristic value of the industrial product is used as an error factor, and a plurality of levels set for each of the plurality of control factors are set. A procedure (203, 303) for obtaining the characteristic values for some combinations selected from the combinations;
A procedure (203, 303) for obtaining a variation in the characteristic value due to the error factor for the partial combination;
A procedure (204, 304) for obtaining an average value and a standard deviation of the characteristic values for all combinations of all levels of the control factor by additiveity from the characteristic values for the partial combinations and variations thereof;
A procedure (205, 305) for obtaining a frequency distribution of the characteristic values for all combinations of all levels of the control factors;
A procedure (206, 306) for inputting the order condition of the industrial product, the production condition of the industrial product and the desired condition of the producer;
An industrial product production support program for causing a computer to execute a procedure (208, 308) for extracting a combination of levels of the control factors satisfying the desired condition from the frequency distribution. A production support program for an industrial product according to claim 8,
The procedure (304) for obtaining the average value and standard deviation of the characteristic values is to obtain them based on the characteristic values corresponding to a plurality of levels set as the error factor and the frequencies set for the levels. Is a production support program for industrial products. A production support program for an industrial product according to claim 9,
The industrial product production support program in which the frequencies of the error factors with respect to a plurality of levels are set based on a normal distribution. A production support program for an industrial product according to claim 9,
The frequency of the error factor with respect to a plurality of levels is set on the basis of a distribution obtained by combining two normal distributions having different standard deviations with the maximum likelihood value as a boundary. An industrial product production support program according to any one of claims 8 to 11,
An industrial product production support program in which the partial combination is selected based on an orthogonal table. A production support program for an industrial product according to any one of claims 8 to 12,
The procedure (205, 305) for obtaining the frequency distribution of the characteristic values is an industrial product production support program for obtaining the frequency distribution as a normal distribution.

Images