JP5022729B2 - Flow number management system, method, and program - Google Patents

Description

  The present invention relates to a flow number management system, method, and program for determining and managing an appropriate level of inventory in, for example, an on-demand SCM (supply chain management) environment.

  In recent years, the like cycle of products tends to become shorter as society becomes more advanced, and fluctuations in demand for products are also increasing. In the distribution industry, having a lot of inventory in such a situation leads to serious financial flaws such as fixing funds, lowering sales, and increasing losses.

  The manufacturing industry has a series of chain operations that procure materials, make goods, and deliver them to customers. This is called the supply chain. This supply chain must work well in order to deliver products when customers demand (on demand). To that end, SCM is now in the spotlight, and many companies are working on reforms.

  Wholesale, which is located downstream of this supply chain, supplies its products to retailers. Information on retail outlets in contact with consumers arrives with some bias to wholesalers. The merit of wholesale trade is to obtain information from many retailers with different positions, motivation and sales. That is, by integrating information from a large number of retail stores, consumer trends are estimated and future demand is predicted.

  However, the conventional act of demand forecasting is merely to explain the past demand structure and assumes that the demand structure does not change, so there is a certain limit to its guarantee.

Therefore, in order to have the minimum necessary inventory without causing a shortage without relying on demand forecasts based on a certain demand structure in the past, accurate selling information is obtained and the appropriate amount of inventory is obtained in real time. There is a need for technology that seeks and responds and adjusts instantly to do so. The present inventors have also proposed an efficient approach as a technique for realizing an on-demand environment for this old and new problem in the IC tag era (see Patent Document 1 and Non-Patent Document 1). ).
JP 2006-344186 A Masayuki Matsui, Hiroki Uchiyama, Hiromi Fujikawa, “Managing inventory fluctuations in on-demand SCM using flow diagram”, Journal of Japan Society for Management Engineering, Vol.56, No.2, pp.139-145 (2005)

  However, in the technique using the flow diagram (Patent Document 1 and Non-Patent Document 1), it is necessary to adjust the parameters. Conventionally, this adjustment has been dependent on the human system, so that the demand structure changes. There was no response. In other words, it is necessary to search and determine various parameters in order to respond to fluctuations in the demand structure. It takes a lot of time and experience to get, and the fact is that it is not flexible.

  Therefore, the present invention solves the above-mentioned problems, and in supply chain management, parameter adjustment to cope with fluctuations in demand structure is automated, and parameter search time is reduced without requiring operator skill. An object of the present invention is to provide a flow number management system, method, and program capable of realizing real-time improvement in an on-demand environment while shortening and improving parameter accuracy and reducing human labor.

In order to solve the above problems, a flow number management system according to the present invention is a flow number management system including a flow number management logic processing means and a parameter automatic design means, wherein the flow management logic processing means includes: Inflow amount data and outflow amount data in the management target , and confirmed order data (preceding data) as expected demand information used for determination of the inflow amount data are acquired as data corresponding to time, and the next t + 1 firm orders amount X _{t + 1} and the error E _t-1 based on the runoff X _t ~ X ₁ to firm orders amount X _t and period t of period t, and the coefficient α to smooth the continuity of the quantity From this , the predicted value F _{t + 1} of the next runoff is F _{t + 1} = X _{t + 1} + α (O _t -X _t ) + (1-α) E _t-1 (where 0 <α ≤ 1 a prediction value calculating means for calculating a) obtains the stock quantity in each period based on runoff data and the inflow amount data with respect to the state of the stock quantity Multiplying the cost factor β according to the characteristics of the inventory condition to obtain the total penalty cost related to the inventory quantity, the movement reference inventory quantity calculating means for calculating the movement reference inventory quantity that minimizes the total penalty cost, and the calculated movement or standard inventory quantity whether a managed state, by plotting the moving reference inventory in the management view (e.g. cumulative sum control chart), the point that the plot is within a range stipulated by the management limit line The movement reference stock quantity management means for judging symbolically (for example, by the V mask method) according to whether or not , the predicted value F _{t + 1} , the outflow data for the current term and the previous term are made to correspond to the respective times. average multiplied by a weighting coefficient γ of weighting from their cumulative obtains the new flow rate, which is the difference between the cumulative inflow and cumulative inflow to this term to the next, the total penalty cost is minimized regarding runoff And input amount calculating means for calculating the inflow as input of the next, calculated been whether the input amount is in the managed state, accumulated on the vertical axis the flow, and the flow speed in the chart has the horizontal axis represents time Input amount management means for determining whether or not the cumulative inflow amount exceeds an upper limit control limit line indicating an upper limit of the cumulative inflow amount, and the parameter automatic design means is based on the flow management logic processing means . Prior to the arithmetic processing, a plurality of sets of the coefficients α, β and γ are automatically generated by random numbers , and coefficient generating means for selecting an arbitrary set under a predetermined condition from the plurality of sets ; While varying the coefficients α, β, and γ in the set selected by the coefficient generation means, the calculation process by the flow management logic processing means is executed for a plurality of generations, and the set of the coefficients of the changed generations is set. Calculate the fitness of the input amount obtained for each generation based on the above, compare the fitness between generations, store and hold the best set, and the number of generations stored or held reaches a predetermined number or is calculated And a calculation control unit that repeats selection of the set and calculation of the fitness until the fitness reaches a predetermined range .

The flow number management program according to the present invention is a flow number management program for causing a computer to function as a flow number management logic processing means and a parameter automatic design means in order to manage a flow number in a management target. The management logic processing means obtains inflow data and outflow data in the management target , and confirmed order data (preceding data) as expected demand information used for determination of the inflow data as data corresponding to time. , the error E _t-1 based on the runoff X _t ~ X ₁ to firm orders amount X _t and t-life of firm order quantity X _{t + 1} and t-life of the next t + 1, and the continuity of the quantity and a coefficient of smoothing alpha, the predicted value F _{t + 1} of the next runoff _{F t + 1 = X t +} 1 + α (O t -X t) + (1-α) E t-1 ( where 0 <α ≦ 1) and the predicted value calculating means for calculating a result, each period based on runoff data and the inflow data Calculate the inventory amount, multiply the inventory status by the cost factor β according to the characteristics of each inventory status to determine the total penalty cost for the inventory amount, and calculate the movement-based inventory amount that minimizes this total penalty cost and moving the reference stock amount calculation means, whether said calculated displacement reference stock amount is in the managed state, plotted in the control chart of the moving reference stock amount (e.g. cumulative sum control chart), the point that the plot Moving reference stock quantity management means that determines symbolically (for example, by the V mask method) depending on whether or not is within the range defined by the control limit line, the predicted value F _{t + 1} , and the outflows for the current term and the previous term Multiply each value by the weighted average weighting coefficient γ corresponding to the time, and calculate the new inflow that is the difference between the accumulated inflow until the next period and the accumulated inflow until this period. to determine the amount, runoff Total and input amount calculating means penalty cost is calculated new inflow amount becomes minimum as input of the next, the input amount, which is calculated as to whether or not a managed state related, accumulated ordinate flow, and the horizontal axis The automatic parameter design means comprises: an input amount management means for determining whether or not the cumulative inflow amount exceeds an upper limit control limit line indicating an upper limit of the cumulative inflow amount , said prior to arithmetic processing by the fluidized management logic processing unit, the coefficient alpha, it is automatically generated by the random number a plurality of sets of β and gamma, from a set of the plurality of predetermined conditions under any a coefficient generating means for selecting a set, the coefficients in the set is selected by said coefficient generating means alpha, while varying the β and gamma, the arithmetic processing by the flow management logic processing means multiple generations And calculate the fitness of the input amount obtained for each generation based on the coefficient set of each generation changed, compare the fitness of the generations, and store and hold the best set. Calculation control means for repeating the selection of the set and the calculation of the fitness until the number of generations reaches a predetermined number or the calculated fitness converges to a predetermined range .

  According to the present invention, in supply chain management, parameter adjustment to cope with fluctuations in demand structure is automated, parameter search time is shortened without requiring operator skill, parameter accuracy is improved, and human labor is required. It is possible to improve the real-time property in the on-demand environment while reducing the amount of power consumption.

  Flow number management is originally considered for the production process (see reference 1) and is static. There is an inflow / outflow relationship at each stage of the supply chain, and the same flow number problem occurs.

  The flow number management problem in the SCM can be rewritten as follows in the flow number management formula (see Document 1). Each stage of the supply chain has a standard inventory amount, and supplies (inflows) according to the input amount in the current period expressed by equation (1), and manages the inventory at each stage.

Input amount for the current period = Outflow amount for the current period + Standard inventory volume-Inventory volume for the previous period (1)
In the flow number management system in the present embodiment, the concept of foresight adaptive control is further introduced, the reference stock quantity is made variable, and the formula (1) is developed to the next input quantity as follows.

Next input amount = Next outflow amount + Movement base stock amount-Current stock amount (2)
At that time, the flow number management problem is how to manage the movement reference stock as a movement indicator to obtain the next input amount and minimize the flow number (inventory). The basic logic for this is to move the movement base stock amount closer to the current stock amount and make the next outflow amount the next input amount. For this purpose, the “newspaper problem” (see Reference 4 and Reference 5) is applied to the development of the movement reference stock quantity and the next input quantity. Take the decision).

  In the present embodiment, the movement reference stock amount (indicator) set by the newspaper seller problem on the flow number chart (refer to documents 2 and 3) is used as the control limit line (cumulative sum control chart (refer to document 6) of the control chart method. ) V-mask method and flow number diagram control limit line)) to check continuously and take action (determination of input amount).

  Hereinafter, embodiments of the present invention will be described in detail with reference to FIGS. Throughout the drawings, the same or equivalent parts and components are denoted by the same or equivalent reference numerals, and the description thereof is omitted or simplified.

(Overall configuration of fluid number management system)
FIG. 1 shows a functional configuration example of the flow number management system according to the present embodiment, and FIG. 2 shows a hardware configuration example of the flow number management system shown in FIG. FIG. 3 shows a processing procedure example (basic logic example) of the flow number management method by the flow number management system.

  As shown in FIG. 1, the flow number management system is roughly composed of a flow number management logic processing means 10 and an automatic parameter design means. The flow number management logic processing means 10 includes a flow number chart generation means 11, a predicted value calculation means 12, a movement reference stock quantity calculation means 13, a movement reference stock quantity management means 14, an input quantity calculation means 15, an input quantity management means 16, Functions such as an input amount improving means 17, an input amount determining means 18, and a profit making means 19 are provided. The parameter automatic design unit includes a calculation control unit 20, a coefficient variation unit 25, and a coefficient generation unit 26.

  The predicted value calculation means 12 acquires the inflow amount data, the outflow amount data, and the confirmed order data (preceding data) in the management target as data corresponding to the time, and determines the next determined order amount, the current determined order amount, and the current term The predicted value of the next outflow is calculated from the outflow and the coefficient α that smoothes the continuity of each quantity. The movement reference inventory quantity calculation means 13 obtains the inventory quantity for each period based on the inflow quantity data and the outflow quantity data, and multiplies the inventory quantity state by a cost coefficient β corresponding to the characteristics of each inventory condition. The total penalty cost is calculated, and the movement base inventory amount that minimizes the total penalty cost is calculated.

  The movement reference inventory quantity management means 14 symbolically indicates whether the movement reference inventory quantity calculated by the movement reference inventory quantity calculation means 13 is in a managed state (for example, a cumulative sum management chart) (for example, the V mask method). Judge). The input amount calculation means 15 multiplies each of these values by the weighted average weighting coefficient γ corresponding to the timing to the predicted value and the outflow amount data, and the total penalty cost related to the outflow amount is obtained from their accumulation. Calculate the minimum amount of the next runoff, and calculate the next runoff as the input for the next term.

  The input amount management means 16 determines whether or not the calculated input amount is in the management state based on the control limit line of the flow number chart. The input amount improving unit 17 improves the accumulated input amount so that it is below the control limit line based on the determination result by the movement reference stock amount management unit or the determination result by the input amount management unit.

  The coefficient generation means 26 of the parameter automatic design means automatically generates a plurality of sets of coefficients α, β, and γ with random numbers prior to the arithmetic processing by each means of the flow number management logic processing unit 10, It is a module that is input to the flow number management logic processing means 10, and also has a filtering function for selecting an arbitrary set from a plurality of sets under a predetermined condition. The filtering by the coefficient generation means 26 uses “Theory A” described later.

  The arithmetic control means 20 is a module that executes the arithmetic processing by the flow number management logic processing means 10 over a plurality of generations while changing the coefficients α, β, and γ in the set selected by the coefficient generating means 26. Specifically, the arithmetic control means 20 is obtained based on the set of changed coefficients for each generation by changing the coefficients α, β and γ by the coefficient changing means 25 and calling the flow number management logic processing means 26. The fitness of the input amount for each generation is calculated, the fitness of the generations is compared by the comparison means 23, and the best set is stored and held in the storage holding means 24 based on the comparison result. Then, the arithmetic control 20 repeats the selection of the set and the calculation of the fitness until the stored number of generations reaches a predetermined number or the calculated fitness converges to a predetermined range.

  In this embodiment, the coefficient variation means 25 varies the coefficients α, β, and γ in the set by a genetic algorithm shown in FIGS. Specifically, initial arrangement (here, generation of a coefficient set by the coefficient generation means 26) and calculation of an objective function (here, calculation of fitness by the flow number management logic processing means 10) are performed (FIG. 20). In step S301 and S302), roulette selection (S303) and crossover / mutation processing (S304) are performed on these, and the above steps S302 to S304 are repeated by loop processing until the processing result satisfies the end determination (S305).

  Here, roulette selection refers to the probability of being selected according to the distribution of fitness obtained by the flow number management logic when selecting an arbitrary set from among a set of coefficients generated by random numbers. In this embodiment, as shown in FIG. 21 (a), the area of roulette (selection is selected according to the fitness level so that a good fitness value can be selected with a high probability. The random selection is performed with a weight that increases the probability of being selected.

  Crossover is a process of changing the values of two data from their relative relationship. For example, when there are values a1 and a2 as shown in FIG. The coordinate transformation is performed so as to move to a1 ′ and a2 ′ in accordance with the vector. The magnitude of this vector can also be set at random. Depending on the magnitude, a1 and a2 may approach each other, be switched, or leave.

  Mutation means an operation of replacing a gene on a chromosome with another gene with a certain establishment. Specifically, data selected at random from a data string as shown in FIG. Change an element with a certain probability.

  As shown in FIG. 2, the flow number management system is realized by a computer system such as a PC, and includes an input unit 31, a control unit 32, a display unit 33, an external interface unit 34, a storage unit 35, and the like. . The input unit 31 is input means such as a keyboard, a pointing device, and a disk drive, for example, and is used for data input, execution order of the application program 41, etc., for the inflow / outflow data 21 and the confirmed order data 22. The display unit 33 is a display unit such as a display or a printer, and is used to display an execution result of the application program 41, an intermediate processing result, or the like. The external interface unit 34 provides an interface for connecting to other devices and systems. The storage unit 35 stores various application programs 41, application data 42, and the like. A program for realizing the functions illustrated in FIG. 1 in the computer system or a program for causing the computer system to execute the processes illustrated in FIG. 3 is stored in the storage unit 35 as the application program 41 and controlled. This can be executed by the unit 32.

(Flow number management logic processing)
First, the process by the flow number management logic processing means 10 described above will be described. The flow number management logic processing means 10 first executes initial values of numerical values (number of terms r, coefficients β ₁ , β ₂ , β ₃ , number of periods n, counter i, etc.) used in the logic when executing the processing shown in FIG. A setting process is executed (step S101).

[Flow number chart generation]
As shown in FIG. 4, for example, the inflow amount, outflow amount, and inventory amount of the warehouse to be managed in the t period are respectively the amount (I _t ) of product A flowing into the warehouse (from the factory) and the product A from the warehouse. Let the amount (O _t ) flowed out (to the market) and the amount of product A stored in the warehouse (L _t ). The confirmed order quantity (X _t ) is demand information for the product A.

  As illustrated in FIG. 5, the flow number chart is a graph in which an inflow accumulation line and an outflow accumulation line are drawn on a plane having a cumulative flow rate on the vertical axis and time on the horizontal axis. When the area surrounded by the two cumulative lines is viewed vertically, the amount of the flow target (Product A) staying in the management target (warehouse) at that time, that is, the number of flows (inventory amount) can be read and viewed horizontally. And the time (lead time) that the target remains within the management target can be read.

  The flow number chart generating means 11 inputs the inflow / outflow data 21 and the confirmed order data 22 (step S102), and generates the flow number chart data of the flow target (product A) (step S103). An example of the flow number chart data generated by the flow number chart generation means 11 is shown in FIG. Each data of inflow, outflow, and firm order is for 20 weeks, and firm order is expected demand information (forecast). In the example shown in FIG. 6, there are two large misalignments, causing a problem.

  For the sake of simplicity, it is assumed that the outflow (state) of the flow target (product A) cannot be controlled, and it is considered to minimize the number of flows / the number of days of ownership (control items) by controlling the inflow. Although the final order quantity can be used to determine the inflow volume, there is a large deviation from the actual actual demand volume (inflow volume) for short cycle products.

[Calculation of predicted value]
The predicted value calculation means 12 uses the flow number chart data generated by the flow number chart generation means 11 to set the final order quantity so that the value of the final order data 22 becomes a predicted value according to the actual demand. Processing is performed (step S104).

The predicted value F _{t + 1} in the t period when the lead time is 1 week (hereinafter, LT ₁ week) can be calculated by Equation (3). Note that LT ₁ week represents the time from when the product A is ordered until it reaches the warehouse.

F _{t + 1} = X _{t + 1} + α (O _t -X _t ) + (1-α) E _t-1 (However, 0 <α ≦ 1) (3)
However, _Et is an error in the t period, and is obtained from Equation (4).

E _t = α (O _t -X _t ) + α (1-α) (O _t-1 -X _t-1 )
+ α (1-α) ² (O _t-2 -X _t-2 ) +… + α (1-α) ^t-1 (O ₁ -X ₁ ) (4)
On the other hand, the predicted value F _{t + n} when the lead time is n weeks can be calculated by equation (5).

F _{t + n} = X _{t + n} + α (O _t -X _t ) + (1-α) E _t-1 (where 0 <α ≦ 1) (5)
Here, each α is a coefficient for smoothing the continuity of the next fixed order quantity, the current fixed order quantity, and the current outflow quantity, and is automatically generated by coefficient generation means described later.

  Moreover, the estimated value in each α value is evaluated based on the error area illustrated in FIG. The error area refers to the area between the accumulated predicted value and the accumulated outflow amount.

  The predicted value calculated by the predicted value calculating unit 12 is sent to the movement reference stock amount calculating unit 13 and the input amount calculating unit 15 together with the flow number chart data.

[Calculation of moving base inventory]
To demand accommodate on-demand, the movement reference stock amount calculation means 13, a reference inventory quantity N and N _t and the variable of as shown in FIG. 8 as an indicator, determined by applying the newspaper salesman problem (step S105).

  The newspaper seller problem is the question of “how many newspapers to sell every day at a newspaper stand”. If there are too many unsold newspapers, you will lose money, and if you buy too little, you will lose money. Since daily sales also fluctuate stochastically, the optimal number of purchases is obtained in the form of maximizing profit by minimizing opportunity loss, which is a combination of unsold loss and out-of-stock loss.

Penalties will be introduced when applying the newspaper vendor problem to inventory management. The total penalty cost C (N _t ) is given by the following equation (6), where L _{t is} the inventory quantity in the t period and Nt is the standard inventory quantity in the t period.

C (N _t ) = β ₁ N _t + β ₂ (N _t -L _t ) ⁺ + β ₃ (L _t -N _t ) ⁺ (6)
However, β _{1 to} β ₃ are cost coefficients for calculating the penalty for the inventory quantity state in each period, and are determined according to the characteristics of each inventory condition. In this embodiment, the coefficient generation means described later automatically Generated. Specifically, the cost factor of the beta ₁ the reference stock amount N _t (optimum inventory), the cost factor beta ₂ is subjected when below the reference stock amount N _t (too little inventory), beta ₃ the reference stock amount N _t This is the cost factor (excess inventory) to be applied when exceeding.

The (N _t −L _t ) is an amount that is less than the reference stock amount N _t , and (L _t −N _t ) is an amount that is greater than the reference stock amount N _t . Further, + on the right shoulder indicates that a negative value in the equation is zero, and (a) ⁺ = max (a, 0), for example.

In order to obtain the optimum reference stock value N _t *, the expected value of the equation (6) only needs to be minimized. Therefore, by arbitrarily determining the coefficients β ₁ , β ₂ , β ₃ , Σf (L _t ) Can be determined. Here, considered .SIGMA.f the (L _t) is the cumulative frequency of the flow number, .SIGMA.f when standard inventory value N _t * is minimized (L _{t) is, (β 3 -β 1) /} (β ₂ + β ₃ ).

  Here, a method is considered in which the movement reference stock value for each week is obtained by referring to only past data divided by several weeks.

FIG. 9 shows an example of the result of obtaining the moving reference stock quantity based only on the past data with coefficients β ₁ = 5, β ₂ = 250, β ₃ = 200, and the number of periods of the moving average is 3 weeks (r = 3). is doing. Of course, forecast inventory data is also available.

A method for actually calculating the movement reference stock quantity from the cumulative stock quantity using Σf (L _t ) is shown below. When the newspaper seller problem is applied to the inventory quantity, the reference inventory quantity is determined from the accumulation of the inventory quantity illustrated in FIG. In this case, the reference stock quantity is determined by the following equation (7).

Σf (Lt) = (β ₃ -β ₁ ) / (β ₂ + β ₃ ) (7)
Assuming that the penalty) is increased to apply when exceeds the β ₃ standard inventory quantity, we try to increase the standard amount of inventory as stock amount does not exceed the standard inventory amount. If β _l is increased, the value on the right side increases. As β ₃ increases, the standard inventory must also increase, so the right side is taken from the bottom of the distribution and is given by Equation (7).

本実施の形態においては、新聞売り子問題を本在庫管理に応用するにあたり、ペナルティー費用を導入した式(6)を用いたが、評価式はこれに限られるものではない。すなわち、次に示す代替式を適用しても本実施の形態を実施することができる。 By the way, in the newspaper seller problem of the above equation (7), there has been a problem of how to determine the penalty coefficient β effectively. On the other hand, generally the demand pattern of goods changes according to a period (for example, every season). Therefore, a method (logic) for automatically managing the inventory quantity by determining the reference inventory quantity N from the penalty coefficient β following the period variation is proposed.

In the present embodiment, when the newspaper seller problem is applied to the present inventory management, the formula (6) in which a penalty cost is introduced is used, but the evaluation formula is not limited to this. That is, this embodiment can be implemented even when the following alternative formula is applied.

  In other words, from the Little's formula and flow number analysis, there is a corresponding relationship between the world of quantity of inventory (L) and the time of required time (W), so this management method (logic) is developed in W It is possible. Therefore, the present embodiment can be implemented by replacing the formula (6) with the following formula (6-1).

C (T _t ) = β ₁ T _t + β ₂ (T _t -W _t ) ⁺ + β ₃ (W _t -T _t ) ⁺ (6-1)
This is considered to be effective particularly when there is a process that causes a delay, such as production.

[Transfer-based inventory management]
As a method for managing inventory fluctuations in real time, the V mask method for cumulative sum control charts used in the field of quality control is effective. A cumulative sum control chart that excels in the ability to detect shift-like changes is used to confirm the presence or absence of a management state. If the hit point is within the range of a line called a V mask, the process is in a managed state, and if it is outside the range of the V mask, the process is not in a managed state. If it is determined that it is not in the management state (that is, the current demand fluctuation is within the planned range), some action must be taken. Demand fluctuations may show large values instantaneously, but basically appear as trend fluctuations. It may continue to rise above the average, or a decrease below the average may occur continuously. The cumulative sum chart used in quality control is also effective for recognizing it. This determines whether or not the blur from the target value is significant by applying a V-shaped mask from the trend line.

Movement reference inventory quantity management unit 14, the movement standard inventory quantity N _t set by moving the reference stock amount calculation means 13, to check with the V-mask method cumulative sum control chart. The movement reference inventory management means 13 provides a V-shaped control limit line (V cut) in the cumulative sum control chart (step S106), and whether the blur from the V mask formed by this V cut is significant. Determine whether or not.

  The movement reference stock quantity management means 14 determines that it is in the management if all the past points plotted in the cumulative sum management chart are within the V mask, and it is out of the management if it does not (Step S107).

If the result of determination in step S107 is that it is within management, the process proceeds to step S111. On the other hand, if it is determined that the process is out of control, if the current process is the Kth process (where K is a predetermined threshold that can be changed as appropriate), the values of the coefficients β ₁ , β ₂ , and β ₃ are changed. Then, the process returns to step S104.

In order to adapt to demand, obtain indicators that change the coefficients β ₁ , β ₂ , and β ₃ while taking into account fluctuations in the forecast values for the next week, current inventory values, and previous spills (past spills). It was. FIG. 11 (mask of the third period of the actual value) and FIG. 12 (mask of the thirteenth period of the indicator) illustrate the presence / absence of the actual value and the indicator management state.

  As illustrated in FIGS. 11 and 12, the indicator shows an improvement effect in the first half portion, but a large deviation from the reference is seen in the second half portion. In order to eliminate this gap in the latter half, improvement using the control limit line as shown below is performed (details will be described later).

[Calculation of input amount]
The input amount calculation means 15 calculates the input amount based on the flow number chart data and the predicted value calculated by the predicted value calculation means 12 (step S111).

Here, using the predicted value F _{t + 1} in the t period, the newspaper seller problem previously applied to the inventory quantity is also applied to the outflow quantity, and a new indicator of the next period called the input quantity is determined. First, the coefficients β ₁ = 5, β ₂ = 200, β ₃ = 200, and the cumulative number of outflows up to the t period in the t period and the cumulative number of outflows up to the (t + 1) period in the (t + 1) period Thus, the predicted value calculated by the predicted value calculation means 12 is entered in the (t + 2) period. The moving average obtained there is taken as the average cumulative inflow until the (t + 2) period.

  Next, put the cumulative number of outflows up to (t + 1) period in (t + 1) period and the total number of outflows up to (t + 2) period in (t + 2) period, then (t + 3) period Is put the predicted value calculated by the predicted value calculation means 12. The moving average found there was taken as the average cumulative inflow in the (t + 3) period, and the average cumulative inflow in the (t + 2) period minus the new inflow (input) in the (t + 3) period. To do.

  That is, in the present embodiment, when calculating the input amount in the input amount calculation means 15, the predicted value and the outflow amount data are multiplied by a weighted average weight coefficient γ corresponding to the respective times, respectively. From the accumulation, the next runoff that minimizes the total penalty cost for the runoff is obtained, and the calculated next runoff is calculated as the input for the next.

Specifically, for the function F (t) that calculates the cumulative number of outflows until the next period,
Next term F (t) = predicted value * γ ₁ + number of outflows this term * γ ₂ + number of outflows in the previous term * γ ₃ (7-1)
Is used to smooth the accumulated amount using the predicted value so that it does not deviate significantly from the past weighted average. Here, each γ is also automatically generated by a coefficient generation means described later.

  The calculation result of the input amount obtained in this way is illustrated in FIG. However, since “~ 1 period” and “~ 2 period” include periods without outflow data, the forecast value was applied and used as the new inflow.

  Further, when applying the newspaper seller problem regarding the input amount, the following equation (8) is used to determine the calculated input amount from the new inflow amount and FIG.

Σf (L _t ) = 1- (β ₃ -β ₁ ) / (β ₂ + β ₃ ) (8)
If β ₃ (penalty to be applied when it exceeds the reference stock amount) is increased, the stock amount is reduced so that the stock amount does not exceed the reference stock amount, that is, the new inflow amount to be obtained is reduced. If β ₃ is increased, the value on the right side is increased. As β ₃ increases, the new inflow must decrease, so the right side is taken from the top of the distribution and is given by equation (8).

[Input amount management and improvement]
In order to manage the next input amount, the input amount management means 16 newly sets an upper control limit line in the flow number chart (step S112). The upper control limit line indicates the upper limit of the cumulative inflow amount, and has a role of preventing an excessive inflow by setting an upper limit of the inflow amount.

  FIG. 15 shows an example in which the upper limit line for the next period with a lead time of 1 week = cumulative outflow amount for this period + reference stock quantity + predicted value for the next period is set in this example.

  When the cumulative inflow of the next period exceeds the upper management limit line as illustrated in FIG. 16A, the input amount improving means 17 increases the upper management limit of the next period of inflow as illustrated in FIG. 16B. Limit to below the line and use it as the new cumulative inflow. If the cumulative inflow for the next period does not exceed the upper control limit line, that inflow is adopted.

  When the reference stock value is within the control limit as a result of the determination by the movement reference stock management 14, the input amount determination unit 18 determines that the next cumulative inflow amount is within the control limit as a result of the determination by the input control unit 16. If it is, the input amount calculated by the input amount calculation means 15 is determined as the next input amount. As a result of determination by the movement reference stock quantity management 13, both or either when the reference stock value is out of the control limit and when the input amount management means 16 determines that the next cumulative inflow amount is out of the control limit When one is satisfied, the value improved by the input amount improving means 17 is determined as the next input amount (steps S113 to S114).

  FIG. 17 shows the improvement effect of the input amount (input amount 1) obtained by the newspaper seller problem by the input amount calculation means 15 and the improvement amount (input amount 2) improved within the control limit by the input amount improvement means 17. ing. Compared to FIG. 6, it can be seen that the improvement effect is large.

In the present embodiment, the indicators β ₁ , β ₂ , and β ₃ are obtained by changing the indicators β ₁ , β ₂ , and β ₃ while taking into consideration the predicted value for the next week, the current inventory value, and fluctuations in the previous spillage. FIG. 18 shows a comparison of fluctuations in inventory when the indicator is used.

[Profit]
The profit making means 19 verifies the effect of improving the input amount from the profit side (step S115).

If the ratio [capacity occupied by the inventory / capacity of the entire warehouse] in the period t is the operation rate ρ _t and the ratio [free capacity / capacity of the entire warehouse] is the idle rate 1−ρ _t , then the cost EC _t Is calculated by the following equation (9) (see Literature 7 and Literature 8).

EC _t = α ₁・ L _t + α ₃ + (α ₂ -α ₃ ) ρ _t (9)
However, α ₁ is an inventory storage cost coefficient (yen / piece), α ₂ is an operating cost coefficient (yen), and α ₃ is an idle cost coefficient (yen).

Further, the gain ER _t (yen) in the t period can be calculated by the following equation (10), where P (yen) is the gain when one product is distributed, and d _t is the supply and demand speed in the t period.

ER _t = P / d _t (10)
At this time, the profit EN _t (yen) in the t period can be calculated by the following equation (11).

EN _t = P / d _t- (α ₁・ L _t + α ₃ + (α ₂ -α ₃ ) ρ _t ) (11)
From the equation (11), the profit EN _t is a function of the operation rate ρ _t . If ρ _t * exists that maximizes profit EN _t , the optimal inventory is obtained and the pair strategy projection can be developed. Moreover, the real-time monitoring becomes possible.

  Returning to the flowchart of FIG. 3, when the processing of steps S104 to S115 is completed as described above, the number of periods n is compared with the counter i. When i> n, that is, the processes for all the periods are completed. Then, the processing result is output to the display unit 33 (step S118), and the processing is terminated. Conversely, if i ≦ n, that is, if an unprocessed period remains, 1 is added to the counter i (step S117), the process returns to step S104, and the process for the next period is started.

  As explained above, the newspaper seller problem is applied to the development of the movement reference stock quantity and the next input quantity, and the movement reference stock quantity and the next input quantity are respectively managed by the V mask method of the cumulative sum control chart and the upward management of the flow number chart. The limit line is continuously checked and an action (determination of input amount) is taken. Therefore, it is possible to finely set an appropriate level of inventory, and to minimize inventory storage costs, warehouse operation costs, and the like.

  In the case of the logic using the demand rate λ shown in equation (7-a) (equation (7-c)), it is possible to further obtain the effect of reducing the number of out-of-stock items and the total inventory amount. .

(Automatic parameter design process)
Next, automatic parameter design processing in the above-described flow number management logic will be described. In this parameter automatic design process, parameters (coefficients α, β and γ) used in the flow number management logic process are automatically generated, and the coefficients α, β are generated according to the result of the flow number management logic process using the parameters. And γ are changed, and the process is repeated to select a more optimal parameter. FIG. 19 is a flowchart showing the procedure of parameter automatic design processing.

  Prior to the calculation processing by the flow number management logic processing means 10, the coefficient generation means 26 automatically generates a plurality of sets of the coefficients α, β and γ with random numbers (S201). Here, 10 individuals are generated for each coefficient. In generating this coefficient, in the present embodiment, filtering based on theory A is performed, and loop processing is repeated until a predetermined number of coefficients conforming to theory A are generated (“N” in S202).

  Here, the theory A is “0.1 <α <1.0”, “0.01 <β <1.00”, “γ1 + γ2 + γ3 = 1, and γ1> for each coefficient, as shown in step S202 of FIG. This function gives a range of “0.33”.

  Next, the cumulative stock, out of stock, and fitness are calculated by the flow number management logic processing means 10 (S203), and an arbitrary set is selected from the set based on the fitness calculation result. In this embodiment, when selecting the set, the coefficient variation means 25 executes the arithmetic processing over a plurality of generations while varying the coefficient in the set using the genetic algorithm described above.

  That is, two roulettes are selected from the ten individuals based on the fitness obtained by the flow number management logic (S204). Take out every two individuals (pairs) selected by this roulette selection (S205), whether the random number x for each pair exceeds the crossover probability (the value of the two individuals is enough to obtain a sufficient effect by the crossover process) (S206), if exceeded ("Y" in step S206), proceed to the crossover processing (S207) of each coefficient, and if not ("N" in step S206), the crossover probability Step S205 is repeated until two individuals satisfying the condition are selected.

  After two individuals exceeding the crossover probability are selected in step S205, each coefficient (parameter) is crossed (S207), and whether or not the crossed coefficient satisfies the above theory A is determined by the filtering function of the coefficient generating means 26 Is determined by calling (S208). If the theory A is not satisfied (“N” in step S208), the crossover process in step S207 is performed again.

  The processes in steps S205 to S208 are repeated until 5 pairs satisfying the crossover end condition are selected (S209 and S210). After 5 pairs satisfying the condition are selected (“Y” in S210), the process is constant. Each parameter is subjected to mutation processing based on the mutation probability (probability that a mutation can occur) (S211). The coefficient generation means 26 determines whether the mutated coefficient satisfies the theory A (S212). When the coefficient generated by the mutation does not satisfy the theory A (“N” in step S212), the mutation process is performed again.

  After that, using the coefficient generated by the mutation process, the cumulative stock, out of stock and fitness are calculated again by the flow number management logic (S213), and the best parameter set is stored based on the obtained fitness. The holding means 24 stores and holds it.

  Specifically, for the fitness obtained in step S213, if there is a previous set of parameters (previous generation parameter set), a comparison is made as to whether there is a better fitness than the previous generation. (S214) If there is a good solution, the latest parameter set is stored. If not, the best solution of the previous generation is stored as it is, and the latest parameter set is discarded (S215).

  The above steps S204 to S215 are repeated until the end condition is satisfied (S216). This termination condition was repeated by repeating the selection of the set and the calculation of the fitness until the number of generations stored and held in the memory holding means 24 reached a predetermined number or until the calculated fitness converged to a predetermined range. After ("Y" in step S216), the result is displayed (S217).

  According to such parameter automatic design processing according to the present embodiment, α, β, and γ, which are parameters of flow number management projection, are optimized and algorithmized using a genetic algorithm, and until then, manual and full search have been performed. The work can be automated. In the present embodiment, the search range of the coefficients β and γ is set in increments of 0.01, and the number of generations can be arbitrarily set from 1 to 1000 generations, so that the accuracy of the solution can be improved.

  Therefore, according to the present embodiment, in the supply chain management and the like, the parameter adjustment for responding to the fluctuation of the demand structure is automated, the parameter search time is shortened and the parameter accuracy is improved without requiring operator's skill. Real-time performance in an on-demand environment can be improved while reducing human labor.

(Example of change)
As mentioned above, although embodiment of this invention was described in detail, this invention can be implemented in other various forms, without deviating from the mind or main characteristics.

  For example, in the present embodiment, the flow number management of parts and products in a warehouse or the like has been described as a model. However, the present invention is not limited to this, and can be widely applied to the management and analysis of flowing things. For example, the present invention may be applied to power supply / demand management or the like, or may be applied to water / food supply / demand management.

  In the parameter automatic design process described above, the genetic algorithm is used for parameter (coefficient) variation. However, the present invention is not limited to this, and filtering is performed while optimizing a set of randomly generated coefficients. Any algorithm that can be used can be employed.

  Further, in the symbolic inspection of the control limit by the flow number management logic processing means 10 described above, the so-called V mask method is adopted, but the present invention is not limited to this. For example, only the limit of β is inspected. A β limit method or the like can also be employed.

  Here, the β limit method means that in the control limit processing using the upper control limit line and the lower control limit line shown in FIG. 15, (predicted value / β) is used as the upper control limit line, and the lower control limit line is used. As (predicted value × β). In this β limit method, since only β is varied, the above-described control limit check in steps S106 and S107 may be omitted. Therefore, according to the β limit method, when the parameter set is changed, the values of α and γ are fixed, only the value of β is generated as a random number, and the process using the control limit line in the flow number management logic described above ( By checking only the value of β at the time of step S113), an optimal parameter set can be easily obtained.

  It should be noted that a function for selecting the control limit method to be used may be added to the system of the present embodiment according to the analysis target, such as switching between the above-described V mask method and β limit method by a user operation.

  As described above, the above-described embodiment is merely an example in all respects, and should not be interpreted in a limited manner. The scope of the present invention is indicated by the claims, and is not restricted by the text of the specification. Further, all modifications and changes belonging to the equivalent scope of the claims are within the scope of the present invention.

(References)
[Reference 1] Japan Business Engineering Association, “Production Management Terminology”, Japanese Standards Association, p.464, 2002 [Reference 2] Nakamura Zentaro, “Utilization of Flow Number Analysis in Work Improvement”, Japan Management Engineering Journal, Vol.36, No.2, pp.93-100, 1985 [reference 3] Jun Usuki, Masatoshi Kitaoka, Masayuki Matsui, “On the flow number management problem and gray theory model”, Bulletin of the University of Electro-Communications Vol.14, 1 No., pp.13-20, 2001 [reference 4] Yutaka Kato, Masanori Ozawa, “From the basics of OR to AHP to optimization” Jikkyo Publisher, 1988 [reference 5] Week, jK, “Optimizing Planned Lead times and Delivery Dates ”, 21st Annual conference Proceedings, APICS, pp.177-188, 1979
[Reference 6] Haruo Watanabe, Eiji Komine, Kenichi Muida, “Sequential sampling inspection and cumulative sum control chart”, Muroran Institute of Technology Bulletin 50, pp.71-82, 2000 [Reference 7] Japanese Patent Laid-Open No. 2002-49449 [Document 8] Japanese Patent Laid-Open No. 2002-197261

It is a schematic block diagram which shows the function structural example of the flow number management system by this embodiment. It is a schematic block diagram which shows the hardware structural example of the flow number management system by this embodiment. It is a follow chart which shows the process sequence example (basic logic example) of the fluid number management method by a fluid number management system. It is a figure which shows the inflow amount in the management object of this embodiment, the outflow amount, the inventory amount, and the fixed order amount. It is the schematic diagram which made the accumulated quantity of the inflow amount, and the accumulated quantity of the outflow amount into the flow number diagram. It is an example in which the cumulative amount of the inflow amount, the cumulative amount of the outflow amount, and the cumulative amount of the confirmed order amount in the management target of the present embodiment are converted into a flow diagram. It is an example of the error area for evaluating the value of the variable (alpha) used for calculation of a predicted value (next term outflow amount). It is a figure which shows the relationship between reference | standard stock quantity, real stock quantity, and penalty expense. It is a figure which illustrates the result of having obtained coefficients β1 = 5, β2 = 250, β3 = 200, the number of periods is three weeks, and the reference stock quantity based only on past data. It is a figure which illustrates the accumulated inventory quantity at the time of calculating a reference inventory quantity by applying the newspaper salesperson problem. It is a figure which shows the example which showed the management state of the actual stock quantity by V mask (V mask of the 3rd period) of a cumulative sum management chart. It is a figure which shows the example which showed the management state of the indicator with V mask (V mask of the 13th period) of a cumulative sum management chart. It is a figure which illustrates the result of having calculated coefficients β1 = 5, β2 = 250, β3 = 200, the number of periods being three weeks, and determining the input amount based only on past data. It is a figure which illustrates the accumulated outflow amount at the time of calculating the input amount by applying the newspaper seller problem. It is a figure for demonstrating the calculation method of the upper control limit line in a flow number chart. It is a figure for demonstrating the improvement method of the cumulative inflow amount by an upper management limit line. It is a figure which illustrates the improvement effect of the input amount improved by the upper management limit line. It is a figure which illustrates the fluctuation | variation comparison of an inventory amount when the indicator calculated | required considering the predicted value of the next period, the present inventory amount, and the fluctuation | variation of the outflow amount until now is used. It is a follow chart which shows the example of a process sequence of the parameter automatic design by a flow number management system. It is a flowchart which shows the process sequence of the genetic algorithm in parameter automatic design processing. It is explanatory drawing which shows the outline | summary of the genetic algorithm in a parameter automatic design process.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Flow number management logic processing means 11 ... Flow number chart generation means 12 ... Predicted value calculation means 13 ... Movement reference stock quantity calculation means 14 ... Movement reference stock quantity management means 15 ... Input quantity calculation means 16 ... Input quantity management means 17 ... Input amount improving means 18 ... Input amount determining means 19 ... Profiting means 20 ... Calculation control means 21 ... Inflow / outflow data 22 ... Final order data 23 ... Comparing means 24 ... Memory holding means 25 ... Coefficient variation means 26 ... Coefficient generation Means 31 ... Input unit 32 ... Control unit 33 ... Display unit 34 ... External interface unit 35 ... Storage unit 41 ... Application program 42 ... Application data

Claims (4)

A flow number management system comprising flow number management logic processing means and parameter automatic design means,
The flow management logic processing means includes:
Inflow amount data and outflow amount data in the management target, and confirmed order data (preceding data) as expected demand information used to determine the inflow amount data are acquired as data corresponding to time, and the next t + 1 firm orders amount X _{t + 1} and the error E _t-1 based on the runoff X _t ~ X ₁ to firm orders amount X _t and period t of t phase and the α coefficient smoothing the continuity of each of these quantities From the next forecast, F _{t + 1}
F _{t + 1} = X _{t + 1} + α (O _t -X _t ) + (1-α) E _t-1 (where 0 <α ≦ 1)
A prediction value calculating means for calculating a,
Based on the inflow volume data and the outflow volume data, the inventory quantity of each period is obtained, and the total penalty cost related to the inventory quantity is obtained by multiplying the inventory quantity state by a cost coefficient β corresponding to the characteristics of each inventory condition. A movement base inventory quantity calculation means for calculating a movement base inventory quantity that minimizes the total penalty cost;
Whether the calculated movement reference stock quantity is in a controlled state, plot the movement reference stock quantity in a control chart , and whether the plotted point is within the range defined by the control limit line A movement-based inventory management means to determine symbolically by
The forecasted value F _{t + 1} , current and previous period runoff data are multiplied by a weighted average weighting coefficient γ corresponding to each value, and the cumulative inflow from the accumulation to the next period and determine the new flow rate, which is the difference between the cumulative inflow to this term, the input amount calculating means for calculating a new flow rate the total penalty cost for runoff is minimized as input of the next,
Whether or not the calculated input amount is in a managed state, the cumulative inflow amount indicates the upper limit of the cumulative inflow amount in the flow number chart in which the vertical axis indicates the cumulative flow rate and the horizontal axis indicates the time. Input amount management means for judging whether or not the upper limit control limit line is exceeded
The parameter automatic design means comprises
Prior to the arithmetic processing by the flow management logic processing means, a plurality of sets of the coefficients α, β and γ are automatically generated by random numbers, and an arbitrary set is selected from the plurality of sets under a predetermined condition. Coefficient generation means to select ;
While varying the coefficients α, β and γ in the set selected by the coefficient generation means, the calculation process by the flow management logic processing means is executed for a plurality of generations, and based on the changed coefficient set for each generation The fitness of the input amount obtained for each generation is calculated, the fitness between generations is compared, the best set is stored and retained, and the number of generations stored or retained reaches a predetermined number or is calculated A flow number management system comprising: an arithmetic control unit that repeats selection of the set and calculation of fitness until the degree converges to a predetermined range . The arithmetic control means includes
A filtering function for selecting the set of coefficients under a predetermined condition;
A function for switching the coefficient to be changed and the control limit line;
Flow number management system according to claim 1, characterized in that it comprises a. A flow number management program for causing a computer to function as a flow number management logic processing means and a parameter automatic design means in order to manage the flow number in the management target ,
The flow management logic processing means includes:
Inflow amount data and outflow amount data in the management target, and confirmed order data (preceding data) as expected demand information used to determine the inflow amount data are acquired as data corresponding to time, and the next t + 1 firm orders amount X _{t + 1} and the error E _t-1 based on the runoff X _t ~ X ₁ to firm orders amount X _t and period t of t phase and the α coefficient smoothing the continuity of each of these quantities From the next forecast, F _{t + 1}
F _{t + 1} = X _{t + 1} + α (O _t -X _t ) + (1-α) E _t-1 (where 0 <α ≦ 1)
A prediction value calculating means for calculating a,
Based on the inflow volume data and the outflow volume data, the inventory quantity of each period is obtained, and the total penalty cost related to the inventory quantity is obtained by multiplying the inventory quantity state by a cost coefficient β corresponding to the characteristics of each inventory condition. A movement base inventory quantity calculation means for calculating a movement base inventory quantity that minimizes the total penalty cost;
Whether the calculated movement reference stock quantity is in a controlled state, plot the movement reference stock quantity in a control chart , and whether the plotted point is within the range defined by the control limit line A movement-based inventory management means to determine symbolically by
The forecasted value F _{t + 1} , current and previous period runoff data are multiplied by a weighted average weighting coefficient γ corresponding to each value, and the cumulative inflow from the accumulation to the next period and determine the new flow rate, which is the difference between the cumulative inflow to this term, the input amount calculating means for calculating a new flow rate the total penalty cost for runoff is minimized as input of the next,
Whether or not the calculated input amount is in a managed state, the cumulative inflow amount indicates the upper limit of the cumulative inflow amount in the flow number chart in which the vertical axis indicates the cumulative flow rate and the horizontal axis indicates the time. Input amount management means for judging whether or not the upper limit control limit line is exceeded
The parameter automatic design means comprises
Prior to the arithmetic processing by the flow management logic processing means, a plurality of sets of the coefficients α, β and γ are automatically generated by random numbers, and an arbitrary set is selected from the plurality of sets under a predetermined condition. Coefficient generation means to select ;
While varying the coefficients α, β and γ in the set selected by the coefficient generation means, the calculation process by the flow management logic processing means is executed for a plurality of generations, and based on the changed coefficient set for each generation The fitness of the input amount obtained for each generation is calculated, the fitness between generations is compared, the best set is stored and retained, and the number of generations stored or retained reaches a predetermined number or is calculated A computer-readable flow number management program, comprising: an operation control unit that repeats selection of the set and calculation of fitness until the degree converges to a predetermined range . In the flow number management logic process, the predicted value calculation means uses the inflow data and outflow data in the management target and the final order data (preceding data) as expected demand information used for the determination of the inflow data. was obtained as the corresponding data, the error E _t-1 based on the runoff X _t ~ X ₁ to firm orders amount X _t and t-life of firm order quantity X _{t + 1} and t-life of the next t + 1, and their the continuity of each quantity and a coefficient α for smoothing the prediction value F _{t + 1} of the next runoff
F _{t + 1} = X _{t + 1} + α (O _t -X _t ) + (1-α) E _t-1 (where 0 <α ≦ 1)
A step of calculating by
In the flow number management logic processing, the movement reference inventory quantity calculation means obtains the inventory quantity for each period based on the inflow quantity data and the spill quantity data, and the cost coefficient corresponding to the characteristics of each inventory condition with respect to the status of the inventory quantity multiplying β to determine the total penalty cost for inventory, and calculating a migration-based inventory that minimizes the total penalty cost;
In the flow number management logic process, the movement reference inventory quantity management means plots the movement reference inventory quantity in the control chart to determine whether or not the calculated movement reference inventory quantity is in a management state. A step of determining symbolically by whether or not it is within a range defined by a control limit line ;
In the flow number management logic processing, the input amount calculation means multiplies the predicted value F _{t + 1} and the outflow data for the current term and the previous term by a weighted average weighting factor γ corresponding to the time of each of these values. The new inflow, which is the difference between the cumulative inflow until the next term and the cumulative inflow until the current term, is calculated from these accumulations , and the new inflow that minimizes the total penalty cost for the outflow is calculated as the input for the next term. And steps to
In the flow number management logic processing, the input amount management means indicates whether or not the calculated input amount is in a managed state, in the flow number chart in which the vertical axis indicates the cumulative flow rate and the horizontal axis indicates the time. Determining whether or not the inflow amount exceeds an upper limit control limit line indicating an upper limit of the cumulative inflow amount ; and
Prior to the calculation process in the flow management logic process , the coefficient generation means automatically generates a plurality of sets of the coefficients α, β, and γ with random numbers, and arbitrarily selects the plurality of sets under a predetermined condition. the method comprising the steps of: selecting a set,
The calculation control means executes the calculation process by the flow management logic process over a plurality of generations while changing the coefficients α, β, and γ in the set selected by the coefficient generation means, and the changed coefficients of each generation The fitness of the input amount obtained for each generation obtained based on the set of the above is calculated, the fitness of the generations is compared, the best set is stored and held, and the number of generations stored and held reaches a predetermined number, or A flow number management method, comprising: causing a computer to repeat the step of selecting the set and calculating the fitness until the calculated fitness converges to a predetermined range .

Images