Determination system

Abstract

To provide a determination system for highly accurately determining the propriety of disposal of a facility or a mold.SOLUTION: The determination system 100 determines whether or not a facility or a mold for manufacturing a supplemental part for an automobile can be discarded. The determination system 100 includes means for calculating a probability of occurrence of an unexpected demand from a product number assigned to a supplemental part by using a mathematical expression included in a learned model, and means for comparing the probability with a threshold value included in the learned model. When the probability is smaller than a threshold, information indicating that the equipment or the mold can be discarded is output, and when the probability is equal to or smaller than the threshold, information indicating that the equipment or the mold is maintained is output.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a determination system, and more particularly to a determination system for determining whether or not continuous production of automotive spare parts is necessary. [Background technology]

[0002] Patent Document 1 discloses a parts management system that determines the discontinuation of supply of parts in a form defined for a specific sales unit in response to the approaching end of sales period. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2005-108021 Summary of the Invention [Problem to be solved by the invention]

[0004] The conventional technology does not disclose specific steps or functions for determining whether or not equipment or molds can be disposed of, which makes it difficult to determine whether or not equipment or molds can be disposed of using computer resources.

[0005] The present disclosure has been made to solve such problems, and provides a determination system that determines with high accuracy whether or not equipment or molds can be disposed of. [Means for solving the problem]

[0006] The judgment system according to the present disclosure is a judgment system for determining whether or not to scrap equipment or molds used to manufacture automotive spare parts, and includes: means for calculating, using a formula included in a trained model, the probability of unexpected demand occurring, representing a situation in which demand for the spare part is greater than expected, from the product number attached to the spare part; and means for comparing the probability with a threshold probability value included in the trained model; if the probability is smaller than the threshold probability value, information indicating that the equipment or mold can be scrapped is output; and if the probability is equal to or greater than the threshold probability value, information indicating that the equipment or mold will be maintained is output; the formula included in the trained model includes explanatory variables into which feature values ​​obtained from the product number are substituted, and coefficients for weighting each explanatory variable; the coefficients are trained and calculated by logistic regression using training data; and the threshold probability value is determined so as to minimize the profit given for each spare part whose prediction by the formula in the training data is correct, and the loss amount based on the loss given for each spare part whose prediction by the formula in the training data is incorrect. [Effects of the Invention]

[0007] According to the present disclosure, it is possible to provide a determination system that determines with high accuracy whether or not equipment or molds can be disposed of. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a block diagram illustrating a configuration of a determination system according to a first embodiment. [Figure 2] 4 is a flowchart illustrating an operation of the determination system according to the first embodiment. [Figure 3] FIG. 2 is a diagram for explaining an example of training data according to the first embodiment. [Figure 4] FIG. 1 is a diagram for explaining an example of a mathematical formula included in a trained model according to the first embodiment. [Figure 5] FIG. 4 is a diagram for explaining a method for setting a threshold probability value according to the first embodiment. [Figure 6]FIG. 4 is a diagram for explaining a method for setting a threshold probability value according to the first embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, the present embodiment will be described with reference to the drawings. However, the present invention is not limited to the following embodiment. In addition, the following description and drawings have been simplified as appropriate for clarity of explanation.

[0010] Embodiment 1 FIG. 1 is a block diagram showing the configuration of a determination system 100 according to the first embodiment. The determination system 100 is, for example, a computer such as a server. The determination system 100 may be realized by, for example, cloud computing. In this case, multiple components of the determination system 100, which will be described later, may be realized by physically different computers. The determination system 100 determines whether or not equipment or a mold (e.g., a die) for manufacturing automotive spare parts can be discarded.

[0011] The determination system 100 includes a processor such as a CPU (Central Processing Unit) not shown. The determination system 100 may include multiple processors. The determination system 100 also includes a storage device such as a memory or a hard disk not shown. The storage device has a function for storing a control program or an arithmetic program to be executed by the processor. The memory stores one or more instructions.

[0012] The determination system 100 may include a user interface (UI) (not shown). The user interface includes input devices such as a keyboard, a touch panel, or a mouse, and output devices such as a display or a speaker. The user interface accepts data input operations by a user (operator) and outputs information to the user.

[0013] The determination system 100 includes, as its components, a calculation unit 11 and a comparison unit 12. Each component can be realized by a processor executing a program stored in a memory.

[0014] The calculation unit 11 uses a formula included in the trained model to calculate the probability of unexpected demand occurring from the product number attached to the spare part of the automobile to be predicted. Unexpected demand represents a situation in which the demand for the spare part is higher than the expected demand.

[0015] The comparison unit 12 compares the calculated probability with a threshold probability value included in the trained model to automatically determine whether the production system needs to be maintained or whether the production system can be discarded. The comparison unit 12 outputs the determination result.

[0016] The trained model includes a mathematical formula for calculating the probability of unexpected demand occurring. The mathematical formula includes explanatory variables representing feature quantities obtained from the product number to be predicted and coefficients for weighting each explanatory variable. The coefficients are trained using training data. Furthermore, the trained model includes a threshold probability value trained using the training data.

[0017] The unexpected demand may be a situation in which the actual demand for the spare parts is higher than the expected demand. When the unexpected demand occurs, the user can appropriately set how many times the actual demand should be higher than the expected demand.

[0018] The determination system 100 may include a function for training a trained model using machine learning. The determination system 100 accumulates time-series data indicating changes in the inventory quantity of spare parts in a database. From this time-series data, the determination system 100 determines the value of a dependent variable indicating whether unexpected demand has occurred for each part number, and assigns a label to each part number according to the definition. For example, a dependent variable value of 1 indicates that unexpected demand has occurred, and a dependent variable value of 0 indicates that unexpected demand has not occurred. The determination system 100 then sets the characteristics of the spare parts associated with each part number as multiple explanatory variables. The explanatory variables may be qualitative variables that are either 0 or 1, categorical variables with multiple categories, or quantitative variables with continuous values.

[0019] A method for determining a weighting coefficient for each explanatory variable will be described. The determination system 100 determines a coefficient by logistic regression using the maximum likelihood method, which is a type of machine learning. This can be performed on a computer using publicly available Python statsmodels or R generalized linear models.

[0020] The determination system 100 sets the occurrence or non-occurrence of unexpected demand as a response variable, and sets the characteristics of the supply parts associated with each product number as explanatory variables. Each explanatory variable x i Coefficient β that weights i and the intercept α are output as the calculation results.

[0021] The sum of the intercept and the value obtained by multiplying each explanatory variable by the coefficient that weights that explanatory variable is called the linear predictor y. The output equation that connects the objective variable, that is, the qualitative variable that is 0 or 1, and each explanatory variable is the logistic function, shown in equation (1). The logistic function is also called the sigmoid function.

number

[0022] The determination system 100 calculates the threshold probability value using a loss function calculated as follows. The determination system 100 first sets the threshold to a probability value between 0 and 1. Among product numbers for which a linear predictor is calculated that results in a probability value equal to or less than the threshold, the determination system 100 assigns a profit represented by a predetermined negative loss amount 1 to product numbers for which the prediction is correct, i.e., product numbers for which unexpected demand will not occur. The determination system 100 assigns a loss represented by a predetermined positive loss amount 2 to product numbers for which the prediction is incorrect, i.e., product numbers for which unexpected demand will occur.

[0023] The determination system 100 also assigns a profit represented by a predetermined negative loss amount 3 to product numbers for which the prediction is correct, i.e., product numbers for which unexpected demand will occur, among product numbers for which a linear predictor is calculated that has a probability value exceeding or equal to the threshold. The determination system 100 assigns a loss represented by a predetermined positive loss amount 4 to product numbers for which the prediction is incorrect, i.e., product numbers for which unexpected demand will not occur.

[0024] The determination system 100 calculates the total loss amount as the sum of negative loss amount 1, positive loss amount 2, negative loss amount 3, and positive loss amount 4 for all product numbers in the learning data. The determination system 100 calculates the total loss amount for linear predictors corresponding to all possible probability values ​​or linear predictors corresponding to representative values ​​extracted from all possible probability values. The determination system 100 then sets the probability value corresponding to the linear predictor that minimizes the total loss amount as the threshold probability value.

[0025] Negative loss amount 1 is set to the amount expected as profit from the sale of discarded production equipment. Positive loss amount 2 is set to the cost of preparing a reproduction system, for example, the amount expected as the cost of remanufacturing a mold. Negative loss amount 3 is set to the amount expected as profit from the sale of manufactured spare parts. Positive loss amount 4 is set to the cost of maintaining a production system, for example, the amount expected as the cost of storing or maintaining a mold.

[0026] The characteristics of the spare part associated with each part number represent information obtained from the part number assignment rules, such as the name of the vehicle model on which the part is installed, the installation location, the part's major classification, and whether the part is a single part or an assembled part. The part's major classification may represent, for example, a powertrain part, a body part, or an electrical part.

[0027] Furthermore, the characteristics of a spare part may be information obtained from the product name assigned to the part number. The characteristics of a spare part may include information indicating whether the spare part is a functional part or a design part. The characteristics of a spare part may include information indicating whether a mold is used to manufacture the spare part. The characteristics of a spare part may include information indicating whether the spare part has a composite material configuration or whether the spare part is made of a single material. The characteristics of a spare part may include information indicating whether the spare part is a relatively large part or a small part. The characteristics of a spare part may include information indicating the purchase price of the spare part.

[0028] The characteristics of the service part may also include information obtained from a drawing corresponding to the part number, and information indicating the design year, design company, design department, etc. of the service part.

[0029] The characteristics of the spare parts may also include information on the number of vehicle models equipped with the spare part linked to the part number, or information obtained from the information on the number of vehicle models equipped with the spare part. The characteristics of the spare parts may also include the total number of vehicle models sold, the number of years since mass production of the vehicle model was discontinued, and the characteristics of the vehicle model (e.g., sedan, sports car, family car, luxury car, special edition vehicle, commercial vehicle, and government vehicle). The characteristics of the spare parts represent the sales price range of new cars, the export ratio of new cars for each vehicle model equipped with the spare part, the export ratio as used cars, the vehicle insurance rate class for each vehicle model equipped with the spare part in each country, the average vehicle age, the number of recalls and free repair costs for each vehicle model equipped with the spare part, etc.

[0030] The characteristics of the spare parts may include the sales ratio and trends (e.g., rate of change) of the spare parts overseas and in Japan obtained from the past demand trends of the target part number, and the annual sales volume trends (e.g., rate of change).

[0031] The characteristics of the spare part may include information such as the name of the primary supplier obtained in association with the part number, the number of secondary and subsequent suppliers obtained from the name of the primary supplier, and characteristics of the supply chain. The characteristics of the supply chain may include, for example, information on whether the supply chain includes suppliers related to industries other than the automotive industry (e.g., the semiconductor industry). The characteristics of the spare part may include information indicating the number of processes, number of models, number of specialized equipment, number of general-purpose equipment, number of types of constituent materials, and the supply period of the constituent materials required to manufacture the spare part.

[0032] Next, an example of the definition of unexpected demand will be explained. First, the demand quantity for the lifetime (e.g., 30 years) from the time of prediction is predicted using conventional techniques based on past demand trend data for the product number to be predicted. Then, it is determined whether the actual demand quantity will exceed the predicted demand quantity over a certain period (e.g., 30 years). The occurrence of unexpected demand is defined as the actual demand quantity exceeding the predicted demand quantity.

[0033] Next, a method for determining whether or not unexpected demand has occurred in the learning data will be described. If the latest inventory amount in the learning data is zero, or if it is determined from extrapolation of inventory trends that the inventory amount will be zero within the next year, it is determined that unexpected demand has occurred. In other words, if the inventory amount of spare parts at the time of prediction is less than a predetermined amount, it may be determined that unexpected demand has occurred. Furthermore, if the annual inventory decrease rate is within the top 5% of all part numbers, it is determined that unexpected demand has occurred.

[0034] An example of the operation of the determination system 100 will be described with reference to FIG. 2. First, the part number of the target spare part is input to the determination system 100 (step S11). Next, the determination system 100 generates features of the spare part associated with the part number and sets the generated features as multiple explanatory variables (step S12). The features of the spare part include information on the vehicle model (e.g., three categories: A, B, and C), the part category determined by the first digit of the part number (e.g., seven categories), the two categories of assembly (combined part) or part, the two categories of whether the decay rate (the ratio of the previous year's inventory to the current inventory) is high or low, the two categories of design part or functional part, and the two categories of whether the part is manufactured using a mold. Note that for explanatory variables with three or more categories, one category may be removed as a criterion. For example, if vehicle model A or the part category engine is removed, the weighting of the other explanatory variables is interpreted as being set relative to the criterion.

[0035] Next, the calculation unit 11 of the determination system 100 calculates the probability of unexpected demand occurring using a formula included in the trained model 20 (step S13). For example, the calculation unit 11 calculates a linear predictor y from the explanatory variables and converts the calculated y into p i =1 / (1+e -y ) and p i Next, the determination system 100 reads the threshold probability value included in the trained model 20 (step S14). Next, the comparison unit 12 of the determination system 100 determines whether the calculated probability is smaller than the threshold probability value (step S15). If the calculated probability is smaller than the threshold probability value, the determination system 100 outputs information that the production system of the spare parts can be discontinued (step S16). If the calculated probability is equal to or greater than the threshold probability value, the determination system 100 outputs information that the production system of the spare parts will be maintained (step S17).

[0036] Next, we will explain how to train the trained model. The learning data shown in Figure 3 indicates data on spare parts that were produced in bulk at a certain prediction point in time. The learning data includes time-series data showing the quantity of the spare parts produced in bulk, changes in the inventory quantity after the spare parts began to be supplied from inventory, and changes in the overseas ratio of the demand quantity. Then, as explanatory variables, two types of information are set: information on whether the attenuation rate (the ratio of the inventory rate to the previous year) is high or not, and information on whether the overseas ratio is high or not.

[0037] The quantity of spare parts to be produced in a lump sum is determined in the following way. First, a predicted attenuation rate β is obtained from a table set up in advance based on past experience, using data on the transition of past shipments of the spare part and the part number group to which the spare part belongs. The lump sum production volume is calculated, for example, using equation (2). N i indicates the number of shipments per fiscal year.

number

[0038] Then, similar to step S12 in Fig. 2, other features of the spare part linked to the part number are generated, and multiple explanatory variables are set. Then, a response variable indicating whether or not unexpected demand has occurred is linked to the part number. For example, if the latest inventory quantity is zero, if it is determined that the inventory quantity will be zero within the next year based on extrapolation of past inventory quantity trends, or if the annual inventory quantity decrease rate is within the top 5% of all part numbers, the response variable is set to 1; otherwise, the response variable is set to 0.

[0039] The determination system 100 then generates the equations included in the trained model by learning using logistic regression that uses the maximum likelihood method, which is a type of machine learning. For example, learning can be performed on a computer by selecting a generalized linear model using statistical software R. This allows a weighting coefficient and an intercept corresponding to each explanatory variable to be calculated. A linear predictor y is calculated from each explanatory variable, the coefficients that weight each explanatory variable, and the intercept. The probability p of unexpected demand occurring is then calculated from the linear predictor y. i is calculated.

[0040] The horizontal axis of Figure 4 represents the linear predictor value, and the vertical axis represents the probability of unexpected demand occurring. If the threshold probability value is set to 0.10, the corresponding threshold of the linear predictor is approximately -2.2. In this case, it is determined that the production system for product numbers whose linear predictor value is to the left of the vertical line in Figure 4 can be discarded. If unexpected demand occurs for some of the product numbers whose linear predictor value is smaller than -2.2, there is a risk of large losses.

[0041] Consider plotting points on Figure 5 that indicate whether or not unexpected demand will occur for each supply part. The horizontal axis value for each point is the value of the linear predictor for that supply part. The vertical axis value for each point is 1 if unexpected demand will occur for that supply part, and 0 if not. In this case, each supply part is associated with one of areas A11, A12, A21, and A22. Unexpected demand will occur for supply parts associated with areas A21 and A22. Information is output that indicates that the production system can be discontinued for supply parts associated with areas A11 and A21. Information is output that indicates that the production system will be maintained for supply parts associated with areas A12 and A22.

[0042] Corresponding to areas A11 and A12 is called negative, and corresponding to areas A21 and A22 is called positive. Corresponding to area A11 is called true negative, corresponding to area A12 is called false positive, corresponding to area A21 is called false negative, and corresponding to area A22 is called true positive. In other words, among negative supply parts, supply parts that are correctly judged are called true negatives. Among positive supply parts, supply parts that are correctly judged are called true positives. Among negative supply parts, supply parts that are incorrectly judged are called false positives. Among negative supply parts, supply parts that are incorrectly judged are called false negatives.

[0043] The loss function assigns a profit for each correct judgment of a spare part and a loss for each incorrect judgment of a spare part. The loss function assigns profit or loss according to Figure 6, for example. For example, if the production system is a mold, -A represents the cost of reproducing the mold. -B represents the cost required to store the mold. +C represents the profit gained from selling the scrap when the mold is discarded. +D represents the profit gained from selling spare parts to meet unexpected demand.

[0044] The determination system 100 calculates the total loss amount for all linear predictor values, and then sets the probability of the linear predictor with the smallest total loss amount as the threshold probability value.

[0045] Next, the effects of the first embodiment will be described. The production system for spare parts is maintained for several years to several decades after the end of mass production of new cars, but it is inefficient because the quantities are small. In response to this, there is a method of predicting the lifetime demand quantity and producing spare parts in bulk, and supplying them from inventory, but there is a problem that if unexpected demand occurs and the parts run out of stock, expensive re-production will be required.

[0046] In embodiment 1, the AI ​​uses a trained model to determine which product numbers are unlikely to experience unexpected demand and encourages the abandonment of the production system, thereby reducing the costs required to maintain the production system.

[0047] The above-mentioned program includes a set of instructions (or software code) that, when loaded into a computer, causes the computer to perform one or more functions described in the embodiments. The program may be stored in a non-transitory computer-readable medium or a tangible storage medium. By way of example and not limitation, computer-readable media or tangible storage media include random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technologies, CD-ROM, digital versatile disk (DVD), Blu-ray® disk or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device. The program may also be transmitted on a transitory computer-readable medium or communication medium. By way of example and not limitation, transitory computer-readable media or communication media include electrical, optical, acoustic, or other forms of propagated signals. [Explanation of symbols]

[0048] 100: Judgment system, 11: Calculation unit, 12: Comparison unit, 20: Trained model

Claims

1. A determination system for determining whether or not to dispose of equipment or a mold for manufacturing automotive spare parts, a means for calculating, using a mathematical formula included in a trained model, a probability of occurrence of unexpected demand representing a situation in which demand for the supply part is greater than expected demand, based on a product number assigned to the supply part; means for comparing the probability with a threshold probability value included in the trained model; Including, If the probability is less than the threshold probability value, information indicating that the equipment or model can be discarded is output, and if the probability is equal to or greater than the threshold, information indicating that the equipment or model should be maintained is output; a mathematical formula included in the trained model includes explanatory variables into which feature values ​​obtained from the product numbers are substituted, and coefficients for weighting each explanatory variable; The coefficients are trained and calculated by logistic regression using training data, The threshold probability value is determined so as to minimize a profit given for each spare part whose prediction by the formula in the learning data is correct, and a loss amount based on a loss given for each spare part whose prediction by the formula in the learning data is incorrect. Judging system.

2. The part number is associated with at least one of the following information: a category to which the replacement part belongs; a vehicle model in which the replacement part is installed; whether the replacement part is a combination part; whether the replacement part is a design part; and whether the replacement part is manufactured as a mold. The determination system according to claim 1 .

3. In the learning data, a label indicating whether or not the unexpected demand has occurred is assigned to each supply part, If the inventory amount of the supply part at the time of prediction is less than a predetermined amount, a label indicating that the unexpected demand has occurred is attached to the supply part. The determination system according to claim 1 or 2.

4. the profit when the probability is equal to or greater than the threshold probability value includes a profit obtained by selling the spare parts in response to the unexpected demand; If the probability is less than the threshold probability value, the profit includes the profit obtained by selling the equipment or mold. The determination system according to claim 1 or 2.

5. the loss if the probability is less than the threshold probability value includes the cost of reproducing the equipment or mold; If the probability is greater than or equal to the threshold probability value, the loss includes the cost of storing the equipment or mold. The determination system according to claim 1 or 2.

Images