Warehouse management system with functions for performing warehouse management and preventive maintenance

Abstract

Warehouse management system (1), comprising: a plurality of devices (3); An information management device (2) for managing information to be supplied to or received from the devices (3), wherein the devices (3) each have replaceable common components that can be shared between the devices (3), wherein the common components with the same specifications used in the devices (3) are referred to as components under inventory management, wherein the information management device (2) is configured to: Collecting and storing state variables of the devices (3); Determine whether the operating states of the components indicate that the components have failed or not; If components have failed, replace the failed components and update the stock quantity of the designated components; If no components have failed, calculate the cumulative failure rate of each of the designated components with reference to data on the cumulative failure rate of the designated components under a standard operating condition and with reference to data on an acceleration coefficient of the cumulative failure rate as a function of an operating condition; verify a lead time for each designated component; calculate an estimated cumulative failure rate as the cumulative failure rate after a lead time has elapsed from a present time with reference to the lead time of each designated component; calculate a sum of the estimated cumulative failure rates by adding the estimated cumulative failure rates; and determine whether an inventory quantity of each type of the designated components is greater than or equal to the sum of the estimated cumulative failure rates plus a margin;and ordering further components if the stock quantity is not greater than or equal to the sum of estimated cumulative error rates plus the margin, the margin comprising a predetermined number that is increased in the case of longer lead times and reduced prices.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention

[0001] The present invention relates to a warehouse management system having the functions of performing warehouse management and preventive maintenance. 2. Description of related technology

[0002] Equipment malfunctions due to component failures or damage often cause equipment to stop and create downtime periods during which the equipment is unusable. To prevent an increase in downtime frequency and downtime periods, the inventory quantity of spare components for each device is typically managed.

[0003] Even if an average failure rate (the number of failures per unit of time) is known to some extent depending on the operating time of each component, the failure rate may vary according to operating conditions, and therefore, it is generally difficult to predict failure unless there is an obvious sign indicating when each component will fail. Therefore, excessive or insufficient inventory can often occur in inventory management for fixture components.

[0004] If a component is expensive, excessive inventory is uneconomical. On the other hand, insufficient inventory will cause the device to be stopped until a replacement component is delivered. If the device is a piece of production equipment, a production process using the device will be significantly delayed, causing a significant loss. Therefore, appropriate inventory management is required, taking into account a failure that has no warning signs because it does not announce itself.

[0005] Since inventory management for maintenance components of devices is an important technology, various related techniques are proposed (for example, Japanese Unexamined Patent Publications (Kokai) JP 2010-113672 A, JP 2004-295667 A and JP 2012-104058 A). SUMMARY OF THE INVENTION

[0006] However, no prior art inventory management system with an inventory management function that considers unsigned defects has been proposed. Therefore, it has been impossible to accurately estimate the number of unsigned defects and perform appropriate inventory management. Therefore, inventory management for components that consider unsigned defects is required.

[0007] According to one aspect of this disclosure, a warehouse management system includes a plurality of devices and an information management device connected to the devices. The devices each have interchangeable common components that can be shared between the devices. The common components, which have the same specifications as those used in the devices, are referred to as warehouse management components.

[0008] The inventory management system includes an information processor for calculating the sum of the cumulative failure rates, which is the sum value of the cumulative failure rates of all the designated components with the same specifications used in the devices at a certain time, taking into account an acceleration depending on an operating condition with respect to the cumulative standard failure rate of each type of the designated components with the same specifications used in the devices under a standard operating condition, and for calculating the appropriate inventory quantity of the designated components at the certain time based on the calculated sum of the cumulative failure rates.

[0009] The features and advantages of the present invention and other features and advantages will become more apparent from the following detailed description of preferred embodiments of the present invention together with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram showing the schematic configuration of a warehouse management system according to an embodiment; Fig. Figure 2a is a graph showing an example of variations in the cumulative failure rate of a component whose dominant failure is a random failure relative to an operating time; Fig. Figure 2b is a graph showing an example of variations in the cumulative failure rate of a component whose dominant failure is a wear failure relative to an operating time; Fig. 3 is a graph showing an example of the relationship between an operating condition and an acceleration coefficient under the operating condition; Fig. 4a is a graph showing an example of a variation of the cumulative failure rate of a component whose dominant failure is a random failure relative to an operating time when the component is operated under operating conditions with different acceleration coefficients; Fig. 4b is a graph showing an example of a variation of the cumulative failure rate of a component whose dominant failure is a wear failure relative to an operating time when the component is operated under operating conditions with different acceleration coefficients; Fig. 5a is a graph showing an example of a variation in each of the sum of cumulative failure rates, the sum of estimated cumulative failure rates, and the appropriate inventory quantity, relative to an elapsed time, of a plurality of components with the same specifications whose dominant failures are random failures, provided in equipment under inventory management; Fig. 5b is a graph showing an example of a variation of the sum of cumulative failure rates, the sum of estimated cumulative failure rates, and the appropriate inventory quantity relative to an elapsed time of a plurality of components with the same specifications whose dominant failures are wear failures, provided in equipment under inventory management; Fig. Figure 5c is a graphical representation when the suitable storage quantity is given by an equation “suitable storage quantity ≥ sum of estimated cumulative error rates + 1” instead of Fig. 5b is calculated; Fig. 6 is a flowchart showing an example of the operation of the Fig. 1 shows the warehouse management system; Fig. 7 is a graph showing characteristic change curves indicating characteristic changes by operation, and a method of estimating the remaining life from a characteristic change curve in which a characteristic value and a deterioration rate coincide; Fig. 8 is a graph showing characteristic change curves under an operating condition in which an acceleration coefficient of a deterioration rate by operation is twice that of Fig. 7, and shows a method for estimating the remaining life from a characteristic change curve in which a characteristic value and a deterioration rate coincide; Fig. 9 is a flowchart showing an example of the operation of the Fig. 1 shown warehouse management system, being a flowchart that leads to Fig. 6 can be added; Fig. 10 is a block diagram showing the schematic configuration of a warehouse management system according to another embodiment; Fig. 11 is a flowchart showing an example of the operation of a machine learning device in the Fig. 10 shows the warehouse management system; Fig. 12 is a flowchart showing another example of the operation of the machine learning device in the warehouse management system; Fig. 13 is a block diagram showing the schematic configuration of a warehouse management system according to still another embodiment; Fig. 14 is a drawing schematically showing the passage of time until an unknown character is found; Fig. 15 is a flowchart showing an example of the operation of a machine learning device in the Fig. 13 shows the warehouse management system; Fig. 16 is a flowchart showing an example of the operation of the Fig. 13 shown warehouse management system, being a flowchart that leads to Fig. 15 can be added; Fig. 17 is a block diagram showing the schematic configuration of a warehouse management system according to still another embodiment; Fig. 18 is a flowchart showing an example of the operation of a machine learning device in the Fig. 17 shows the warehouse management system; Fig. 19 is a flowchart showing another example of the operation of the machine learning device in the Fig. 17 shows the warehouse management system; Fig. 20 is a block diagram showing the schematic configuration of a warehouse management system according to still another embodiment; Fig. 21 is a block diagram showing the schematic configuration of a warehouse management system according to still another embodiment; and Fig. 22 is a block diagram showing the schematic configuration of a warehouse management system according to still another embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] Next, embodiments of this disclosure will be described with reference to the drawings. In the reference drawings, the same reference numerals denote the same or similar structural or functional components. For ease of understanding, the drawings are appropriately scaled. The embodiments shown in the drawings are merely examples embodying the present invention, and the present invention is not limited to the embodiments shown.

[0011] Fig. 1 is a block diagram showing the schematic configuration of a warehouse management system 1 according to an embodiment.

[0012] As in Fig. 1, the warehouse management system 1 includes an information management device 2 and a plurality of devices 3 connected to the information management device 2. The information management device 2 manages information to be delivered to or received from the devices 3. More specifically, the information management device 2 is a device that collects state information of the devices 3 from the devices 3 through a communication controller, a memory, and the like of each device 3 to perform recording, analysis, and the like of the state information in a centralized manner to output the required information.

[0013] The information management device 2 includes, for example, a communication network such as the Internet and an intranet, and an information processor 30. Each of the devices 3 and the information processor 30 is composed of a computer, a server, or the like, including a memory such as a ROM (Read Only Memory) and a RAM (Random Access Memory), a CPU (Control Processing Unit), and a communication controller, which are connected to each other via a bus. Furthermore, the functions and operations of the devices 3 and the information processor 30 are each realized through the cooperation of the CPU and memory mounted in the computer, the server, or the like, and a control program stored in the memory.

[0014] Devices under inventory management and preventive maintenance are generally limited, and the target devices 3 are installed, for example, in the same manufacturing cell, the same building, the same factory site, the same areas to which maintenance components can be delivered within a predetermined period of time, or the like. This is because there is no point in having a spare component unless the spare component can be delivered within a predetermined period of time to the location of a device that has been stopped due to a component failure or is in a non-working state. Thus, the inventory management system 1 according to this disclosure performs inventory management of common components with the same specifications used in the two or more designated devices 3.

[0015] It should be noted that the term "component" as described in this specification and within the scope of the claims encompasses at least one component or unit. A component refers to an element constituting a machine, electrical and electronic equipment, and the like. A unit refers to an assembly of elements.

[0016] The term "apparatus" as described in this specification and the scope of the claims includes at least one of a device, a facility, a machine, or a system. Furthermore, the term "the same specifications" as described in this specification and the scope of the claims includes the meaning of approximately the same specifications, in which the specifications are substantially the same, except for the meaning that each specification is the same.

[0017] As will be apparent from the following description, the warehouse management system 1 according to this disclosure must collect a huge volume of information from each device 3. Thus, a configuration such as a cloud computing system that widely collects information from the devices 3 via the Internet places an excessively large burden on a communication network and makes it difficult to process the information from each device 3 at high speed in real time. Therefore, the information management device 2 constituting the warehouse management system 1 is preferably arranged between the devices 3 and the cloud (not shown) to perform the management and processing of the information close to the devices. However, as shown in Fig. 1, a group of devices 3 are connected to the information management device 2 as devices under inventory management and preventive maintenance, but the present invention is not limited to Fig. 1. A plurality of devices 3 divided into groups may be connected to the information management device 2, and the devices 3 may be subjected to inventory management and preventive maintenance on a group-by-group basis.

[0018] The devices 3 connected to the information management device 2 have at least one type of interchangeable common components commonly used in the devices 3, and the common components are referred to as inventory management components. The inventory management components (hereinafter referred to as designated components) are managed based on each component group having the same specifications as those used in the devices 3. However, a group of inventory management and preventive maintenance devices 3 does not necessarily have to have the same specifications.

[0019] Regarding a type of designated components with certain specifications, it is known that their cumulative failure rate, which is an operating time integral of the failure rate (the number of failures per time) of the designated components, is given by Fig. 2a or Fig. 2b. In this description, the term "operation" replaces the terms operation, power supply, and the like, although the terms operation, power supply, and the like may be more appropriate for a condition than the term "operation," depending on the type and use of the designated components.

[0020] In both Fig. 2a and Fig. 2b, the horizontal axis represents a value obtained by dividing the operating time by MTTF (mean time to failure: average lifetime) of the designated components operated under standard operating conditions. Fig. Figure 2a shows an example of the cumulative error rate when a random error is dominant. Fig. Figure 2b shows an example of the cumulative failure rate when a wear failure is dominant.

[0021] In the Fig. 2a and Fig. In Figure 2b, solid lines represent the cumulative failure rate of the designated component used under standard operating conditions. Dashed lines represent the case where an acceleration coefficient of the cumulative failure rate is doubled, for example, due to a higher operating voltage or a higher temperature than the standard operating conditions, and the cumulative failure rate reaches 0.5 in half the operating time. Alternate long and short dashed lines represent the case where an acceleration coefficient of the cumulative failure rate is halved, for example, due to a lower operating voltage or a lower temperature than the standard operating condition, and the cumulative failure rate reaches 0.5 in twice the operating time.

[0022] In each of the Fig. 2a and Fig. 2b, the operating time until the cumulative error rate reaches a certain value (e.g. 0.5) is shortened by the acceleration coefficient or extended to 1 / acceleration coefficient. Fig. Figure 3 shows an example of the relationship between the operating conditions and the acceleration coefficient. In the example of Fig. 3, the operating voltage is used as the operating condition, and the acceleration coefficient is the cube of "operating voltage / standard operating voltage." Operating conditions that accelerate the increase in the cumulative failure rate and the consumption of the service life due to operation include operating current, temperature, temperature cycle, vibration, and the like, in addition to the Fig. 3. If the component is a light-emitting device, light emission is a typical operating condition that accelerates the increase in the cumulative failure rate and lifetime consumption. To calculate the increase in the cumulative failure rate and lifetime consumption due to operation, it is necessary to consider an acceleration coefficient for each accelerated operating condition, except for the case where the acceleration coefficient is close to 1.

[0023] If an acceleration coefficient that shortens or lengthens the time until the cumulative error rate reaches a certain value is different from the other acceleration coefficients, the acceleration coefficient is referred to as the "first acceleration coefficient" in this specification.

[0024] Each of the Fig. 4a and Fig. Figure 4b is an example of a temporal variation of the cumulative error rate of one of the designated components with the cumulative error rate characteristics shown in the Fig. 2a and Fig. 2b. In both Fig. 4a and Fig. 4b, the horizontal axis represents a value at which an elapsed time from the start of operation is divided by MTTF of the designated components operating under the standard operating condition. Fig. Figure 4a shows an example of the cumulative error rate when a random error is dominant. Fig. Figure 4b shows an example of the cumulative failure rate when a wear-out failure is dominant. MTTF is on the order of, for example, one hundred thousand hours. Although it is unnatural for a device to continue operating for several tens of thousands of hours under a fixed operating condition and to stop operating for a few tens of thousands of hours under a fixed operating condition, as in the Fig. 4a and Fig. 4b for easier understanding, each of the Fig. 4a and Fig. 4b a case where the designated component is operated for a long time under the fixed operating conditions.

[0025] Each of the Fig. 4a and Fig. Figure 4b is a graph of the cumulative error rate when the designated component is operated in the following procedure. First, as shown in the Fig. 4a and Fig. As shown in Figure 4b, the designated component is operated under the standard operating condition. The operating condition is switched to have a cumulative failure rate acceleration coefficient of 0.5 at any time, and then the operating condition is switched to have a cumulative failure rate acceleration coefficient of 2 at any time. After that, the designated component is not operated from a certain point in time during a standstill period. While the designated component is being operated again under the standard operating condition, a failure occurs, and the designated component is replaced with a new designated component. After the cumulative failure rate is reset to zero, the designated component is operated under the standard operating condition.

[0026] When switching the operating state, the cumulative error rate at that time is switched to the graph of the cumulative error rate with an acceleration coefficient of a new operating state (as in Fig. 2a and Fig. 2b) and thus the cumulative error rate increases over time. Therefore, Fig. 4a and Fig. 4b Examples of the cumulative failure rate taking into account an acceleration coefficient depending on the operating condition relative to a cumulative standard failure rate of the designated components with the same specifications under the standard operating conditions. As can be seen from the Fig. 4a and Fig. As can be seen from Figure 4b, the cumulative error rate differs largely from the actual rate without considering the acceleration coefficient depending on the operating condition.

[0027] Even if the cumulative failure rate of a designated component is known, taking into account an acceleration coefficient depending on an operating condition, the time at which the component actually fails depends on a probability and is not known until it occurs. However, calculating the sum of the cumulative failure rates of all designated components with the same specifications used in a group of devices 3 under inventory management and preventive maintenance, taking into account acceleration factors depending on the operating conditions (i.e., the sum of the cumulative failure rates), as in the Fig. 5a and Fig. 5b to obtain an index for the appropriate inventory quantity of designated components.

[0028] Fig. Figure 5a is a graph showing an example of a temporal variation in the sum of cumulative failure rates of designated components whose dominant failures are random failures and the appropriate inventory quantity. Fig. Figure 5b is a graph showing an example of a temporal variation in the sum of cumulative failure rates of designated components whose dominant failures are wear failures and the appropriate stock quantity. In each of the Fig. 5a and Fig. In Figure 5b, the thick dashed line represents the sum of the cumulative failure rates (i.e., Σ of the cumulative failure rate) of all designated components with the same specifications used in a group of fixtures 3 under inventory management and preventive maintenance. The thin dashed line represents the sum of the estimated cumulative failure rates (i.e., Σ of the estimated cumulative failure rate) after the expiration of a delivery time (i.e., time required for delivery) of the designated components from a certain point in time.

[0029] The sum of the estimated cumulative failure rates is defined as the estimated cumulative failure rates after the expiration of a delivery period, which are estimated from time variations in the cumulative failure rates of the designated components with the same specifications as in the Fig. 4a and Fig. 4b. The appropriate storage quantity, shown by the thick solid line in each of the Fig. 5a and Fig. 5b is a positive integer to which the sum of the estimated cumulative error rates is rounded up and generally represents a history of the appropriately managed inventory quantity.

[0030] To make the drawings easier to understand, Fig. 5a and Fig. 5b shows the sums of cumulative failure rates of ten designated components with the same specifications operating under the same operating conditions. In the example of Fig. 5a starts operating the ten components simultaneously from the beginning.

[0031] On the other hand, in the example of Fig. 5b operating for five of the ten components first, and the remaining five components are operated sequentially at intervals of MTTFx0.1 with a delay of MTTFx0.3 from the beginning to prevent the occurrence of a wide valley in the sum of the cumulative failure rates due to the small number of components.

[0032] As in Fig. 5a and Fig. As shown in Figure 5b, the sum of the cumulative failure rates varies greatly because the number of designated components is small when calculating the sum of the cumulative failure rates, but the rate of change decreases with an increase in the number of designated components. A delivery time is set to be long, i.e., with one-fiftieth of the MTTF, so the graphs of the sum of the cumulative failure rates and the sum of the estimated cumulative failure rates are shown separately. Due to the long delivery time in the example of Fig. 5b the inventory quantity is zero for fairly long periods of time.

[0033] Under the conditions described above, managing inventory levels based on the sum of the cumulative failure rates—in other words, making provisions for replacement components when the actual inventory level becomes less than the sum of the estimated cumulative failure rates—prevents a situation where, when a designated component fails, there is no replacement component for the failing designated component. Therefore, inventory levels are managed appropriately.

[0034] To convey the sense that inventory quantity is managed based on the sum of the cumulative failure rates, for example, if the cumulative failure rate of each designated component is 0.01 and there are 1,000 designated components with the same specifications, the sum of the cumulative failure rates is 10 (ten). This means that, on average, ten of the components can fail at any time, and at least ten replacement components of the designated components are required.

[0035] If the number of failures is low, an increased number of components has been used for a long period, and the number of failures may subsequently increase. In this case, inventory management can be performed taking into account the need to increase the inventory quantity.

[0036] Since the probability of a defect occurring depends on a statistical probability, an actual result (the actual number of defects) varies to a greater extent as the number of components (population) decreases. Therefore, to reduce the risk of insufficient inventory, an adjustment can be made such that a predetermined number is added to the minimum required inventory quantity as a margin, and a calculated number is called the appropriate inventory quantity, or the product of the minimum required inventory quantity and a predetermined ratio is added to the inventory quantity as a margin, and a calculated number is called the appropriate inventory quantity.When the number of designated components is relatively small, the former management method is preferred in most cases, in which the sum of the minimum required inventory quantity and a predetermined number is referred to as the appropriate inventory quantity.

[0037] Fig. Figure 5c represents the appropriate stock quantity when implementing a stock management procedure whereby provisions are made for a designated component as soon as the subtraction of the sum of the estimated cumulative defect rates from the actual stock quantity becomes equal to or less than 1, with reference to the example of Fig. 5b, in other words, when the stock management procedure is implemented whereby provisions are made for a designated component as soon as the actual stock quantity becomes lower than the sum of the estimated cumulative defect rates.

[0038] In the example of Fig. 5c compared with the example of Fig. 5b, the inventory level never becomes zero in any time period. Although the inventory level is increased by almost 1 on average, the risk of a delay in the device's recovery due to insufficient inventory is reduced. The margin can be adjusted based on the type of designated components, depending on their price and delivery time, taking into account the balance between an increase in inventory level and the risk of insufficient inventory. In general, the longer the delivery time and the cheaper the price, the larger the margin is preferable.

[0039] Fig. 6 is a flowchart showing an example of the operation of the warehouse management system according to this embodiment. In the following description based on Fig. 6, however, an information management device 2 also functions as an information processor 30.

[0040] After starting the inventory management, the information management device 2 collects and stores state quantities of the devices 3, which include the operating states and states of designated components provided in the devices 3 under inventory management, through the communication controllers and the like of the devices 3 (step S101).

[0041] The information management device 2 determines whether the operating conditions of the designated components acquired in step S101 are normal or not, in other words, whether the designated components have not failed (step S102). If they are normal, the information management device 2 calculates the cumulative failure rate of each designated component from the data regarding the operating conditions of the designated components collected in step S101, taking into account an acceleration coefficient depending on the operating condition of each designated component with reference to data on the cumulative failure rate of the designated components under a standard operating condition and with reference to data on an acceleration coefficient of the cumulative failure rate depending on an operating condition, and records a transition in the cumulative failure rate (step S103).

[0042] Further, the information management device 2 checks a delivery time for each designated component that has been stored in advance (step S104). Next, the information management device 2 calculates and stores an estimated cumulative failure rate, which is a cumulative failure rate after the elapse of the delivery time from a certain point in time (a current point in time in general) of each designated component, by referring to the delivery time of each designated component, the data on the cumulative failure rate of the designated component in the standard operating condition, the data on the acceleration coefficient of the cumulative failure rate depending on the operating condition, and the data on the transition of the cumulative failure rate recorded in step S103 (step S105).

[0043] Next, the information management device 2 calculates and records the sum of the estimated cumulative failure rates of the designated components with the same specifications on a component basis by adding the estimated cumulative failure rates of the designated components with the same specifications calculated and recorded on a component basis in step S105 (step S106). The information management device 2 checks whether the designated components, which were provided for in advance, have been delivered or not (step S107). If the designated components have been delivered, the stock quantity of the delivered components is updated in a memory (step S108), and the operation proceeds to step S109. If NO occurs in step S107 as a result of the determination of the presence or absence of delivery, the operation proceeds directly from step S107 to step S109.

[0044] In step S109, the information management device 2 determines whether “the inventory quantity of each type of the designated components” ≥ “the sum of estimated cumulative failure rates of each type of designated components with the same specifications” + Δn by comparing the inventory quantity of each type of the designated components stored in the memory and the sum of estimated cumulative failure rates of each type of the designated components with the same specifications calculated and recorded in step S106.

[0045] Δn is a margin represented by a positive integer including zero. Δn = 0 is used in the examples of Fig. 5a and Fig. 5b, and Δn = 1 is used in the example of Fig. 5c, but the margin is not necessarily an integer. The determination equation is only an example; the margin can be calculated based on an equation of "the inventory quantity of each type of designated components" ≥ m × "the sum of estimated cumulative defect rates of each type of designated components that have the same specifications," where m ≥ 1. "Δn" and "m" can be changed depending on the type of designated components.

[0046] If YES occurs in step S109, a suitable inventory quantity is ensured. The information management device 2 determines whether or not a command to stop inventory management is issued (step S110). If the stop command is issued, inventory management is completed. If no stop command is issued, the information management device 2 returns to step S101 and continues inventory management.

[0047] Note that in step S102, if one of the designated components is determined to be failing, the information management device 2 notifies the outside of a replacement of the failing component as a result of determining the operating conditions of the designated components (step S111). A device 3 with the failing component is restored, and the inventory quantity of the designated components is updated (step S112). If the replaced component is a new component, the information management device 2 resets the cumulative failure rate of the designated component to zero (step S113), and the operation returns to step S101 to continue inventory management.

[0048] If NO occurs in step S109, there is an increased possibility that insufficient inventory may occur thereafter. Thus, the information management device 2 prepares for a required inventory quantity of the designated components determined as NO in step S109 to satisfy the determination equation (step S114), and the operation returns to step S101 to continue inventory management.

[0049] By repeating steps S101 to S114, the inventory quantity of the designated components is managed in an appropriate manner.

[0050] As described above, the inventory management system 1 according to this embodiment has the function of calculating the sum of the cumulative failure rates, that is, the sum value of the cumulative failure rates of all designated components with the same specifications used in the group of devices 3 at a certain time, taking into account the acceleration factors depending on the operating conditions, relative to the standard cumulative failure rate of each type of designated components with the same specifications used in the group of devices 3 under the standard operating condition, and the function of calculating the appropriate inventory quantity of the designated components based on the calculated sum of the cumulative failure rates. These functions are intended to perform appropriate inventory management while taking into account an unsigned failure.Therefore, it is possible to prevent delay in maintenance or restoration due to insufficient inventory, but also to prevent excessive inventory.

[0051] Note that the information processor 30, such as the computer or the server arranged in the information management device 2, performs the above functions. More specifically, the information processor 30 performs a series of processes described in Fig. 6, including the calculation of the cumulative error rates and the sum of the cumulative error rates, detection, recording, and the like. The information processor 30, which performs the above functions, may be connected to the information management device 2 or may be installed in any of the devices 3 connected to the information management device 2.

[0052] In a case where, among at least one type of the common designated components provided in the devices 3 connected to the information management device 2, there is a component having a gradually deteriorating characteristic (hereinafter referred to as a characteristic deteriorating component) which needs to be replaced when the characteristic reaches a predetermined reference value, the timing of replacing the characteristic deteriorating component is preferably determined from the characteristic of the characteristic deteriorating component at a certain time and a characteristic deterioration rate taking into account an acceleration coefficient depending on an operating condition relative to a standard characteristic deterioration rate of the characteristicdeteriorating component is estimated under a standard operating condition. The inventory quantity of the designated components is preferably calculated based on the sum of the cumulative failure rates, taking into account the time for replacement in addition to the inventory quantity. The information processor 30 also performs such functions.

[0053] Fig. Figure 7 shows an example of characteristic standard change curves indicating time variations in the properties of characteristic deteriorating components whose properties deteriorate when operated under a standard operating condition. In the example of Fig. 7, an operating current required to operate the characteristic deteriorating component and obtain an output of a predetermined power is used as a characteristic. If the operating current required to achieve the predetermined output exceeds 14A, it is determined that the characteristic deteriorating component is at the end of its service life and needs to be replaced.

[0054] For an actual characteristic change, it is assumed that an operating current for obtaining a predetermined output at a certain time is 10A and an operating current for obtaining the predetermined output signal is 8A before 4500 hours of operation are reached. If the actual characteristic change is consistent with the values ​​in Fig. Comparing the standard characteristic change curves shown in Figure 7, the leftmost standard characteristic change curve of the six standard characteristic change curves has a higher deterioration rate (i.e., a division of a characteristic change by an operating time) than the actual characteristic change. The rightmost standard characteristic change curve has a lower deterioration rate than the actual characteristic change. The fourth standard characteristic change curve from the left has almost the same deterioration rate as the actual characteristic change. Accordingly, the remaining service life is estimated to be on the order of 4600 hours from the fourth standard characteristic change curve, which almost follows the actual characteristic change.

[0055] When an operating condition with a two-fold acceleration of the consumption of a lifetime (ie, an operating condition with an acceleration coefficient of 2) is used during an increase in the operating current from 8A to 10A to obtain the predetermined output at a certain time, it is necessary to estimate the remaining lifetime by comparing it with characteristic change curves shown in Fig. 8, in which the horizontal axis (time axis) of Fig. 7 is compressed to half. Fig. 8 uses an operating condition that has a twice accelerated life consumption, the remaining life is estimated to be on the order of 2300 hours when the component is used under an unchanged operating condition. However, when the component is used under the standard operating condition (i.e., an operating condition with an acceleration coefficient of 1), the remaining life is estimated to be on the order of 4600 hours. If the operating condition is changed during operation, it is necessary to use an average acceleration coefficient (=j(acceleration coefficient)dt / jdt), which is a division of a time integral value of a life consumption acceleration coefficient at each point over time by the duration over time.

[0056] Fig. 7 and Fig. 8 show a limited number of characteristic change curves and a gradient curve between the characteristic change curves can be obtained by interpolation.

[0057] Note that a lifetime consumption acceleration coefficient is referred to as a “second acceleration coefficient” when a distinction is needed between the lifetime consumption acceleration coefficient and the other acceleration coefficients.

[0058] As described above, when the at least one type of commonly designated components provided in the devices 3 includes the characteristic deteriorating component, the remaining lifetime can be estimated based on the characteristic change curve having the same characteristic deterioration rate as the characteristic change of the characteristic deteriorating component in a certain period (in a predetermined period before a certain time point) from the characteristic change curves taking into account an acceleration coefficient depending on the operating state of the characteristic deteriorating component, as shown in Fig. 8. The characteristic deterioration rate of the characteristic deteriorating component in a certain period can be converted into a deterioration rate under the standard operating condition, and a remaining lifetime can be estimated based on the characteristic standard change curve with the same deterioration rate as the characteristic change in the certain period after the conversion outside the characteristic standard change curves, as shown in Fig. 7. For example, in an operating condition with an acceleration coefficient of 2, the actual operating time corresponds to twice the operating time under the standard operating condition. Since a deterioration rate is a division of a characteristic change by the operating time (= characteristic change / operating time), a deterioration rate under the standard operating condition is half of an actual deterioration rate. Thus, a remaining lifetime can be estimated from one of the standard characteristic change curves, as shown in Fig. 7, which has the same deterioration rate as half the deterioration rate of the actual characteristic change. However, when an operating condition is changed during operation, it is necessary to use a time integral value of a lifetime consumption acceleration coefficient at each time point (=∫(Acceleration coefficient)dt) as the operating time under the standard operating condition.

[0059] As described above, the deterioration characteristics in the specific period are compared with the characteristic change curves, taking into account the acceleration coefficient depending on the operating condition. The remaining service life and replacement time for the characteristic deteriorating component are estimated from the characteristic change curve with the same deterioration rate as the deterioration characteristic. The inventory quantity of components designated for maintenance is managed based on the estimated values, enabling more appropriate inventory management performance.

[0060] In the inventory management system according to this embodiment, the information management device 2 can make provisions for replacement of the characteristic deteriorating component, taking into account the replacement time estimated from the degree of deterioration. Furthermore, the information management device 2 outputs preventive maintenance information recommending replacement of the characteristic deteriorating component and serving to replace the component in a planned manner before the deterioration characteristic reaches the predetermined reference value at which the characteristic deteriorating component must be replaced, so as to prevent the device 3 from stopping the component and adversely affecting production using the device 3.

[0061] Fig. 9 is an example of a flowchart used to Fig. 6, and shows inventory management and preventive maintenance steps for the characteristic deteriorating component as an additional operation. By inserting steps S115 to S129 of Fig. 9 between steps S109 and S110 of Fig. 6, inventory management and preventive maintenance can be carried out appropriately on the characteristic, deteriorating component. In the following description based on Fig. 9, the information management device 2 also functions as the information processor 30.

[0062] In Fig. 9, after step S109 of Fig. 6, it is determined whether the designated components include a characteristic deteriorating component or not (step S115).

[0063] If it is not determined in step S115 that the designated components include a characteristic deteriorating component, the operation of the information management device 2 returns to step S101 of Fig. 6. However, if it is determined that the designated components include a characteristic deteriorating component, the information management device 2 collects and stores the state quantity of each device 3, including the operating condition and the characteristic of the characteristic deteriorating component, through the communication controller and the like of the device 3 under inventory management (step S116). The information management device 2 calculates a deterioration rate of the characteristic of the characteristic deteriorating component and an acceleration coefficient of the deterioration rate depending on the operating condition (step S117).

[0064] Then, the information management device 2 compares the characteristic of the characteristic deteriorating component in a certain period with characteristic change curves taking into account the acceleration coefficient, as shown in Fig. 8 (step S118). The information management device 2 selects a curve having the same characteristic deterioration rate as the characteristic of the characteristic deteriorating component in the specified period, and calculates the remaining life of the characteristic deteriorating component from the selected curve, taking into account the acceleration coefficient depending on the operating condition (step S119).

[0065] Further, the information management device 2 determines whether the characteristic deteriorating components, for which preparations were made due to short lifespans and the like, have been delivered or not (step S120). If the characteristic deteriorating components have been delivered, the inventory quantity of the delivered components is updated (step S121), and the operation proceeds to step S122. If they have not yet been delivered, the operation proceeds directly to step S122.

[0066] In step S122, the information management device 2 determines whether preventive maintenance of a characteristic deteriorating component to be replaced soon has been performed in accordance with preventive maintenance information recommending replacement of the characteristic deteriorating component a predetermined time before the replacement time. As a result of the determination, if preventive maintenance has been performed, the information management device 2 updates the inventory quantity of the characteristic deteriorating component used in the preventive maintenance (step S123). The information management device 2 resets the remaining service life of the replaced characteristic deteriorating component to the average service life or the like (step S124), and the operation proceeds to step S125.If preventive maintenance has not been performed, the operation proceeds directly from step S122 to step S125.

[0067] In step S125, the information management device 2 determines whether the remaining service life calculated in step S119 is longer than the time by which a predetermined time T1 is added to a delivery time of the characteristic deteriorating component. In this specification, the delivery time indicates a period of time required from an order for the component to its delivery.

[0068] Regarding the predetermined time T1, which is provided in a sense as a buffer, it is generally conceivable to calculate a remaining lifetime of a characteristic deteriorating component under the assumption that, for example, the characteristic deteriorating component is operated under an operating condition with an average acceleration coefficient of 2, remains in the operating condition, and the average acceleration coefficient in the future continues to be 2. In fact, however, the characteristic deteriorating component may be operated under an operating condition with, for example, an average acceleration coefficient = 4 after being operated under the operating condition with the average acceleration coefficient = 2.In such a case, the actual remaining service life is reduced to half of the expected remaining service life, and therefore the predetermined time T1 must be set with this in mind. It should be noted that continuously monitoring the operating condition and determining the remaining service life relatively frequently based on operating condition data minimizes the difference between the expected remaining service life and the actual remaining service life.

[0069] If it is not determined in step S125 that "remaining life ≥ delivery time + predetermined time T1" is satisfied, the information management device 2 determines whether or not arrangements have been made for the characteristic deteriorating components (step S126). If no arrangement has been made yet, the information management device 2 performs arrangements for the characteristic deteriorating components (step S127) and records that arrangements have been made for the characteristic deteriorating components, and the process proceeds to step S128. However, if it is determined in step S125 that "remaining life ≥ delivery time + predetermined time T1" is satisfied, or it is determined that arrangements have been made for the characteristic deteriorating components in step S126, the process proceeds directly to step S128.

[0070] Step S128 determines whether "remaining life ≥ predetermined time T2" ​​is satisfied. If the remaining life is shorter than the predetermined time T2, the information management device 2 recommends replacing the characteristic deteriorating component and outputs preventive maintenance information to recommend replacing the characteristic deteriorating component (step S129).

[0071] As examples of the preventive maintenance information output, a warning text may be displayed on a monitor or an approximate remaining life may be displayed on a monitor. To ensure notification, the preventive maintenance information may be accompanied by sound and light. A plurality of different predetermined times T2 may be set, and the notification may be switched from a notice to a warning with a decrease in the remaining life. If it is determined in step S128 that “remaining life ≥ predetermined time T2” is satisfied, step S110 of Fig. 6 is executed without executing step S129.

[0072] As described above, by repeating a series of operations including steps S101 to S114 of Fig. 6 and steps S115 to S129 of Fig. 9, with respect to the designated components that do not show signs, the stock quantity may be appropriately managed based on the sum of the cumulative failure rates, taking into account the acceleration coefficients depending on the operating conditions. Regarding the characteristic deteriorating components whose properties gradually deteriorate, the stock quantity may be appropriately managed based on the replacement time estimated from the characteristic deterioration rates, taking into account the acceleration coefficients depending on the operating conditions, and preventive maintenance may be appropriately performed by recommending replacement as the replacement time approaches.

[0073] In this embodiment, the operating procedure ( Fig. 6) for the inventory management of components that do not show any signs and the operational process ( Fig. 9) for the inventory management and preventive maintenance of the characteristic, deteriorating components are carried out in chronological order, although the present invention is not limited to this order. For example, it may be determined at the outset whether the designated components contain characteristic, deteriorating components or not. As a result of the determination, if the designated components do not contain characteristic, deteriorating components, the operation of Fig. 6. If the designated components contain characteristic, deteriorating components, the operation of Fig. 9. In other words, the Fig. 6 and Fig. 9 shown operating sequences are executed in parallel.

[0074] The information processor 30 such as the computer or the server required to calculate the deterioration rate and the remaining life, which collects and records information for the calculation, records calculation results to make a determination based on the calculation results, and the like, as shown in Fig. 9, is installed in the information management device 2. However, the information processor 30 may be connected to the information management device 2 or may be installed in any of the devices 3 connected to the information management device 2. The information processor 30 may be identical to the information processor 30 used for calculating the cumulative error rates and the sum of the cumulative error rates, determining, recording, and the like, as shown in Fig. 6 shown, is used, or may be different.

[0075] Fig. 10 is a block diagram showing the schematic configuration of a warehouse management system 1 according to another embodiment. In the Fig. In the warehouse management system shown in Figure 10, at least one first machine learning device 4-1 is connected to the information management device 2 described above.

[0076] The first machine learning device 4-1 includes a first learning unit 5-1, a state variable observation unit 6, and a result acquisition unit 7. The first learning unit 5-1 includes an error calculation unit 8 and a learning model update unit 9. A data recorder 10 is provided outside the first machine learning device 4-1 to record data with results (labeled data) therein.

[0077] The state quantity observation unit 6 of the first machine learning device 4-1 observes outputs of components, such as communication controllers and memories, which constitute the devices 3 connected to the information management device 2, and outputs of at least one of a variety of sensors provided in the devices 3 and a variety of sensors provided around the devices 3 as the state quantities of the devices 3, including the operating conditions, states, and the like of the designated components.

[0078] The result obtaining unit 7 of the first machine learning device 4-1 obtains information about an error occurring in the devices 3.

[0079] The first learning unit 5-1 learns a standard cumulative error rate curve (see, for example, the solid lines of the Fig. 2a and Fig. 2b), which represents a temporal variation of the cumulative failure rate of each type of designated components under a standard operating condition, and a first acceleration coefficient (see for example Fig. 3), which represents the acceleration ratio of the cumulative failure rate as a function of an operating condition with respect to the cumulative standard failure rate of each type of designated components under the standard operating condition, in conjunction with the observed state variables of the devices 3 and the obtained failure information on the devices 3.

[0080] An information processor 30 installed inside or outside the information management device 2 calculates the sum of cumulative failure rates with reference to the cumulative standard failure rate curve and with reference to the first acceleration coefficient of each type of designated components learned by the first learning unit 5-1, and operates in accordance with the Fig. 6 to make appropriate inventory management in view of an unsigned error.

[0081] As concrete examples of the operating condition, state, and the like of the designated component to be considered as the state quantity of the device 3, there are operating current, applied voltage, operating frequency, operating temperature, ambient temperature, temperature cycle, humidity, vibration, and the like. Depending on the specifications of the component, the state quantity to be observed by the state quantity observation unit 6 may include light output, rotational speed, acceleration, audible sound, atmospheric gas composition, voltage, and the like.

[0082] In addition to observing the state variables of the devices 3, the state variable observation unit 6 can process data in order to make the data usable in the first learning unit 5-1 when needed.

[0083] There are various machine learning methods. Although a machine learning method applicable to this embodiment is not specifically limited, in the example of Fig. 10 a data recorder 10 is provided outside the first machine learning device 4-1 to record labeled data to be used in supervised learning.

[0084] The first machine learning device 4-1 has a learning model to learn the cumulative standard failure rate curve and the first acceleration coefficient. To perform learning, the first machine learning device 4-1 calculates the error between information about the learning model and the obtained actual failure information, i.e., a result (label), and updates the learning model according to the error.

[0085] The data recorder 10 stores marked data obtained up to this point and provides the marked data to the error calculation unit 8. The marked data may be provided to the error calculation unit 8 via a memory card, a communication line, or the like.

[0086] Machine learning can be initiated from a state where no labeled data is available through reinforcement learning. However, supervised learning using the labeled data obtained up to that point has the advantage that learning can be performed in a relatively short time.

[0087] According to this embodiment, even if the cumulative failure rate and the acceleration coefficient of the cumulative failure rate are unknown or inaccurate depending on the operating condition, it is possible to appropriately perform inventory management depending on the type of designated components in consideration of a failure that shows no sign. Since the first machine learning device 4-1 has a series of learning processes, the cumulative failure rate and the acceleration coefficient of the cumulative failure rate have increased accuracy depending on the operating condition, and therefore inventory management can be performed more reliably.

[0088] Fig. 11 is a flowchart showing an example of a machine learning operation according to this embodiment. In the following description based on Fig. 11, the information management device 2 also functions as the information processor 30.

[0089] When starting machine learning, the state quantity observation unit 6 of the first machine learning device 4-1 collects and records output signals from components such as a communication controller and a memory, each constituting the device 3, and outputs from at least one of a variety of sensors provided in the device 3 and a variety of sensors provided around the device 3, as time variations in the state quantities of the device 3, including the operating condition, the state, and the like of the designated component (step S201).

[0090] The first machine learning device 4-1 determines whether the first learning unit 5-1 already has a learning model or not (step S202). If the first learning unit 5-1 has a learning model, the operation proceeds to step S204. If the first learning unit 5-1 does not have a learning model, the first machine learning device 4-1 generates an initial learning model from the tagged data stored in the data recorder 10 (step S203), and the operation proceeds to step S204.

[0091] In step S204, the information management device 2 calculates and stores a cumulative failure rate of each designated component, taking into account a first acceleration coefficient, with reference to the learning model provided in the first learning unit 5-1. The first machine learning device 4-1 obtains, through the information management device 2, failure information of each designated component in the inventory management (step S205).

[0092] The failure information is information about which of the designated components failed at what time. For example, the failure information can be obtained from the state variable of the device 3, which represents the state of the designated component. Alternatively, an operator can input information about the replacement of a failing component into the information management device 2 at the time of replacing the failing component, and the input information can be used as the failure information.

[0093] The first machine learning device 4-1 determines whether a failing designated component exists based on the obtained failure information (step S206). If it is determined that a failing designated component exists, actual data on the cumulative failure rate of the corresponding designated component is updated (step S207), and the updated actual data is added to the data recorder 10 (step S208). The error calculation unit 8 of the first machine learning device 4-1 calculates the error between the cumulative failure rate calculated from the learning model and the actual cumulative failure rate data (step S209). The learning model updating unit 9 of the first machine learning device 4-1 updates the learning model to reduce the error (step S211).If it is determined in step S206 that there is no failing designated component, the designated component has an increased actual operating time until its failure, so the first machine learning device 4-1 updates the actual data on the cumulative failure rate of each designated component (step S210), and the operation proceeds to step S211.

[0094] After completing the learning model update in step S211, the first machine learning device 4-1 determines whether or not a machine learning end command is present (step S212). If a machine learning end command is present, the machine learning is completed. If a machine learning end command is missing, the operation returns to step S201 to continue the machine learning.

[0095] As described above, by repeating a series of processes from step S201 to step S212, the first learning unit 5-1 can learn the cumulative standard error rate curve of each type of designated components with the same specifications and with the first acceleration coefficient of the cumulative error rate depending on the operating state with high accuracy.

[0096] A machine learning device (hereinafter referred to as a second machine learning device 4-2) of another warehouse management system having the same configuration as that shown in Fig. The warehouse management system 1 shown in Figure 10 can observe outputs from components such as communication controllers and memories that constitute devices 3 connected to an information management device 2, and outputs from at least one of a variety of sensors provided in the devices 3 and a variety of sensors provided around the devices 3 as the state variables of the devices 3, including the operating conditions, properties, and the like of characteristic deterioration components. The second machine learning device 4-2 (not shown) can generate a characteristic standard change curve of the characteristic deteriorating component (see Fig. 7) and learn a second acceleration coefficient representing the acceleration ratio of the characteristic deterioration rate depending on an operating condition with respect to the standard characteristic change curve by machine learning in conjunction with the observed state variables of the devices 3 and characteristic change information about the characteristic deteriorating components obtained as a part of the state variables.

[0097] As concrete examples of the operating conditions, properties, and the like of the characteristic deteriorating component to be observed as the state quantity of the device 3, there are operating current, applied voltage, operating frequency, operating temperature, ambient temperature, temperature cycle, humidity, vibration, and the like. Depending on the type of component, the state quantity to be observed by the state quantity observation unit 6 may include a torque output characteristic, a light output characteristic, a heat generation characteristic, a rotation speed, an acceleration, an audible sound, an atmospheric gas composition, a voltage, and the like.

[0098] As with the first machine learning device 4-1 described above, the second machine learning device 4-2 includes a second learning unit 5-2, a state variable observation unit 6, a result acquisition unit 7, and a data recorder 10. The second learning unit 5-2 includes an error calculation unit 8 and a learning model update unit 9.

[0099] The information management device 2 can determine the remaining life of each characteristic deteriorating component by executing the Fig. 9, referring to the standard characteristic change curve and the second acceleration coefficient of each type of characteristic deteriorating components learned by the second learning unit 5-2 of the second machine learning device 4-2. Furthermore, even if the standard characteristic change curve and the second acceleration coefficient of the characteristic deteriorating components are unknown or inaccurate, the information management device 2 can appropriately perform inventory management and preventive maintenance using a series of learning processes conducted by the second machine learning device 4-2.

[0100] Fig. Figure 12 is a flowchart showing an example of a process for learning a standard characteristic change curve and a second acceleration coefficient of a characteristic deteriorating component. In the following description based on Fig. 12, however, the information management device 2 also functions as the information processor 30.

[0101] The second machine learning device 4-2 has a learning model for learning a cumulative standard failure rate curve and a second acceleration coefficient. To perform learning, the second machine learning device 4-2 calculates the error between the learning model and the actual characteristic change information obtained by the result acquisition unit 7 of the second machine learning device 4-2 and updates the learning model according to the error.

[0102] In the example of Fig. 12, when starting the machine learning, the second machine learning device 4-2 observes and records the outputs of components such as a communication controller and a memory constituting each device 3, and the outputs of at least one of a variety of sensors provided in each device 3 and a variety of sensors provided around the device 3, through the state quantity observation unit 6 as a time variation in the state quantity of the device 3, including the operating condition, operating characteristic, and the like of a characteristic deteriorating component (step S301).

[0103] The second machine learning device 4-2 determines whether a characteristic deterioration width of the characteristic deteriorating component, that is, a characteristic change from a predetermined time before a certain time, which is the characteristic change information of the characteristic deteriorating component, is larger than a predetermined value (step S302). This is because if the characteristic deterioration width is not sufficiently larger than a measurement error, the deterioration rate calculated as a division of the characteristic deterioration width by the operation time will have an increased error, thereby deteriorating the accuracy of the learning. If the characteristic deterioration width of the characteristic deteriorating component is smaller than the predetermined value, step S301 is executed again.

[0104] When it is determined that the characteristic deterioration width of the characteristic deteriorating component is larger than the predetermined value, the second machine learning device 4-2 adds data with a result (a characteristic deterioration width) (more specifically, the state quantity of the device 3, including the operating condition, the operating characteristic, and the like of the characteristic deteriorating component) to the data recorder 10 (step S303).

[0105] Next, the second machine learning device 4-2 determines whether the second learning unit 5-2 already has a learning model or not (step S304). If the second learning unit 5-2 has a learning model, the operation proceeds to step S306. If the second learning unit 5-2 does not have a learning model, the second machine learning device 4-2 generates an initial learning model from the tagged data stored in the data recorder 10 (step S305), and the operation proceeds to step S306.

[0106] In step S306, the information management device 2 calculates and plots a deterioration rate of the characteristic deteriorating component under a standard operating condition using a second acceleration coefficient with reference to the learning model provided in the second learning unit 5-2. Further, the information management device 2 extracts and plots a standard characteristic change curve having the same deterioration rate after calculation as the characteristic of the characteristic deteriorating component with reference to the learning model (step S307), and determines whether or not there is a history of extracting a standard characteristic change curve (step S308).If an extraction history exists, the error calculation unit 8 calculates the error between the deterioration rate calculated in step S306 and a deterioration rate estimated from the most recently extracted standard change characteristic curve (step S309) and updates the learning model to reduce the error (step S310). On the other hand, if it is determined in step S308 that there is no extraction history, the operation returns to step S301 and machine learning continues.

[0107] After completing the learning model update in step S310, the second machine learning device 4-2 determines whether or not a machine learning end command is present (step S311). If a machine learning end command is present, machine learning is completed. If a machine learning end command is missing, the operation returns to step S301 to continue machine learning.

[0108] As described above, by repeating a series of processes between steps S301 and S311, the second learning unit 5-2 can learn the standard characteristic change curve of each type of characteristic deteriorating component with the same specifications and the second acceleration coefficient with respect to lifetime consumption depending on the operating condition with high precision. Furthermore, the information management device 2 can more appropriately perform inventory management and preventive maintenance by calculating the replacement time of the characteristic deteriorating component with reference to the characteristic change curve and the second acceleration coefficient learned by the second learning unit 5-2 of the second machine learning device 4-2.

[0109] Note that this machine learning method is not limited to supervised learning, which is Fig. 12 is shown as an example.

[0110] In addition, a Fig. 13, according to another embodiment, search for an error character of a designated component whose error character is unknown. As shown in Fig. As shown in Figure 13, a third machine learning device 4-3 is connected to an information management device 2. The third machine learning device 4-3 includes a third learning unit 5-3 and a state variable observation unit 6. The third learning unit 5-3 has a learning model 11. Fig. Figure 14 is a diagram schematically showing the passage of time until an unknown character is found.

[0111] The Fig. The third machine learning device 13 shown in FIG. 13 observes output signals from components such as a communication controller and a memory constituting each device 3 having a designated component, and outputs from at least one of a plurality of sensors provided in the device 3 and a plurality of sensors provided around the device 3 through the state quantity observation unit 6 as the state quantity of the device 3, including the characteristic, the state, and the like of the designated component.Whenever a result indicating the occurrence of a failure in the designated component is received, except for a characteristic deteriorating component provided in the device 3, the third learning unit 5-3 of the third machine learning device 4-3 collects state quantity data of each type of failing component with the same specifications between the time a predetermined time precedes the occurrence of the failure and the time of the occurrence of the failure outside the state quantities of the devices 3, including the characteristics, states, and the like of the failing components. The third learning unit 5-3 extracts a feature different from the state quantities of non-failing components from the collected state quantity data to search for a failure sign.

[0112] Although the learning method is not specifically restricted, unsupervised learning is suitable for searching for an unknown error character from a large data set. As in Fig. 13, unsupervised learning does not require obtaining results (i.e., error information). A large number of state variables of the devices 3 related to the operating conditions, including the states and characteristics of the designated components, are input, and the third learning unit 5-3 of the third machine learning device 4-3 establishes a learning model related to the distribution of the input data. The third learning unit 5-3 performs a comparison between the constructed learning model and the collected state variable data (input data) to detect a variation in the state variable (a fault sign).

[0113] Unsupervised learning learns the distribution of input data and extracts essential configuration behind the data without requiring corresponding output data (training data), and can therefore find out an unknown error character.

[0114] The state quantity of the device 3 and its variation, including the characteristics, state, and the like of the failing component, observed by the state quantity observation unit 6 of the third machine learning device 4-3, vary depending on the type of components. As concrete examples of the state quantity of the device 3, there are an electrical input and output characteristic (e.g., change in resistance, etc.), a light output characteristic (e.g., decrease in electro-optical conversion efficiency, etc.), a heat generation characteristic (e.g., increase in heating power, etc.), a torque output characteristic (e.g., decrease in torque, etc.), an acoustic output characteristic (e.g., decrease in acoustic output, etc.), temperature (e.g., temperature increase, etc.), vibration (e.g., increase in vibration, etc.), displacement (e.g.,Occurrence of displacement and the like) and the like.

[0115] Fig. 15 is a flowchart showing an example of a process in which the third machine learning device 4-3 extracts a feature different from the state variables of non-failing components from collected state variable data before the occurrence of a failure to detect a failure sign. In the following description based on Fig. 15, the information management device 2 also functions as an information processor 30.

[0116] The third machine learning device 4-3 learns the distribution of state variable data and a model regarding the regularity of the data. The third machine learning device 4-3 detects a change in the state variable data before the occurrence of a fault, in other words, a sign before the occurrence of a fault, and outputs a learning result to the information management device 2.

[0117] In the example of Fig. 15, the state quantity observation unit 6 of the third machine learning device 4-3, when starting the machine learning, observes outputs of components such as a communication controller and a memory that constitute each device 3, and outputs of at least one of a variety of sensors provided in the device 3 and a variety of sensors provided around the device 3, as the state quantity of the device 3, including the characteristic, state, and the like of a designated component (step S401).

[0118] If it is determined in step S402 that a learning model related to the distribution of state variables is not established, the third learning unit 5-3 of the third machine learning device 4-3 next builds or updates a model (learning model) related to the distribution of state variables (input data) and the regularity of the data (step S403). If it is determined in step S402 that a learning model has been built by batch learning, sufficient online learning, or the like, the operation proceeds to step S404 without executing step S403.

[0119] Subsequently, if it is determined that a failure has occurred due to the state quantity observed by the state quantity observation unit 6 and the like (step S404), the third machine learning device 4-3 determines whether a failing component is a characteristic deteriorating component, without requiring the search for its failure sign (step S405). If the failing component is not a characteristic deteriorating component, the third machine learning device 4-3 extracts state quantity data of the device 3 having the failing component between a predetermined time preceding the failure occurrence and the time of the failure occurrence (step S406).The third machine learning device 4-3 detects the difference by comparing between the extracted state quantity and the learning model (step S407) and outputs the difference to the information management device 2 or another output device (not shown) as a failure flag from the failing component (step S408).

[0120] Next, the third machine learning device 4-3 determines whether or not a machine learning end command is present (step S409). If a machine learning end command is present, the machine learning is completed.

[0121] If a machine learning end command is missing, the operation returns to step S401 to continue machine learning.

[0122] If it is determined in step S404 that no failure has occurred in the designated component, or the failing component is determined in step S405 to be a characteristic deteriorating component, the operation returns to step S401 to continue machine learning.

[0123] As described above, by repeating a series of processes between steps S401 and S409, the third learning unit 5-3 can output data that certainly or probably indicates the failure character of the designated components whose failure character is unknown or uncertain to the information management device 2 in a repeated manner.

[0124] The third machine learning device 4-3 according to this embodiment additionally has, for example, the following functions and also outputs data indicating the failure character of the designated components with certainty or probability.

[0125] The third learning unit 5-3 of the third machine learning device 4-3 extracts a feature different from state variables of non-failing components from collected state variable data before the occurrence of a failure.When it is determined that the different feature is a failure sign of a designated component, the third learning unit 5-3 learns a failure occurrence probability distribution relative to the time after the occurrence of the failure sign and the dependency of the failure occurrence probability distribution relative to the time in an operating state associated with the state of the failure sign, which includes an occurrence range and an occurrence frequency of the failure sign, the state quantity data of the device related to the operating state of the failing component between the occurrence of the failure sign and the occurrence of the failure, and an actual time of the occurrence of the failure observed by the state quantity observation unit 6.The third learning unit 5-3 outputs preventive maintenance information to recommend replacement of the designated component with the failure sign (hereinafter referred to as a component showing the sign), and outputs failure occurrence prediction information about the component showing the sign with reference to the learning result until the designated component showing the sign actually fails. The third machine learning device 4-3 having the above functions is similar to the first machine learning device 4-1 shown in FIG. Fig. 10. The information management device 2 can manage the inventory quantity for the designated components for maintenance by taking into account the failure occurrence prediction information in addition to the failure rates and the times of replacing the components.

[0126] Fig. 16 is an example of a flowchart used to Fig. 15, and shows an additional operation flow in which, as described above, when a failure sign is seen, the third machine learning device 4-3 learns a failure occurrence probability distribution relative to the time after the failure sign appears, and the dependence of the failure occurrence probability distribution relative to the time in an operating state in association with the state of the failure sign, the state quantity of the device related to the operating state of the failing component between the occurrence of the failure sign and the failure occurrence, and an actual time of failure occurrence. The third machine learning device 4-3 recommends replacing the component showing the sign with reference to the learning result until the component actually showing the sign fails, and the learning result is reflected in the inventory management of the components showing the sign.In the following description based on . Fig. 16, the information management device 2 also functions as an information processor 30.

[0127] By inserting a series of processes from step S410 to step S424 of Fig. 16 between step S408 and step S409 of Fig. 15, stock management and preventive maintenance may be carried out appropriately on a component displaying the symbol. If the Fig. 16 is executed, the third learning device 4-3 has the same components as the one shown in Fig. 10 shown first learning device.

[0128] In Fig. 16 after step S408 of Fig. 15, the third machine learning device 4-3 records the difference between the state quantity data of the designated component from the time at which the predetermined time precedes the occurrence of the failure and the model (hereinafter referred to as an unsupervised learning model) regarding the distribution and regularity of state quantities obtained by unsupervised learning from state quantity data of non-failing designated components with the same specifications as the failing designated component (step S410). The third machine learning device 4-3 determines whether or not the difference is recognized as a failure sign of the failing designated component (step S411).If the difference is not recognized as a failure sign, the third machine learning device 4-3 determines whether the difference has a reproducibility of a predetermined ratio or more than one failure sign of the designated component with reference to information about the difference recorded in step S410 (step S412). If the difference has a reproducibility of the predetermined ratio or more, the third machine learning device 4-3 recognizes the extracted difference between the state quantity and the unsupervised learning model as a failure sign (step S413) and observes and records the state of the failure sign, including an occurrence range and occurrence frequency of the failure sign, and the state quantity of the device 3 with respect to the operating state of the failing designated component after the occurrence of the failure sign (step S414).

[0129] If the difference in step S412 does not have a reproducibility of the predetermined ratio or more and is not recognized as an error sign, the operation proceeds to step S409 of Fig. 15. If it is determined in step S411 that the difference is recognized as a fault sign of the failing component, the operation proceeds to step S414 of Fig. 16 continued.

[0130] Next, the third machine learning device 4-3 determines whether or not a fault has actually occurred in the component exhibiting the sign (step S415). If a fault has occurred, the third machine learning device 4-3 adds the state of the fault sign and the state quantity of the device 3, which have an operating time between the appearance of the fault sign and the actual fault as a result (mark), as marked data (step S416).

[0131] Furthermore, the third learning unit 5-3 of the third machine learning device 4-3 learns a failure occurrence probability distribution relative to time after the occurrence of the failure sign and the dependency of the failure occurrence probability distribution relative to time at an operating state through supervised learning in conjunction with the state of the failure sign, the device's state quantity data regarding the operating state of the failing component between the occurrence of the failure sign and the occurrence of the failure, and the operating time between the occurrence of the failure sign and the occurrence of the failure, and determines whether or not a supervised learning model already existed (step S417). If no supervised learning model exists, the third learning unit 5-3 generates an initial supervised learning model from data recorded in the data recorder 10 (step S418).If a supervised learning model exists in step S417, the operation proceeds to step S419 without executing step S418.

[0132] Subsequently, the third learning unit 5-3 calculates the error between a fault occurrence probability calculated by an error function provided in the supervised learning model and an actual fault occurrence probability value (step S419). The third learning unit 5-3 updates the supervised learning model to reduce the error (step S420), and the operation proceeds to step S409 of Fig. 15 continued.

[0133] If it is not determined in step S415 that the designated component exhibiting the failure sign is failing, the third learning unit 5-3 determines whether or not a supervised learning model exists (step S421). If no supervised learning model exists, the third learning unit 5-3 generates an initial supervised learning model from data recorded in the data recorder 10 (step S422) and calculates a temporal variation of the failure occurrence probability or a cumulative failure rate with reference to the supervised learning model (step S423). On the other hand, if it is determined in step S421 that a supervised learning model exists, the operation proceeds to step S423 without executing step S422.

[0134] After step S423, the third learning unit 5-3 outputs at least one of the preventive maintenance information to recommend the replacement of the component displaying the sign and failure occurrence prediction information required for the inventory management of the component displaying the sign to the information management device 2 based on the calculated time variation in the failure occurrence probability or the cumulative failure rate (step S424), and the operation proceeds to step S409 of Fig. 15. The cumulative failure rate can be calculated by integrating the probability of failure occurrence with respect to time.

[0135] As described above, it is by repeating steps S401 to S424, which are steps S410 to S424 of Fig. 16, added to steps S401 to S409 of Fig. 15, it is possible to find out the error character of the designated components whose error characters are unknown or uncertain. If the state of the designated component before the error is recognized as the error character by the additional operation sequence specified in Fig. 16, the learning is further repeatedly performed in conjunction with the state of the error symbol, the state variable of the device 3 related to the operating state of the component triggering the symbol, and the temporal variation of the error occurrence probability. With reference to the learning results, for example, the cumulative error rate and the acceleration coefficient of the designated component can be determined depending on the operating condition, as shown in Fig. 2b, can be calculated more accurately. This allows preventive maintenance and inventory management to be carried out more reliably.

[0136] Fig. 17 is a block diagram showing the schematic configuration of a warehouse management system 1 according to yet another embodiment. In this embodiment shown in the drawing, at least a fourth machine learning device 4-4 is connected to the above-described information management device 2.

[0137] The fourth machine learning device 4-4 includes a fourth learning unit 5-4, a state variable observation unit 6, a determination result acquisition unit 12, and a decision-making unit 15. The fourth learning unit 5-4 includes a reward calculation unit 13 and a value function update unit 14.

[0138] To be more specific, the state quantity observing unit 6 observes the state quantity of each device 3 connected to the information management device 2. The determination result obtaining unit 12 obtains a determination result regarding true and false of an estimation result regarding a failing component, that is, a failing portion of the device 3. The estimation result regarding the failing component is to estimate which component or part fails in the failing device 3.

[0139] The fourth learning unit 5-4 receives an output from each of the state quantity observing unit 6 and the determination result obtaining unit 12, and learns the estimation result of the failing component in the device 3 in association with the state quantity of the device 3 and the determination result regarding correct and incorrect of the estimation result regarding the failing component.

[0140] The decision-making unit 15 determines the estimation result of the failing component to be output from the fourth machine learning device 4-4 by referring to the learning result by the fourth learning unit 5-4.

[0141] Furthermore, the reward calculation unit 13 of the fourth learning unit 5-4 has a value function for determining the value of the estimation result of the failing component, and calculates a positive reward when the estimation result is correct, while calculating a negative reward when the estimation result is incorrect. The value function update unit 14 of the fourth learning unit 5-4 updates the value function according to the reward calculated by the reward calculation unit 13.

[0142] The fourth machine learning device 4-4 having the above configuration diagnoses a failure of the device 3 with respect to the learning result of the fourth learning unit 5-4.

[0143] Fig. 18 is a flowchart showing an example of the operation of the Fig. 17 shown warehouse management system 1. In this embodiment, which is shown in Fig. 18, the information management device 2 also functions as an information processor 30.

[0144] When starting machine learning, the state quantity observation unit 6 of the fourth machine learning device 4-4 observes the state quantity of the device 3 connected to the information management device 2, including the properties, states, and the like of components (step S501). The state quantity observation unit 6 processes the state quantity data as needed and outputs the state quantity data to the fourth learning unit 5-4.

[0145] The fourth learning unit 5-4 compares the state quantity data with a state quantity in a normal operation recorded in advance to determine whether a malfunction or failure has occurred in the device 3 (step S502). If it is determined that a malfunction or failure has occurred in the device 3, the fourth learning unit 5-4 extracts the state quantity of the device 3, including a history of the state quantity of the failing device 3, such as error information, characteristic change information, and the like, stored up to the time of the malfunction occurrence (step S503).The fourth learning unit 5-4 estimates a failing component from the extracted state variable by referring to a value function that is a learning result of the fourth learning unit 5-4 to determine the value of a missing component estimation result (step S504). Based on the failing component evaluation result, the decision-making unit 15 determines the missing component estimation result by referring to the learning result by the fourth learning unit 5-4 and outputs the failing component evaluation result to the information management device 2 (step S505).

[0146] After receiving information about the occurrence of a malfunction or error in a specific device 3, the information management device 2 next checks the failing component of the device 3 (step S506). The determination result acquisition unit 12 of the fourth machine learning device 4-4 receives the check result as a determination result and outputs the determination result to the reward calculation unit 13 (step S507). The failing component is replaced or the like to repair the failing device 3 (step S508).

[0147] Subsequently, the fourth learning unit 5-4 determines whether the actual failing component checked in step S506 agrees with the failing component estimation result output by the decision-making unit 15 (step S509). If the actual failing component agrees with the failing component evaluation result, the reward calculation unit 13 provides a positive reward (step S510). If the actual failing component does not agree with the failing component evaluation result, the reward calculation unit 13 provides a negative reward (step S511). The value function updating unit 14 updates a value function for determining the value of the failing component estimation result in accordance with the reward (step S512).Further, the fourth machine learning device 4-4 adds failure information about the failing device 3 to a failing information history for updating (step S513).

[0148] Next, the fourth machine learning device 4-4 determines whether or not a machine learning end command is present (step S514). If a machine learning end command is present, machine learning is completed. If a machine learning end command is absent, the operation returns to step S501 to continue machine learning. If it is determined in step S502 that no malfunction or malfunction has occurred in the device 3, the operation returns to step S501, and the fourth machine learning device 4-4 continues machine learning.

[0149] As described above, by repeating a series of processes from step S501 to step S514, the fourth learning unit 5-4 can learn the fault component estimation result on the device 3 in conjunction with the state quantity of the device 3, including the history of the state quantity of the device 3, and the determination result corresponding to the result of the evaluation of the failing component. Therefore, the fourth learning unit 5-4 can diagnose the fault in the device 3.

[0150] The error component estimation result output by the fourth machine learning device 4-4 may include not only a first candidate with the highest probability, but also a plurality of candidates listed in descending order of probability. In this case, the higher the candidate with which the actually failing component matches is ranked, the greater the positive reward provided by the reward calculation unit 13. If the actually failing component does not match any of the listed candidates, the reward calculation unit 13 provides a negative reward. Based on the rewards provided under such conditions, the value function update unit 14 updates the value function.

[0151] Fig. Figure 19 is a flowchart showing an example of the operation of listing a plurality of candidates as described above. Comparing Fig. 19 with Fig. 18, steps S505, S509 and S510 of Fig. 18 by steps S515, S516 and S517 of Fig. 19 replaced.

[0152] Enumerating a plurality of candidates for the fault component estimation result output by the fourth machine learning device 4-4 has the effect of further improving learning effects with a smaller number of errors. Furthermore, even if the first candidate is incorrect, the next candidate can be checked immediately, thus reducing the risk of the device taking a long time to recover due to an inability to identify a failing component.

[0153] The Fig. The warehouse management system 1 shown in Figure 17 learns a fault diagnosis capability through reinforcement learning, which updates a value function with the goal of maximizing a total reward in the future. However, another machine learning method can be used to learn the fault diagnosis capability instead of reinforcement learning.

[0154] In each of the embodiments described above, the first machine learning device 4-1, the second machine learning device 4-2, the third machine learning device 4-3, the fourth machine learning device 4-4, and the like are provided separately in the warehouse management system 1, but the machine learning devices do not necessarily constitute different hardware units. A single machine learning device may consist of at least one of the first to fourth machine learning devices 4-1 to 4-4.

[0155] In the Fig. 10, Fig. 13 and Fig. 17, the machine learning device (4-1, 4-3, and 4-4) is connected to the information management device 2 independently of the information management device 2, but the present invention is not limited thereto. At least one of the first to fourth machine learning devices 4-1 to 4-4 may be installed in the information management device 2, a server or a computer connected to the information management device 2, or one of the devices 3.

[0156] Fig. 20 is a block diagram showing the schematic configuration of a warehouse management system 1 according to still another embodiment.

[0157] In the Fig. In the warehouse management system 1 shown in Figure 20, at least one of a plurality of devices connected to an information management device 2 is a laser device 16 with laser oscillators 17, and at least one of the first to fourth machine learning devices 4-1 to 4-4 (e.g., a machine learning device 4) is connected to the information management device 2. In Fig. 20, solid arrows indicate information or electrical transmission. Dashed arrows indicate light beams. Alternate long and short dashed arrows indicate a portion of a cooling water flow for cooling the laser oscillators 17. A controller 19 supplies and receives information to and from each component and various sensors that make up the laser device, although this is omitted from the drawing for simplicity.

[0158] The following various state quantities can be output from components of the laser oscillators 17, laser power supplies 18, a controller 19, the optical systems 20 and the like constituting the laser device 16, from various sensors such as light intensity sensors 21, electric power sensors 22, flow rate sensors 24, temperature sensors 26, a temperature / humidity sensor 27 and the like provided in the laser device 16, and from various sensors such as a temperature / humidity sensor 27 provided around the laser device 16.

[0159] As examples of the various state variables to be output, there are a laser output intensity of at least one of the laser devices 16 and the laser oscillators 17, a laser output pulse condition of at least one of the laser devices 16 and the laser oscillators 17, a reflected light intensity fed back to at least one of the optical systems 20 and the laser device 16, at least one of an operating current and an operating voltage supplied from the laser power supply 18 to the laser oscillator 17, the temperature of a component constituting the laser device 16 including the laser oscillator 17, at least one of a temperature and / or a humidity within the laser device 16, at least one of a temperature and a humidity around the laser device 16, an acceleration applied to at least one of the laser device 16 and a component of the laser device 16,at least one of the type of auxiliary gas and a gas supply condition, at least one of a transmittance and an attenuation of the optical system 20, at least one of the installation position and height of the laser device 16, and the like.

[0160] In the machine learning device 4 connected to the information management device 2, a state quantity observation unit 6 observes at least one of the various state quantities described above as the state quantity of the laser device 16, and a determination result acquisition unit 12 receives error information and characteristic change information of each component constituting the laser device 16.Further, a learning unit 5 provided in the machine learning device 4 may learn cumulative failure rate curves and a first acceleration coefficient of components whose failure character is unknown to perform inventory management based on the sum of cumulative failure rates of the designated components with the same specifications, and may learn characteristic change curves and a second acceleration coefficient of a characteristic deteriorating component whose characteristic gradually deteriorates with operation to perform inventory management and preventive maintenance for the characteristic deteriorating component.The learning unit 5 can detect a failure sign of a component whose failure sign is unknown, and can learn a temporal variation of a failure occurrence probability after the detected failure sign occurs and the dependency of a component exhibiting the failure sign under an operating condition to perform inventory management and preventive maintenance of the component exhibiting the failure sign. The learning unit 5 can learn a value function to determine the value of a failure component estimation result to increase a failure diagnosis capability for identifying a failing component.

[0161] The laser device 16 includes a component whose failure rate and acceleration coefficient are difficult to accurately calculate by testing the component itself. For example, for a component that is susceptible to damage by reflected light, the failure rate of the component cannot be accurately calculated. It is necessary to calculate the failure rate of the component in a state where the component is installed in the laser device 16. In the warehouse management system 1 according to this embodiment, since the cumulative failure rate and acceleration coefficient of the component are learned by machine learning in a state where the component is installed in the laser device 16, it is possible to calculate the failure rate and acceleration coefficient of the component installed in the laser device 16 with high accuracy.Therefore, the warehouse management system according to the present invention can be advantageously applied to the laser device 16.

[0162] For example, the following configuration and method are applied to a fault diagnosis of the laser device 16.

[0163] The laser device 16 includes the laser oscillators 17, the light intensity sensor 21 for detecting a light output from the laser device 16, and the other light intensity sensors 21 each for detecting light outputs from the laser oscillators 17. The two laser oscillators 17 are in Fig. 20 is shown for brevity, but the number of laser oscillators 17 may be larger.

[0164] Further, the laser device 16 includes the electric current sensors 22 for detecting the power supplied from the laser power supplies 18 to the laser oscillators 17, the flow rate sensors 24 and water temperature sensors 25 for detecting the flow rate and temperature of the cooling water supplied from a chiller 23 for cooling water-cooled plates thermally connected to the laser oscillators 17, and the temperature sensors 26 for detecting the temperature of the laser oscillators 17.

[0165] The machine learning device 4 observes output values ​​of the above-described various sensors as state variables to receive error information and characteristic change information of each component constituting the laser device 16. Therefore, when an intended light output cannot be detected in response to an output command from the controller 19 of the laser device 16, it is possible to easily perform fault diagnosis to identify which of the laser oscillators 17, the laser power supplies 18, the cooling device 23, the optical systems 20, and the sensors has failed or deteriorated.

[0166] On the other hand, it is often difficult to perform fault diagnosis of the components without a sensor as described above, but the machine learning device 4 makes it possible to make a precise fault diagnosis using limited information from the various sensors through machine learning.

[0167] Fig. 21 is a block diagram showing the schematic configuration of a warehouse management system 1 according to still another embodiment.

[0168] In the Fig. In the warehouse management system 1 shown in FIG. 21, a plurality of information management devices 2 are connected to each other via a cloud 28. To each information management device 2, a plurality of devices 3 and at least one of the above-described first to fourth machine learning devices 4-1 to 4-4 (for example, a machine learning device 4) are connected. The machine learning devices 4 connected to the information management devices 2 share learning result information, such as the above-described learning model and the value function, through the cloud 28 to which the information management devices 2 are connected. The machine learning devices 4 connected to the individual information management device 2 share the learning result information, such as the above-described learning model and the value function, through the information management device 2 to which the machine learning devices 4 are connected.

[0169] Since the machine learning devices 4 share information such as the learning result information, it is possible to perform the learning in a shorter time and obtain more reliable learning results.

[0170] Fig. 22 is a block diagram showing the schematic configuration of a warehouse management system 1 according to still another embodiment.

[0171] In the Fig.In the warehouse management system 1 shown in FIG. 22, an automatic warehouse 29 and a plurality of devices 3 are connected to an information management device 2. In this case, orders for designated components are automatically made based on at least one of the inventory management information such as the appropriate inventory quantity of the designated components and preventive maintenance information about designated components and inventory information about the automatic warehouse 29, which are input via the information management device 2. When a designated component is supplied from the automatic warehouse 29 for maintenance or an ordered component is delivered to the automatic warehouse 29, the inventory information about the automatic warehouse 29 is updated, and the information management device 2 is notified of the updated inventory information.

[0172] Since the configuration automatically makes provisions for the necessary maintenance components in accordance with the inventory quantity of components, remaining service lives, and the like, it is possible to prevent errors in the arrangements. The configuration also has the advantage of returning the usage record of the maintenance components to the information management device 2 in real time.

[0173] The problems to be solved by any of the above-mentioned different warehouse management systems 1 are described below.

[0174] Patent Document 1 (Japanese Unexamined Patent Publication (Kokai) JP 2010-113672 A) discloses an inventory management method in which "a process of calculating the first required inventory quantity when components exhibiting failure signs fail at a first failure rate, a process of calculating the second required inventory quantity when components not exhibiting a failure sign still fail at a second failure rate, and a process of calculating the required inventory quantity of the components with the failure sign by summing the first and second required inventory quantities are performed." However, Patent Document 1 only describes a technique of calculating the required inventory quantity based on the number of components exhibiting the failure signs and the failure rate of the components not yet exhibiting a failure sign.A technique for calculating the appropriate stock quantity for components that fail suddenly without showing any failure sign is neither described nor suggested.

[0175] Patent Document 2 (Japanese Unexamined Patent Publication (Kokai) JP 2004-295667 A) discloses a replacement component prediction device having "an inventory quantity prediction unit for predicting the required inventory quantity for components based on information about an operating time acquired by an information acquisition unit and a failure rate recorded in a failure rate recorder." However, the technique described in Patent Document 2 does not consider that, in the inventory quantity prediction, the failure rate may vary depending on an operating condition.In addition, Patent Document 2 does not describe that when a failure rate is kept lower than a failure rate calculated from the recorded failure rate, the proportion of components used increases for a long time, so that the inventory amount needs to be increased, considering that the failure rate is likely to increase thereafter.

[0176] Patent Document 3 (Japanese Unexamined Patent Publication (Kokai) JP 2012-104058 A) discloses an inventory management device and method in which "a failure occurrence prediction unit that predicts a failure occurrence point in time from now and the number of components used at the time of failure occurrence based on a failure information history" predicts "a variation in the inventory quantity of components from now on, taking into account the time of occurrence of the predicted failure and the number of components used at the time of failure occurrence." However, since the technique described in Patent Document 3 does not predict the occurrence of a failure by considering information about operating conditions of the components that affect a failure rate of the components, the failure occurrence is predicted with low accuracy.

[0177] Even if an average variation of a failure rate of a component provided in a device depending on an operating time is known to some extent, the failure rate generally changes according to an operating condition, and therefore, it is extremely difficult to predict the time of component failure, especially when a failure character is uncertain. Even if the failure time is difficult to predict, excessive inventory, that is, stocking too many expensive components, is uneconomical. On the other hand, insufficient inventory prevents the prompt replacement of a failing component and causes a device to stop until the failing component is delivered.If the fixture is a production equipment, the excessive inventory will cause an increase in manufacturing cost, while the insufficient inventory will cause a production stoppage using the fixture and a significant delay in a production plan, resulting in a significant loss.

[0178] As described above, both excessive inventory and insufficient inventory pose problems. Therefore, a first task is to develop an inventory management system that can detect the appropriate inventory quantity, taking into account a variation in a failure rate depending on a component's operating condition, for components whose failure times are difficult to predict because the components fail without exhibiting a specific failure sign, or components that fail in a failure mode, exhibit no failure sign, or exhibit an unknown failure sign, from components shared among multiple devices.

[0179] Some components have characteristics that gradually deteriorate during operation. When the characteristic of a component reaches a predetermined reference value, the component must be replaced. For components with gradually deteriorating characteristics during operation (i.e., characteristic deteriorating components), an inventory management system must manage the appropriate inventory quantity, including maintenance components, taking into account the replacement time predicted by a characteristic deterioration rate with the effect of an operating condition. The characteristic deteriorating component is preferably replaced in a planned manner before the component fails to prevent a malfunction in a device before it occurs.

[0180] Furthermore, the failure rate and the influence of various operating conditions on the failure rate, or the characteristic deterioration rate and the influence of various operating conditions on the characteristic deterioration rate, are not always known. Sometimes they are unknown or have insufficient accuracy, even though they are known to a certain extent. In this case, manually creating information with high accuracy requires enormous man-hours. Therefore, a second challenge is to provide a warehouse management system that can generate the information and data required for appropriate warehouse management and preventive maintenance as automatically as possible through machine learning.

[0181] For defects that do not exhibit defect signs, defect signs may be overlooked. If a defect sign is found during an automatic search, information regarding the defect rate after the defect sign has occurred is preferably collected automatically and reflected in warehouse management and preventive maintenance to establish a more reliable and accurate warehouse management system.

[0182] Furthermore, when a device malfunctions, identifying a failing component that needs to be replaced facilitates device recovery in a shorter time and obtains the information and data necessary for proper inventory management and preventive maintenance by collecting component failure information without human assistance. Thus, a third objective is to provide a warehouse management system with such a diagnostic function.

[0183] As described above, one of the above embodiments proposes a warehouse management system 1 that can perform appropriate inventory management and preventive maintenance of components shared by devices 3 connected in an information management device 2, both for a failure that does not exhibit specific failure signs and for a failure that exhibits a sign of characteristic deterioration, taking into account the dependence of a failure rate on an operating condition. Another embodiment proposes a warehouse management system 1 that performs machine learning to enable more appropriate inventory management and preventive maintenance.

[0184] Furthermore, an inventory management system 1 according to this disclosure manages the inventory quantity for each kind of components having the same specifications used in a plurality of devices under inventory management and preventive maintenance based on the sum of cumulative failure rates, which is the sum of the cumulative failure rates of all the components, taking into account an acceleration coefficient depending on an operating state of each component, thereby enabling appropriate inventory management with high accuracy in consideration of a failure that does not have specific failure signs.

[0185] Regarding the characteristic deteriorating components, the remaining lifetime is estimated from a deterioration rate and an acceleration coefficient of the deterioration rate depending on an operating condition to perform inventory management, and replacement of the component is recommended as preventive maintenance before the component reaches the end of its lifetime.

[0186] Generally, information about a temporal variation in a cumulative failure rate and an acceleration coefficient of the cumulative failure rate depending on an operating condition, information about a temporal variation in a deterioration rate and an acceleration coefficient of the deterioration rate depending on an operating condition, and information indicating the relationship between a malfunction state and a failing portion of a device are not sufficiently collected in a database. Therefore, the warehouse management system 1 according to this disclosure repeatedly learns the information through machine learning and improves the reliability of the information, thereby enabling more appropriate warehouse management and preventive maintenance.

[0187] The present invention is described above using the preferred embodiments, but it will be obvious to those skilled in the art that the above embodiments may be modified, omitted, or added in various ways without departing from the scope of the present invention.

[0188] To achieve at least one of the objects of this disclosure, various aspects and effects thereof are provided as follows.

[0189] For example, a first aspect of this disclosure provides a warehouse management system including a plurality of devices and an information management device connected to the devices. The devices each have replaceable common components that can be shared between the devices, and the common components with the same specifications used in the devices are referred to as warehouse management components.

[0190] The inventory management system further includes an information processor for calculating the sum of cumulative failure rates, which is the sum value of the cumulative failure rates of all designated components with the same specifications used in the devices at a certain time, taking into account an acceleration that depends on an operating condition, with respect to a cumulative standard failure rate of each type of the designated components with the same specifications used in the devices under a standard operating condition, and for calculating the appropriate inventory quantity of the designated components at the certain time based on the calculated sum of the cumulative failure rates.

[0191] According to the first aspect, since the inventory quantity of components in the inventory management is managed based on the sum of the cumulative failure rates taking into account accelerations depending on the operating conditions, it is possible to perform appropriate inventory management taking into account failures that show no sign, thereby preventing the occurrence of a delay in recovery (maintenance) due to failure repair with insufficient inventory, while also preventing excessive inventory.

[0192] A second aspect of this disclosure provides the warehouse management system according to the first aspect, wherein, when the designated components having the same specifications as those used in the devices include at least one characteristic deteriorating component whose characteristic gradually deteriorates with operation and which needs to be replaced when the characteristic reaches a predetermined reference value, the information processor determines the timing of replacing the characteristic deteriorating component based on the characteristic of the characteristic deteriorating component at a certain time and a deterioration rate of the characteristic taking into account an acceleration depending on an operating condition, with respect to a standard deterioration rate of the characteristic of the characteristic,deteriorating component under standard operating conditions and calculates the appropriate storage quantity of the designated components taking into account the time of replacement.

[0193] According to the second aspect, with respect to a characteristic deteriorating component whose characteristic gradually deteriorates, the inventory quantity of the characteristic deteriorating component is managed in consideration of the replacement timing estimated based on the characteristic at a certain time and a characteristic deterioration rate taking into account an acceleration depending on an operating condition, thereby enabling more appropriate inventory management.

[0194] A third aspect of this disclosure provides the warehouse management system according to the second aspect, wherein the information processor outputs preventive maintenance information to recommend replacement of the characteristic deteriorating component before the characteristic of the characteristic deteriorating component reaches the predetermined reference value at which the characteristic deteriorating component needs to be replaced.

[0195] According to the third aspect, with respect to a component whose characteristic deteriorates significantly, replacement is recommended before the deterioration characteristic reaches a reference value at which the characteristic deteriorating component must be replaced, so that a failure in a device is prevented before it occurs, enabling appropriate preventive maintenance.

[0196] A fourth aspect of this disclosure provides the warehouse management system according to the first aspect, further comprising at least a first machine learning device connected to the information management device.

[0197] The first machine learning device includes: a state quantity observation unit for observing outputs of components constituting the devices and outputs of at least one of a plurality of sensors provided in the devices and a plurality of sensors around the devices as state quantities of the devices with respect to the operating state and the states of the designated components; a result acquisition unit for obtaining error information occurring in the devices; and a first learning unit for learning a cumulative standard error rate curve representing a time variation of the cumulative error rate of the designated component under the standard operating condition, and a first acceleration coefficient representing an acceleration ratio of the cumulative error rate depending on the operating condition with respect to the cumulative standard error rate of the designated component under the standard operating condition, in conjunction with the state quantity of the devices observed by the state quantity observation unit and with the error information observed by the result acquisition unit, wherein the information processor calculates the sum of the cumulative error rates by referring to the cumulative standard error rate curve of the designated component under the standard operating condition and to the first acceleration coefficient learned by the first learning unit.

[0198] According to the fourth aspect, the following effects are achieved. In most cases, the cumulative standard error rate and acceleration coefficient of a cumulative standard error rate depending on an operating condition are not known for all designated components. Manually calculating the cumulative standard error rate and the first acceleration coefficient requires enormous man-hours, whereas machine learning to obtain the cumulative standard error rate and the first acceleration coefficient essentially requires no man-hours. Even if the cumulative standard error rate and the first acceleration coefficient are known, the cumulative standard error rate and the first acceleration coefficient are often calculated by an acceleration life test using a relatively small number of samples and do not always have high accuracy.By repeating machine learning, the cumulative standard failure rate and the first acceleration coefficient can be obtained with higher accuracy, and the sum of the cumulative failure rates is calculated with increased accuracy, enabling more suitable warehouse management.

[0199] A fifth aspect of this disclosure provides the warehouse management system according to the second aspect or the third aspect, further comprising at least one second machine learning device connected to the information management device.

[0200] The second machine learning device includes: a state quantity observing unit for observing an output from a component constituting the device and from at least one of a plurality of sensors provided in the device and a plurality of sensors provided around the device as a state quantity of the device in relation to the operating state and the characteristic of the characteristic deteriorating component; a result obtaining unit for obtaining characteristic change information regarding the characteristic of the characteristic deteriorating component; and a second learning unit for learning a standard characteristic change curve representing a temporal change in the characteristic of the characteristic deteriorating component under the standard operating condition, and a second acceleration coefficient representing an acceleration ratio of the deterioration rate depending on the operating condition with respect to the standard characteristic change curve of the characteristic deteriorating component under the standard operating condition in conjunction with the state variable of the device observed by the state variable observation unit and the characteristic change information observed by the result acquisition unit, wherein the information processor estimates the timing of replacing the characteristic deteriorating component by referring to the standard characteristic change curve of the characteristic deteriorating component and the second acceleration coefficient learned by the second learning unit.

[0201] According to the fifth aspect, the following effects are obtained. Regarding a component whose characteristic deteriorates due to the operation of a device, a standard characteristic change curve and a second acceleration coefficient of the component are often calculated through an acceleration life test using a relatively small number of samples. In most cases, a deterioration rate and acceleration coefficients of the deterioration rate depending on various operating conditions are not known with high accuracy. However, by repeating machine learning, the characteristic change curve under a standard operating condition and the acceleration factor depending on the operating condition can be obtained with higher accuracy, thereby enabling more appropriate inventory management.

[0202] The sixth aspect of this disclosure provides the warehouse management system according to the first or fourth aspect, further comprising at least one third machine learning device connected to the information management device.

[0203] The third machine learning device includes: a state quantity observation unit for observing outputs of components constituting the devices and outputs of at least one of a plurality of sensors provided in the device and a plurality of sensors provided around the device as state quantities of the devices with respect to the operating state and characteristics of the designated components; and a third learning unit that, each time it receives a result indicating the occurrence of a failure in the designated components other than the characteristic deteriorating component, collects state quantity data on the devices with the failing designated components observed by the state quantity observing unit based on the type of the designated components with the same specifications between the time a predetermined time precedes the occurrence of the failure and the time of the occurrence of the failure, and extracts a feature from the collected pre-failure data that is different from the state quantities of the non-failing designated components to search for a failure sign.

[0204] According to the sixth aspect, a fault sign can be detected through unsupervised learning. Unsupervised learning generally learns the distribution of input data and extracts an essential configuration behind the data without requiring corresponding output data (training data). Therefore, by collecting state variable data on devices with failing designated components between a predetermined time before the occurrence of a fault and the time of the fault occurrence as the input data for unsupervised learning, an unknown fault sign for the failing designated components can be found.

[0205] A seventh aspect of this disclosure provides the warehouse management system according to the sixth aspect, wherein, if the extracted feature is recognized as the fault sign, the third learning unit learns a fault occurrence probability distribution with respect to time and the dependence of the fault occurrence probability distribution with respect to time of the operating condition in connection with the state of the fault sign, the state variable data of the device relating to the operating state of the failing designated component, between the occurrence of the fault sign and the occurrence of the fault, and an operating time between the occurrence of the fault sign and the actual occurrence of the fault, the third learning unit, until the designated component displaying the failure sign actually fails, outputs preventive maintenance information to recommend the replacement of the designated component displaying the failure sign, and outputs predictive failure information about the designated component displaying the failure sign with reference to a learning result, and the information processor estimates the appropriate inventory quantity for the designated components taking into account the failure occurrence prediction information.

[0206] According to the seventh aspect, when an unknown failure sign is detected, the time of failure occurrence can be estimated through machine learning. Therefore, it is possible to prepare a maintenance component and replace a failing component before the failure occurs, taking the failure sign into account, thus enabling more appropriate inventory management.

[0207] An eighth aspect of this disclosure provides the warehouse management system according to any one of the first to seventh aspects, further comprising at least a fourth machine learning device connected to the information management device.

[0208] The fourth machine learning device includes: a state quantity observation unit for observing outputs of components constituting the devices and outputs of at least one of a plurality of sensors provided in the device and a plurality of sensors provided around the device as state quantities of the devices with respect to the operating state and states or characteristics of the designated components; a determination result obtaining unit for obtaining a determination result related to correct and incorrect estimation results regarding a failing designated component in the device having the failing designated component; a fourth learning unit for receiving an output from the state quantity observation unit and an output from the determination result acquisition unit, and for learning the estimation result regarding the failing designated component in conjunction with the state quantity of the device and the estimation result regarding the failing designated component obtained by the determination result acquisition unit; and a decision-making unit for determining the estimation result regarding the failing designated component to be output by the fourth machine learning device with reference to the learning result by the fourth learning unit, wherein the fourth learning unit comprises a value function for determining the value of the estimation result with respect to the failing designated component, and the fourth learning unit comprises a reward calculation unit for delivering a positive reward if the estimation result is correct, whereas a negative reward is delivered if the estimation result is incorrect, and a value function updating unit for updating the value function in accordance with the reward, and the information processor performs a fault diagnosis of the devices with reference to a learning result of the fourth learning unit.

[0209] According to the eighth aspect, when a device fails, a failing component in the failing device can be estimated through repeated machine learning. Thus, the failing device can be restored in a short time. Furthermore, failure information about the device can be obtained automatically without human assistance.

[0210] A ninth aspect of this disclosure provides the warehouse management system according to the eighth aspect, wherein the estimation result regarding the failing designated component to be output by the fourth machine learning device includes not only a first candidate having the highest probability, but also a plurality of candidates listed in decreasing order of probability, and wherein, The higher the candidate is ranked among the candidates with which the actual failing designated component matches, the greater the positive reward provided by the reward calculation unit, whereas the reward calculation unit provides the negative reward if the actual failing component does not match any of the listed candidates.

[0211] According to the ninth aspect, higher learning effects can be obtained with a small number of errors. Furthermore, even if a first candidate is incorrect, the next candidate can be checked immediately, thus reducing the probability that a long time will be required to recover the device because a failing component cannot be identified.

[0212] A tenth aspect of this disclosure provides the warehouse management system according to any one of the fourth to ninth aspects, wherein at least one of the devices connected to the information management device is a laser device having a laser oscillator, at least one machine learning device of the first machine learning device, the second machine learning device, the third machine learning device and the fourth machine learning device is connected to the information management device and at least one of a laser output intensity of at least one of the laser device and the laser oscillator, a laser output pulse condition of at least one of the laser device and the laser oscillator, a reflected light intensity returning to at least one element from an optical system constituting the laser device or the laser oscillator, at least one of an operating current and an operating voltage supplied from a laser power supply from the laser device to the laser oscillator, the temperature of a component constituting the laser device including the laser oscillator, at least one of the temperature and humidity within the laser device, at least one of the temperature and humidity around the laser device, and an acceleration applied to at least one of the laser devices and a component of the laser device,at least one value from the type of assist gas and a gas supply condition, at least one value from the transmittance and attenuation of the optical system, and at least one value from the installation position and height of the laser device is output as the state variable of the laser device from a component constituting the laser device or from a plurality of sensors provided in the laser device and a plurality of sensors provided around the laser device, and the at least one machine learning device observes the state variable.

[0213] According to the tenth aspect, since the cumulative error rate and the acceleration coefficient of a component are learned in a state where the component is installed in the laser device, it is possible to calculate the error rate and the acceleration coefficient of the installed component in the laser device with high precision.

[0214] An eleventh aspect of this disclosure provides the warehouse management system according to any one of the fourth to tenth aspects, wherein a plurality of information management devices are interconnected via a cloud, the plurality of devices and at least one machine learning device is connected to each of the information management devices and the machine learning devices connected to the information management devices share learning result information through the information management device or the cloud to which the information management devices are connected.

[0215] According to the eleventh aspect, since the machine learning devices share information on learning results such as a learning model and a value function, it is possible to perform learning to obtain information such as a cumulative failure rate or an acceleration coefficient dependent on an operating condition and a characteristic deterioration rate required to estimate the inventory quantity of designated components in a shorter time.

[0216] A twelfth aspect of this disclosure provides the warehouse management system according to any one of the first to eleventh aspects, further comprising an automated warehouse connected to the information management device.

[0217] The information processor automatically makes provisions for the designated component based on at least one of the inventory management information including the appropriate inventory quantity of the designated components and preventive maintenance information for the designated components and inventory information for the automatic warehouse input by the information management device.

[0218] According to the twelfth aspect, since the information processor automatically makes arrangements for the necessary maintenance components, it is possible to prevent an error in the arrangements, and the usage record of a maintenance component is returned to the information management device in real time.

Claims

[1] Warehouse management system (1), comprising: a plurality of devices (3); an information management device (2) for managing information to be supplied to or received from the devices (3), wherein the devices (3) each have replaceable common components that can be shared between the devices (3), wherein the common components with the same specifications that are used in the devices (3) are referred to as components under inventory management, wherein the information management device (2) is arranged to: Collecting and storing state variables of the devices (3); Determine whether component operating states indicate that the components have failed or not; if components have failed, replacing the failed components and updating a stock quantity of the designated components; if no components have failed, calculating the cumulative failure rate of each of the designated components with reference to data on the cumulative failure rate of the designated components under a standard operating condition and with reference to data on an acceleration coefficient of the cumulative failure rate depending on an operating condition, checking a delivery time for each designated component, calculating an estimated cumulative failure rate as a cumulative failure rate after elapse of a delivery time from a current time with reference to the delivery time of each designated component, calculating a sum of the estimated cumulative failure rates by adding the estimated cumulative failure rates; and determining whether or not an inventory quantity of each type of the designated components is greater than or equal to the sum of estimated cumulative failure rates plus a margin;and ordering additional components if the inventory quantity is not greater than or equal to the sum of estimated cumulative defect rates plus the margin, where the margin comprises a predetermined number which is increased for longer delivery times and reduced prices; [2] A warehouse management system (1) according to claim 1, wherein the designated components having the same specifications as those used in the devices (3) include at least one characteristic deteriorating component having a characteristic that gradually deteriorates with operation and that needs to be replaced when the characteristic reaches a predetermined reference value, wherein the information processor (30) determines the timing of replacing the characteristic deteriorating component based on the characteristic of the characteristic deteriorating component at a certain time and a deterioration rate of the characteristic, taking into account an acceleration depending on an operating condition with respect to a standard deterioration rate of the characteristic of the characteristic,deteriorating component under standard operating conditions and calculates the appropriate storage quantity of the designated components taking into account the time of replacement. [3] The warehouse management system (1) according to claim 2, wherein the information processor (30) outputs preventive maintenance information to recommend replacing the characteristic deteriorating component before the characteristic of the characteristic deteriorating component reaches the predetermined reference value at which the deteriorating component must be replaced. [4] Warehouse management system (1) according to claim 1, further comprising: at least one first machine learning device (4-1) connected to the information management device (2), wherein the first machine learning device (4-1) comprises: a state quantity observation unit (6) for observing outputs of components constituting the devices (3) and outputs of at least one of a plurality of sensors provided in the devices (3) and a plurality of sensors provided around the devices (3) as state quantities of the devices (3) with respect to the operating state and the states of the designated components; a result obtaining unit (7) for obtaining error information occurring in the devices (3); and a first learning unit (5-1) for learning a cumulative standard error rate curve representing a time variation of the cumulative error rate of the designated component under the standard operating condition, and a first acceleration coefficient representing an acceleration ratio of the cumulative error rate depending on the operating condition with respect to the cumulative standard error rate of the designated component under the standard operating condition, in conjunction with the state quantity of the devices (3) observed by the state quantity observation unit (6) and the observed error information by the result acquisition unit (7), wherein the information processor (30) calculates the sum of the cumulative error rates with reference to the cumulative standard error rate curve of the designated component under the standard operating condition and with reference to the first acceleration coefficient learned by the first learning unit (5-1). [5] Warehouse management system (1) according to claim 2 or 3, further comprising: at least one second machine learning device (4-2) connected to the information management device (2), wherein the second machine learning device (4-2) comprises: a state quantity observation unit (6) for observing an output from a component constituting the device (3) and the output from at least one of a plurality of sensors provided in the device (3) and a plurality of sensors provided around the device (3) as a state quantity of the device (3) with respect to the operating condition and the characteristic of the characteristic deteriorating component; a result obtaining unit (7) for obtaining characteristic change information on the characteristic of the characteristic deteriorating component; and a second learning unit (5-2) for learning a standard characteristic change curve representing the temporal change in the characteristic of the characteristic deteriorating component under the standard operating condition, and a second acceleration coefficient representing an acceleration ratio of the deterioration rate depending on the operating condition with respect to the standard characteristic change curve of the characteristic deteriorating component under the standard operating condition, in conjunction with the state quantity of the device (3) observed by the state quantity observation unit (6) and the characteristic change information acquired by the result acquisition unit (7), wherein the information processor (30) estimates the timing of replacing the characteristic deteriorating component with reference to the standard characteristic change curve of the characteristic deteriorating component and the second acceleration coefficient learned by the second learning unit (5-2). [6] Warehouse management system (1) according to claim 1 or 4, further comprising: at least one third machine learning device (4-3) connected to the information management device (2), wherein the third machine learning device (4-3) comprises: a state quantity observation unit (6) for observing outputs of components constituting the devices (3) and outputs of at least one of a plurality of sensors provided in the devices (3) and a plurality of sensors provided around the devices (3) as state quantities of the devices (3) with respect to the operating state and characteristics of the designated components; and a third learning unit (5-3) that, whenever it receives a result indicating the occurrence of a failure in the designated components other than the characteristic deteriorating component having a property gradually deteriorating during operation, collects state quantity data about the devices (3) having the failing designated components observed by the state quantity observation unit (6) based on the type of the designated components with the same specifications between the time a predetermined time precedes the occurrence of the failure and the time of the occurrence of the failure, and extracts a feature from the collected data before the failure that is different from the state quantities of the non-failing designated components to search for a failure sign. [7] Warehouse management system (1) according to claim 6, wherein, when the extracted feature is recognized as the fault sign, the third learning unit (5-3) learns a fault occurrence probability distribution with respect to time and the dependence of the fault occurrence probability distribution with respect to time on the operating state in connection with the state of the fault sign, the state quantity data of the device (3) relating to the operating state of the failing designated component between the occurrence of the fault sign and the occurrence of the fault, and an operating time between the occurrence of the fault sign and the actual occurrence of the fault, wherein, until the designated component exhibiting the failure sign actually fails, the third learning unit (5-3) outputs preventive maintenance information to recommend replacing the designated component exhibiting the failure sign, and outputs failure occurrence prediction information about the designated component exhibiting the failure sign with reference to a learning result, and wherein the information processor (30) estimates the appropriate storage quantity for the designated components taking into account the failure occurrence prediction information. [8] Warehouse management system (1) according to one of claims 1 to 7, further comprising: at least one fourth machine learning device (4-4) connected to the information management device (2), wherein the fourth machine learning device (4-4) comprises: a state quantity observation unit (6) for observing outputs of components constituting the devices (3) and outputs of at least one of a plurality of sensors provided in the devices (3) and a plurality of sensors provided around the devices (3) as state quantities of the devices (3) with respect to the operating condition and states or characteristics of the designated components; a determination result obtaining unit (12) for obtaining a determination result regarding a correct or incorrect estimation result regarding a failing designated component in the device (3) having the failing designated component; a fourth learning unit (5-4) for receiving an output from the state quantity observation unit (6) and an output from the determination result acquisition unit (12) and for learning the estimation result regarding the failing designated component in conjunction with the state quantity of the device (3) and the estimation result regarding the failing designated component obtained by the determination result acquisition unit (12); and a decision-making unit (15) for deciding the estimation result to be output by the fourth machine learning device (4-4) with respect to the failing designated component with reference to the learning result by the fourth learning unit (5-4), wherein the fourth learning unit (5-4) has a value function for determining the value of the estimation result with respect to the failing designated component, and the fourth learning unit (5-4) comprises a reward calculation unit (13) for delivering a positive reward if the estimation result is correct, whereas a negative reward is delivered if the estimation result is incorrect, and a value function updating unit (14) for updating the value function in accordance with the reward, and the information processor (30) performs a fault diagnosis of the devices (3) with reference to a learning result of the fourth learning unit (5-4). [9] The warehouse management system (1) according to claim 8, wherein the estimation result regarding the failing designated component to be output by the fourth machine learning device (4-4) includes not only a first candidate having the highest probability but also a plurality of candidates listed in descending order of probability, and the reward calculation unit (13) provides a larger positive reward the higher the candidate among the candidates with which the actual failing designated component coincides, whereas when the actual failing component does not coincide with any of the listed candidates, the reward calculation unit (13) provides the negative reward. [10] Warehouse management system (1) according to one of claims 4 to 9, wherein at least one of the devices (3) connected to the information management device (2) is a laser device (16) with a laser oscillator (17) and of a laser output intensity from the laser device (16) and / or the laser oscillator (17), a laser output pulse state from the laser device (16) and / or the laser oscillator (17), a reflected light intensity returning to an optical system (20) forming a component of the laser device (16) and / or the laser device (16), at least one, an operating current and / or an operating voltage supplied to the laser oscillator (17) by a laser power supply (18) forming a component of the laser device (16), the temperature of a component forming a component of the laser device (16), including the laser oscillator (17), at least one of the temperature and / or humidity inside the laser device (16), the temperature and / or humidity around the laser device (16), an acceleration exerted on the laser device (16) and / or a component of the laser device (16), the type of auxiliary gas and / or a gas supply state, the transmittance and / or the attenuation of the optical system (20), and an installation position and / or the height of the laser device (16) from a component forming a component of the laser device (16),or at least one of a plurality of sensors provided in the laser device (16) and a plurality of sensors provided around the laser device (16) is output as the state quantity of the laser device (16), and the state quantity observation unit (6) observes the state quantity. [11] Warehouse management system (1) according to one of claims 4 to 10, wherein a plurality of information management devices (2) are connected to each other through a cloud, the plurality of devices (3) and at least one machine learning device (4) is connected to each of the information management devices (2) and the machine learning devices (4) connected to the information management devices (2) share learning result information through the information management device (2) or the cloud to which the information management devices (2) are connected. [12] Warehouse management system (1) according to one of claims 1 to 11, further comprising: an automatic warehouse connected to the information management device (2), wherein the information processor (30) automatically makes provisions for the designated component based on at least one of the following information: Stock management information including the appropriate stock quantity of the designated components and preventive maintenance information recommending the replacement of a characteristic deteriorating component among the designated components that has a property that gradually deteriorates during operation, or preventive maintenance information recommending the replacement of the designated component with a failure character, and Warehouse information for the automatic warehouse input by the information management device (2).

Images