Production failure analysis assistance device, production failure analysis assistance method, and storage medium

Abstract

A production failure analysis assistance device 6 comprises: a non-defective product data set collection unit 61 that collects a non-defective product data set GD relating to at least one of a first production device 2A and a non-defective product GG when the first production device 2A is producing the non-defective product GG; a non-defective product model construction unit 62 that constructs a non-defective product model GM which has learned the non-defective product data set GD as teacher data; and a deviation data identification unit 64 that inputs, to the non-defective product model GM, a defective product data set BD relating to at least one of a second production device 2B and a defective product BG when the second production device 2B is producing the defective product BG, and causes the non-defective product model GM to identify deviation data DD which deviates from the non-defective product data set GD in the defective product data set BD.

Description

Production defect analysis support device, production defect analysis support method, and storage medium

[0001] The present disclosure relates to a production defect analysis support device and the like.

[0002] Patent Document 1 discloses an intelligent process control device that uses three models with different characteristics, namely, a theoretical model, a neuro model, and a fuzzy control model, in combination to provide guidance to operators in charge of plant operation.

[0003] Japanese Patent Application Publication No. 8-335101

[0004] In Patent Document 1, by utilizing the advantages of each of three models with different characteristics, it is possible to provide high-quality guidance overall, but because three models are involved, it is difficult to identify the basis for the guidance.

[0005] The present disclosure has been made in consideration of the above circumstances, and aims to provide a production defect analysis support device and the like that can provide data useful for analysis when a production device is producing defective products.

[0006] In order to solve the above problem, a production defect analysis support device according to one aspect of the present disclosure includes a deviation data identification unit that inputs a defective product dataset relating to at least one of the production device and the defective product when the production device is producing defective products into a good product model that has learned a good product dataset when good products are being produced, and identifies deviation data in the defective product dataset that deviates from the good product dataset.

[0007] According to this aspect, by using a good product model that has been trained on a good product data set when good products are being produced, it is possible to identify deviation data that deviates between a defective product data set and a good product data set when a production device is producing defective products.

[0008] Another aspect of the present disclosure is a production defect analysis support method, which inputs a defective product dataset relating to at least one of a production device and the defective product when the production device is producing defective products into a good product model that has learned a good product dataset when good products are produced, and identifies deviation data in the defective product dataset that deviates from the good product dataset.

[0009] Yet another aspect of the present disclosure is a storage medium storing a production defect analysis support program that causes at least one processor to input a defective product dataset related to at least one of a production device and the defective product when the production device is producing a defective product to a good product model that has learned a good product dataset when good products are being produced, and identify deviation data in the defective product dataset that deviates from the good product dataset.

[0010] Any combination of the above components, or any conversion of these expressions into methods, devices, systems, recording media, computer programs, etc., are also encompassed within the present disclosure.

[0011] According to the present disclosure, it is possible to provide data that is useful for analyzing when a production device is producing defective products.

[0012] 1 is a schematic diagram showing a conveying device that conveys products in a production device; 2 is a schematic diagram showing a production defect analysis support device; 3 is a schematic diagram showing a change in the conveying speed of the products in FIG. 1 as an example of a change in production conditions; and 4 is a schematic diagram showing an example in which the functional blocks of the production defect analysis support device are realized in a distributed manner by a manufacturer and an equipment provider.

[0013] Hereinafter, with reference to the drawings, a detailed description of embodiments of the present disclosure (hereinafter also referred to as "embodiments") will be given. In the description and / or drawings, identical or equivalent components, members, processes, etc. will be designated by the same reference numerals, and redundant description will be omitted. The scale and shape of each part shown in the drawings are set for convenience to simplify the description and should not be interpreted as limiting unless otherwise specified. The embodiments are merely examples and do not limit the scope of the present disclosure in any way. Not all features and combinations thereof presented in the embodiments are necessarily essential to the present disclosure. For convenience, the embodiments are presented broken down into components for each function and / or functional group that realizes the features. However, one component in an embodiment may actually be realized by a combination of multiple separate components, or multiple components in an embodiment may actually be realized by a single integrated component. Furthermore, although multiple embodiments and variants may be disclosed in parallel, any components of each embodiment and / or each variant may be combined in any manner as long as they do not interfere with each other's functions.

[0014] The production defect analysis support device according to this embodiment can support the analysis of production defects in any production device. The production device is a device that produces any product. Examples of production devices include industrial machinery, industrial equipment, and the like used at industrial sites such as corporate factories, including mining machinery, chemical machinery, plastics machinery, hydraulic machinery, transport machinery, and steelmaking machinery, as handled by the Japan Society of Industrial Machinery Manufacturers. These include semiconductor manufacturing equipment, machine tools, printing machines, coating machines, molding machines (e.g., injection molding machines), and industrial robots. Note that production devices are not limited to devices that complete products. They may also be devices that produce parts or semi-finished products before the product is completed, or devices that perform part of the processing or treatment in the production process of such products, semi-finished products, or parts. In this embodiment, a gravure printing machine is primarily exemplified as a production device to be analyzed or adjusted.

[0015] FIG. 1 schematically shows a conveying device 2 that conveys a product 3 in a production device such as a gravure printing machine. Examples of the product 3 include linear objects such as strings and wires, and planar objects such as paper, cloth, film, foil, and rubber. In this embodiment, a roll-to-roll conveying device 2 is shown that conveys a planar substrate as the product 3 in a conveying direction (from the most upstream unwinding roll 25 toward the most downstream winding roll 26). The conveying device 2 constitutes part of a production device that performs any production process on the conveyed product 3, such as a coater or applicator (coating machine) that coats the product 3, a printing machine that prints on the product 3, or a stretching device that applies tension to the product 3 to stretch it.

[0016] The conveying device 2 conveys the workpiece 3 in the conveying direction by a number of conveying rollers 20. The conveying rollers 20 mainly comprise a drive roller 21 (a collective term for the five drive rollers 21A to 21E shown in the figure) which is rotationally driven by a motor 11 (a collective term for the five motors 11A to 11E shown in the figure (and four motors 11F to 11I described later)), a driven roller 22 (a collective term for the five driven rollers 22A to 22E shown in the figure) which pinches the workpiece 3 between itself and the drive roller 21 and rotates in conjunction with the drive roller 21, and a number of guide rollers 23 (two guide rollers in FIG. 1) which are arranged on the conveying path of the workpiece 3 and guide the workpiece 3. The conveying device 2 includes a dancer roller 20 (only the rollers are indicated by reference numerals), a dancer roller constituting the dancer 24 (a collective term for the four dancers 24A to 24D shown in the figure), an unwinding roll 25 (a collective term for the two unwinding rolls 25F and 25G shown in the figure) that is rotationally driven by the motor 11 (specifically, motors 11F and 11G) to unwind the workpiece 3 along the conveying direction, and a take-up roll 26 (a collective term for the two take-up rolls 26H and 26I shown in the figure) that is rotationally driven by the motor 11 (specifically, motors 11H and 11I) to wind up the workpiece 3. Fig. 1 shows a simple example of the configuration using a plurality of conveying rollers 20, and in an actual conveying device 2, any number and type of rollers can be provided in any arrangement on the conveying path between the unwinding roll 25 and the take-up roll 26.

[0017] Each pair of drive rollers 21A to 21E and driven rollers 22A to 22E transports the workpiece 3 sandwiched between them in the transport direction. Each drive roller 21A to 21E is driven to rotate by a corresponding motor 11A to 11E. Each driven roller 22A to 22E rotates in the opposite direction to the corresponding drive roller 21A to 21E at substantially the same speed as the corresponding drive roller 21A to 21E. For example, the drive roller 21A is driven to rotate counterclockwise by the motor 11A, and the driven roller 22A rotates clockwise in conjunction with the drive roller 21A at substantially the same speed.

[0018] As will be described later, in this embodiment, the roller measurement unit 4 (specifically, roller measurement units 4A to 4E) measures the rotational speed of each drive roller 21. However, because the rotational speed of each drive roller 21 and the rotational speed of each driven roller 22 are substantially the same, the roller measurement unit 4 may measure the rotational speed of each driven roller 22 instead of or in addition to the rotational speed of each drive roller 21. Here, the rotational speed of the drive roller 21 and / or the driven roller 22 is the tangential speed on the circumference (typically on the circumference) of the roller. Note that if the diameters of each pair of drive rollers 21 and driven rollers 22 are the same, not only the rotational speed but also the number of rotations per unit time will be the same for both rollers.

[0019] In the illustrated example, the pair of first drive roller 21A and first driven roller 22A is a pair of conveying rollers provided immediately after the most upstream unwinding roll 25, and is also referred to as in-feed rollers. In addition, in the illustrated example, the pair of fifth drive roller 21E and fifth driven roller 22E is a pair of conveying rollers provided immediately before the most downstream winding roll 26, and is also referred to as out-feed rollers.

[0020] In the illustrated example, the pairs of second to fourth drive rollers 21B to 21D and second to fourth driven rollers 22B to 22D are transport roller pairs provided between the infeed roller and the outfeed roller in the transport direction and are attached to processing sections that perform any processing on the workpiece 3 moving in the transport direction. Although detailed illustration is omitted, examples of the processing sections include a coating processing section that applies a coating to the workpiece 3, a printing processing section that prints on the workpiece 3, and a stretching processing section that applies tension to the workpiece 3 and stretches it. By providing each of these processing sections with a transport roller pair that is rotationally driven by a motor 11 (specifically, motors 11B to 11D), the transport speed when the workpiece 3 is processed can be adjusted with high precision.

[0021] A large number of guide rollers 23 are arranged on the transport path of the workpiece 3 to guide the workpiece 3. In other words, the guide rollers 23, together with the other transport rollers 20, form the transport path of the workpiece 3. The guide rollers 23 are free rollers that are not provided with a rotation drive unit such as a motor 11. As described above, the guide rollers 23 rotate by coming into contact with the workpiece 3 moving in the transport direction due to the pairs of drive rollers 21 and driven rollers 22 (and the unwind roll 25 and winding roll 26 described below). The workpiece 3 can then move smoothly in the transport direction by being guided by the rotating guide rollers 23.

[0022] The dancers 24 are provided to apply appropriate tension to each portion of the workpiece 3. In the illustrated example, the first dancer 24A is provided between the unwinding roll 25 and the infeed roller, the second dancer 24B is provided between the infeed roller and the processing section, the third dancer 24C is provided between the processing section and the outfeed roller, and the fourth dancer 24D is provided between the outfeed roller and the take-up roll 26. Each dancer 24 includes a dancer roller provided at a position deviated from the conveyance path of the workpiece 3. The dancer roller is biased or pressurized in a direction away from the conveyance path of the workpiece 3 by a thrust application unit (not shown), such as an air cylinder. When the thrust applied by the thrust application unit is approximately constant, the dancer roller applies approximately constant tension corresponding to the thrust to the workpiece 3. Note that the thrust applied by the thrust application unit may be variable, and the dancer roller may apply variable tension corresponding to the thrust to the workpiece 3.

[0023] An unwinding roll 25 provided at the start point of the workpiece 3 and / or the conveying device 2 unwinds the workpiece 3 in the conveying direction. In the illustrated example, an unwinding roll 25F in use that actually unwinds the workpiece 3 and an unused or used unwinding roll 25G that is not actually unwinding the workpiece 3 are schematically shown. The two unwinding rolls 25F and 25G are connected to each other by a turning mechanism such as a turret. When the remaining amount of workpiece 3 on the unwinding roll 25F in use (also referred to as the old shaft) decreases to a specified amount, it is replaced with the unused unwinding roll 25G (also referred to as the new shaft) by a turning operation. When in use, the unwinding rolls 25F and 25G are rotated by the corresponding motors 11F and 11G, respectively, to unwind the workpiece 3 in the conveying direction.

[0024] The winding roll 26, which is provided at the end point of the product 3 and / or the conveying device 2, winds up the product 3. In the illustrated example, a winding roll 26H in use that actually winds up the product 3 and an unused or used winding roll 26I that does not actually wind up the product 3, are schematically shown. The two winding rolls 26H and 26I are connected to each other by a turning mechanism such as a turret, and when the amount of product 3 wound on the winding roll 26H in use (also referred to as the old shaft) reaches a specified amount, it is replaced with the unused winding roll 26I (also referred to as the new shaft) by a turning operation. The winding rolls 26H and 26I are rotated by the corresponding motors 11H and 11I, respectively, during use, to wind up the product 3.

[0025] In the conveying device 2 described above, each motor 11 (11A to 11I) that rotates each drive roller 21 (21A to 21E), each unwinding roll 25 (25F and 25G), and each winding roll 26 (26H and 26I) is controlled so that the conveying speed of the workpiece 3 at the position of each roller 21, 25, 26 (i.e., the rotational speed of each roller 21, 25, 26) becomes a desired value.

[0026] The rotational speed of each motor 11 and the rotational speed of each roller 21, 25, 26 can be converted into each other based on mechanical specifications such as the diameter of each roller 21, 25, 26 and the reduction ratio of each reducer attached to each motor 11. Therefore, if the mechanical specifications accurately represent the actual roller diameter, reduction ratio, etc., the rotational speed of each roller 21, 25, 26 (i.e., the conveying speed of the workpiece 3) can be accurately calculated from the rotational speed of each motor 11. Specifically, each motor 11 (11A to 11I) is integrally provided with a motor measuring unit 12 (12A to 12I) such as a rotary encoder that measures the rotational speed. Then, based on the mechanical specifications such as the roller diameter and reduction ratio that are registered in advance, the measurement value of each motor measuring unit 12 (the rotational speed of each motor 11) is converted into the conveying speed of the workpiece 3 (the rotational speed of each roller 21, 25, 26).

[0027] However, since the actual roller diameter, reduction ratio, etc. may deviate from the machine specifications, the rotational speed of each motor 11 measured by each motor measuring unit 12 may not always be correctly converted into the conveying speed of each corresponding part of the workpiece 3 (the rotational speed of each roller 21, 25, 26). Therefore, in addition to each motor measuring unit 12 that measures the rotational speed of each motor 11, each roller measuring unit 4 that measures the rotational speed of each roller 21, 25, 26 may be provided.

[0028] The conveyance control device 5 controls the conveyance operation of the conveyance device 2 as described above. For example, the conveyance control device 5 adaptively controls the motors 11, dancers 24, and the like in each part of the conveyance device 2 based on various measurement data obtained through the motor measurement unit 12, the roller measurement unit 4, and any other sensors not shown, and various setting data (including preset default setting data, manual setting data set by a user, and auto setting data automatically set by a computer). Hereinafter, these measurement data and setting data will also be collectively referred to as input data.

[0029] The transport control device 5 may constitute part of an overall control device (not shown) that is responsible for overall control not limited to the transport operation of the transport device 2. Such an overall control device may include various control devices, including a temperature control device, a viscosity control device, a defect detector, and other monitoring devices, in addition to the transport control device 5. Measurement data and setting data in various control devices including the transport control device 5 are hereinafter collectively referred to as input data. The production defect analysis support device according to this embodiment, which will be described later, may be configured integrally with the transport control device 5 or the overall control device (or the various control devices that constitute them), or may be configured separately.

[0030] FIG. 2 schematically illustrates a production defect analysis support device 6 according to this embodiment. The production defect analysis support device 6 includes a non-defective product dataset collection unit 61, a non-defective product model construction unit 62, a defective product dataset collection unit 63, a deviation data identification unit 64, a presentation unit 65, and an automatic adjustment unit 66. Some of these functional blocks may be omitted as long as the production defect analysis support device 6 can achieve at least some of the actions and / or effects described below. These functional blocks may be realized by the cooperation of hardware resources, such as a central processing unit (CPU), memory, input devices, output devices, and peripheral devices connected to the computer, and software executed using these hardware resources. Regardless of the type or location of the computer, each of the above functional blocks may be realized by the hardware resources of a single computer or by combining hardware resources distributed across multiple computers.

[0031] As will be described later, the production defect analysis support device 6 is used in a learning phase in which a good product model GM is constructed, and an analysis phase in which a production defect is analyzed using the good product model GM. In order to schematically represent the difference between these phases, the good product data set collection unit 61, the good product model construction unit 62, and the flow of information and materials related thereto, which are mainly related to the learning phase, are indicated by dotted arrows, and the defective product data set collection unit 63, the deviation data identification unit 64, the presentation unit 65, the automatic adjustment unit 66, and the flow of information and materials related thereto, which are mainly related to the analysis phase, are indicated by solid arrows.

[0032] In the learning phase, the production defect analysis support device 6 constructs a good product model GM based on a good product data set GD when the first production device 2A is producing a good product GG, and in the analysis phase, inputs a defective product data set BD when the second production device 2B is producing a defective product BG into the good product model GM and outputs deviation data DD for analyzing production defects.

[0033] Here, the first production device 2A and the second production device 2B are devices of the same type whose production conditions P1, P2 and data sets GD, BD are comparable. For example, the first production device 2A and the second production device 2B may both be the conveying device 2 and / or a gravure printing machine as shown in FIG. 1 or any other comparable production device. The configurations of the first production device 2A and the second production device 2B do not need to be completely identical as long as they are substantially comparable, but it is preferable that they be substantially identical to increase the accuracy of the comparison. The first production device 2A and the second production device 2B are typically production devices of the same type and / or with the same model number and / or the same design. Specifically, the first production device 2A and the second production device 2B may be two different production devices (hereinafter, for convenience, also referred to as Machine No. 1 and Machine No. 2) with the same model number and / or the same design. Alternatively, the first production device 2A and the second production device 2B may be the same production device at different times or phases (i.e., a learning phase and an analysis phase).

[0034] The first production device 2A operates under first production conditions P1, and the second production device 2B operates under second production conditions P2. The production conditions P1 and P2 may be configured as a set of parameters specifying the basic operating modes of the production devices 2A and 2B when they produce the products GG and BG. When the production devices 2A and 2B are the conveying device 2 and / or the gravure printing press shown in FIG. 1 , the production conditions P1 and P2 include various production parameters, such as the conveying speed or production speed of the workpiece 3, the tension applied to the workpiece 3 by the dancer 24, and the ink viscosity and pressing pressure of the driven roller 22 in the printing processing unit (not shown) attached to the second to fourth drive rollers 21B to 21D. Note that if environmental parameters such as temperature, humidity, and brightness of the location where the production devices 2A and 2B are installed are controllable, these may also be included in the production conditions P1 and P2.

[0035] The first production device 2A is assumed to be operating under the first production conditions P1 and producing a non-defective product GG in the learning phase for constructing the non-defective product model GM. Here, the determination of the non-defective product GG, that is, the determination that the product produced by the first production device 2A satisfies a predetermined non-defective product standard, may be performed manually by an assessor or automatically by a non-defective / non-defective determination device (which may be a constructed non-defective product model GM) (not shown).

[0036] In this learning phase, the good-quality data set collection unit 61 collects a good-quality data set GD related to at least one of the first production device 2A and the good-quality product GG when the first production device 2A operating under the first production condition P1 is producing the good-quality product GG. The good-quality data set GD is, for example, a collection of various measurement data obtained by directly or indirectly measuring the first production device 2A and / or the good-quality product GG.

[0037] The non-defective product data set GD in the case where the first production device 2A is the conveying device 2 and / or the gravure printing machine shown in FIG. 1 includes various measurement data such as the rotational speed of each conveying roller 20 measured by the roller measurement unit 4, the rotational speed of each motor 11 measured by the motor measurement unit 12, the tension of each part of the workpiece 3 detected by a tension detector (not shown), the conveying speed of each part of the workpiece 3 measured by a camera (not shown) or the like, various monitoring data (e.g., print amount, printing pressure, drying temperature, ink viscosity) related to the printing process by a printing processing unit (not shown) serving as a processing unit attached to the second to fourth drive rollers 21B to 21D, the number and feature quantities of defects detected by a camera (not shown) or the like, and measurement values ​​of environmental parameters such as temperature, humidity, brightness, etc. of each part of the first production device 2A or the installation location of the first production device 2A.

[0038] The good product model construction unit 62 constructs a good product model GM that has been trained using as training data the diverse and enormous good product data set GD collected as described above by the good product data set collection unit 61. As will be described later, the good product model GM is a model that can output deviation data DD showing a deviation between at least the good product data set GD when the first production device 2A operating under the first production condition P1 is producing good products GG and the defective product data set BD when the second production device 2B operating under the second production condition P2 is producing defective products BG.

[0039] In the analysis phase following the learning phase, the second production device 2B is assumed to be operating under second production conditions P2 and producing defective products BG. The second production conditions P2 are preferably substantially the same as the first production conditions P1 under which the first production device 2A, which is the same type as the second production device 2B, produced non-defective products GG during the learning phase. In this case, the second production device 2B is operating under substantially the same conditions as the first production device 2A when producing non-defective products GG, and therefore should be able to produce non-defective products GG similar to those of the first production device 2A. However, due to individual differences between the first production device 2A and the second production device 2B, as well as differences in their installation errors and installation environments, the second production device 2B may be unable to produce non-defective products GG under the second production conditions P2, which are the same as the first production conditions P1, and may end up producing defective products BG.

[0040] In such a case, it is necessary to adjust the second production conditions P2 so that the second production device 2B can produce a non-defective product GG. The main purpose of the analysis phase described below is to provide deviation data DD that is useful in adjusting the second production conditions P2.

[0041] In the analysis phase, the defective product dataset collection unit 63 collects a defective product dataset BD related to at least one of the second production device 2B and the defective product BG when the second production device 2B, operating under second production conditions P2 that are substantially the same as the first production conditions P1, produces the defective product BG. The defective product dataset BD is, for example, a collection of various measurement data obtained by directly or indirectly measuring the second production device 2B and / or the defective product BG. Note that the determination of the defective product BG, i.e., the determination that the product produced by the second production device 2B does not satisfy the predetermined non-defective product standard, may be performed manually by an assessor or automatically by a pre-constructed non-defective product model GM, as described below.

[0042] When the second production device 2B is the conveying device 2 and / or the gravure printing machine shown in FIG. 1, the defective product dataset BD includes various measurement data such as the rotational speed of each conveying roller 20 measured by the roller measurement unit 4, the rotational speed of each motor 11 measured by the motor measurement unit 12, the tension of each part of the workpiece 3 detected by a tension detector (not shown), the conveying speed of each part of the workpiece 3 measured by a camera (not shown) or the like, various monitoring data (e.g., print volume, printing pressure, drying temperature, ink viscosity) related to the printing process by a printing processing unit (not shown) serving as a processing unit attached to the second to fourth drive rollers 21B to 21D, the number and feature quantities of defects detected by a camera (not shown) or the like, and measurement values ​​of environmental parameters such as temperature, humidity, brightness, etc. of each part of the second production device 2B or the installation location of the second production device 2B.

[0043] The deviation data identification unit 64 inputs the defective product data set BD collected by the defective product data set collection unit 63 when the second production device 2B operating under the second production conditions P2 is producing a defective product BG into the good product model GM, and identifies deviation data DD in the defective product data set BD that deviates from the good product data set GD.

[0044] As described above, the good product model GM has been trained using the good product data set GD as training data, and therefore can accurately identify deviation data DD that deviates from the good product data set GD in the defective product data set BD as input data.

[0045] For example, the good product model GM may compare each piece of measurement data included in the defective product data set BD with each piece of measurement data included in the good product data set GD one-to-one, and output data for which the deviation is greater than a predetermined threshold as deviation data DD. The threshold here is preferably set or optimized for each type of measurement data to be compared, and is preferably set smaller for measurement data that has a greater impact on the quality of the product (i.e., whether the product will be a good product GG or a defective product BG).

[0046] Furthermore, a more advanced good product model GM may, in addition to or instead of a one-to-one comparison of individual measurement data, compare the defective product data set BD and the good product data set GD as a set from multiple perspectives, and may output, as deviation data DD, one or a small number of measurement data in the defective product data set BD that are estimated to have a strong causal relationship with the defective product BG, after taking into consideration in a complex or comprehensive manner the correlation or dependency between the individual measurement data contained in each, and the influence on the quality of the product of measurement data groups arbitrarily defined within each data set.

[0047] The above-described good product model GM may be interpreted as defining a good product range within which a product is determined to be a good product GG and / or a defective product range within which a product is determined to be a defective product BG in a multidimensional data space constructed by compounding a large number of measurement data constituting each data set (e.g., good product data set GD) while taking into account the correlations or dependencies between the data sets. When each measurement data or measurement data group included in the defective product data set BD is outside the good product range and / or within the defective product range, it is identified by the good product model GM as a candidate for deviation data DD. Among these candidates for deviation data DD, for example, data with a relatively large distance or deviation from the good product range in the data space is ultimately identified as deviation data DD and output by the deviation data identification unit 64 and / or the good product model GM.

[0048] The good product model GM that defines the good product range and / or the defective product range in this manner may determine that the product produced by the second production device 2B is a defective product BG when the data set BD (or the individual measurement data or measurement data group that constitutes it) derived from the second production device 2B input through the deviation data identification unit 64 is significantly outside the good product range and / or the defective product range. Conversely, the good product model GM may determine that the product produced by the second production device 2B is a good product GG when the data set BD derived from the second production device 2B is significantly outside the good product range and / or the defective product range.

[0049] Such a good product model GM outputs a quality determination result DR of the product in response to input of a data set BD relating to at least one of the second production device 2B and its product, and may also output deviation data DD in parallel if the product is a defective product BG. Note that if the quality determination result DR is "good" (if the product of the second production device 2B is a good product GG), the good product model GM does not output deviation data DD (in other words, there is no deviation data DD that can be output).

[0050] As described above, the deviation data identifying unit 64 and / or the good product model GM identifies one or a small number of deviation data DD in which the cause or result of the defective product BG is apparent in the defective product data set BD. In other words, the deviation data identifying unit 64 and / or the good product model GM suggests or instructs one or a small number of deviation data DD that should be noted when improving the undesirable situation in which the second production device 2B is producing the defective product BG.

[0051] In order to efficiently output the deviation data DD that is useful for improving such production defects, it is preferable that the good product data set GD and the defective product data set BD that form the basis of the deviation data DD be collected by each data set collection unit 61, 63 during non-steady-state times when the production conditions P1, P2 of each production device 2A, 2B change.

[0052] 3 schematically shows the change in the conveying speed of the workpiece 3 in FIG. 1 as an example of changes in the production conditions P1 and P2. In this figure, the dotted line shows the change in the command value for the conveying speed from the conveying control device 5, and the solid line shows the change in the measured value of the conveying speed from the motor measuring unit 12 and the roller measuring unit 4. The measured value of the conveying speed shown by the solid line is an example of the measurement data included in each of the data sets GD and BD collected by the aforementioned each data set collecting units 61 and 63.

[0053] 3 specifically illustrates an example of "acceleration," in which the conveying speed of the workpiece 3 is increased, for example, when the conveying device 2 starts operating or is started up. When the measurement data (the speed measurement values ​​indicated by the solid lines) change significantly or conspicuously in this way, differences depending on the quality of the product (i.e., whether the product is a good product GG or a defective product BG) tend to appear in the measurement data. Therefore, by collecting measurement data (i.e., data sets GD and BD) when the production conditions P1 and P2 change as shown in FIG. 3, deviation data DD that reveals the difference between the good product GG and the defective product BG can be efficiently generated.

[0054] As described above, the good product data set GD is preferably collected by the good product data set collection unit 61 when the first production conditions P1 of the first production device 2A change. The defective product data set BD is preferably collected by the defective product data set collection unit 63 when the second production conditions P2 of the second production device 2B change in the same way as the first production conditions P1 of the first production device 2A when the good product data set GD was collected.

[0055] When the data set collectors 61, 63 collect each data set GD, BD, it is desirable to record metadata related to the changes or phases along with the data set GD, BD so that the deviation data identifying unit 64 and / or the good product model GM can accurately compare the collected data sets GD, BD for each corresponding change or phase as shown in Fig. 3. Examples of such metadata include whether the data is non-steady-state data collected during a non-steady state when there is a significant change in the production conditions P1, P2 or steady-state data collected during a steady state when there is no significant change; for the non-steady-state data, the type of change or phase (e.g., acceleration, deceleration, start, stop, tension setting change, temperature change); and statistical data (e.g., maximum value, minimum value, average value, median, standard deviation, variance, rate of change, frequency) that represent the type of change in each measurement data over the change phase.

[0056] The presentation unit 65 (deviation data presentation unit) presents the deviation data DD and / or the pass / fail judgment result DR output by the good product model GM using the defective product data set BD as input data to the adjuster OP of the second production device 2B. The presentation unit 65 may display the deviation data DD and / or the pass / fail judgment result DR on a display DP attached to a personal computer (not shown) or the like. Here, the presentation unit 65 may present the deviation data DD extracted from the defective product data set BD to the adjuster OP, or may emphasize the deviation data DD when presenting the defective product data set BD itself to the adjuster OP. The presentation unit 65 may also present to the adjuster OP an analysis or explanation of the presented deviation data DD, as well as recommended adjustment work for the second production device 2B and / or the second production conditions P2 to reduce the deviation data DD.

[0057] The presentation unit 65 may present the above-described various data to the automatic adjustment unit 66. The automatic adjustment unit 66 automatically adjusts the second production device 2B and / or the second production conditions P2 so as to reduce the deviation data DD.

[0058] According to the present embodiment as described above, when the second production device 2B produces defective products BG, the deviation data DD having a deviation between the defective product data set BD and the non-defective product data set GD can be specified by using the non-defective product model GM learned with the non-defective product data set GD as teacher data when the first production device 2A produces non-defective products GG.

[0059] As described above, the first production device 2A and the second production device 2B to be compared may be two different production devices of the same model number and / or the same design, etc., or may be the same production device with different time points or phases. Even in the case of the latter same production device, it is conceivable that non-defective products GG cannot be produced under the first production conditions P1 when the non-defective product data set GD was collected, due to differences in time zones, changes in the surrounding environment such as temperature, progress of deterioration, etc. According to the present embodiment, in such a case where defective products BG are produced under the first production conditions P1 as they are, the old first production conditions P1 can be updated to new appropriate second production conditions P2 based on the deviation data DD output by the non-defective product model GM.

[0060] The production defect analysis support device 6 schematically shown in FIG. 2 may have substantially all its functional blocks provided at the production site (factory, etc.) where non-defective products GG and / or defective products BG are produced, but at least some of the functional blocks may be provided by the device provider who provides the production device, etc. to the producer. FIG. 4 schematically shows an example in which the functional blocks of the production defect analysis support device 6 are realized dispersedly by the producer and the device provider. The producer is the person who produces non-defective products GG and / or defective products BG using the first production device 2A and / or the second production device 2B shown in FIG. 2, and the device provider is the analyst who constructs the non-defective product model GM based on the non-defective products GG produced by the producer and outputs the deviation data DD for analyzing the defective products BG produced by the producer using the non-defective product model GM.

[0061] As shown schematically in Figure 4, the production defect analysis support device 6 is divided into two temporally different phases: "when constructing a good product model" and "when producing defective products." When constructing a good product model, a good product data set collection unit 61 implemented on the producer side collects a good product data set GD for good products GG produced by the producer or the first production device 2A. This good product data set GD is provided by the producer to the equipment provider in order to construct a good product model GM. A good product model construction unit 62 implemented on the equipment provider side constructs the good product model GM based on the good product data set GD provided by the producer.

[0062] When defective products are produced, a defective product dataset collection unit 63 implemented on the producer side collects a defective product dataset BD for defective products BG produced by the producer or the second production device 2B. This defective product dataset BD is provided by the producer to the equipment provider in order to output deviation data DD. A deviation data identification unit 64 implemented on the equipment provider side inputs the defective product dataset BD provided by the producer into the good product model GM constructed in the previous phase, outputting deviation data DD. This deviation data DD is provided by the equipment provider to the producer, and is presented to the producer (adjuster OP, etc.) via a display DP or the like by a presentation unit 65 implemented on the producer side.

[0063] The present disclosure has been described above based on the embodiments. Various modifications are possible to the combinations of the components and processes in the exemplary embodiments, and it will be obvious to those skilled in the art that such modifications are included within the scope of the present disclosure.

[0064] The configuration, operation, and function of each device and method described in the embodiments can be realized by hardware resources, software resources, or a combination of hardware and software resources. Examples of hardware resources include processors, ROM, RAM, and various integrated circuits. Examples of software resources include operating systems, applications, and other programs.

[0065] The present disclosure relates to a production defect analysis support device and the like.

[0066] 2 Conveying device, 2A First production device, 2B Second production device, 3 Produced item, 5 Conveying control device, 6 Production defect analysis support device, 61 Good product data set collection unit, 62 Good product model construction unit, 63 Defective product data set collection unit, 64 Deviation data identification unit, 65 Presentation unit, 66 Automatic adjustment unit, P1 First production condition, P2 Second production condition.

Claims

1. A production defect analysis support device that includes a deviation data identification unit that inputs a defective product dataset related to at least one of the production equipment and the defective products when the production equipment is producing defective products into a good product model that has learned from a good product dataset when good products are being produced, and identifies deviation data in the defective product dataset that deviates from the good product dataset.

2. The production defect analysis support device according to claim 1, further comprising: a good product model construction unit that constructs the good product model trained using the good product data set relating to at least one of a first production device and the good product when the first production device is producing a good product, as training data; and wherein the deviation data identification unit inputs the defective product data set relating to at least one of a second production device and the defective product when the second production device is producing a defective product into the good product model, and identifies deviation data in the defective product data set that deviates from the good product data set.

3. The production defect analysis support device according to claim 2, wherein the defective product data set is collected when the second production device is operating under production conditions substantially the same as the production conditions of the first production device when the good product data set was collected.

4. The production defect analysis support device according to claim 3, wherein the good product data set is collected when the production conditions of the first production device change, and the defective product data set is collected when the production conditions of the second production device change in the same way as the production conditions of the first production device when the good product data set was collected.

5. The production defect analysis support device according to claim 3, wherein the first production device and the second production device are different production devices of the same type.

6. The production defect analysis support device according to claim 3, wherein the first production device and the second production device are the same production device.

7. A production defect analysis support device according to any one of claims 1 to 6, wherein the non-defective product model determines whether a product is good or bad in response to input of a data set relating to at least one of a production device and the product, and outputs the deviation data if the product is defective.

8. The production defect analysis support device according to any one of claims 1 to 6, further comprising a deviation data presentation unit that presents the deviation data to an adjuster of the production device.

9. A production defect analysis support device according to any one of claims 1 to 6, further comprising an automatic adjustment unit that automatically adjusts the production device so as to reduce the deviation data.

10. A production defect analysis support method that inputs a defective product dataset related to at least one of the production equipment and the defective products when the production equipment is producing defective products into a good product model that has been trained with a good product dataset when good products are being produced, and identifies deviation data in the defective product dataset that deviates from the good product dataset.

11. A storage medium storing a production defect analysis support program that causes at least one processor to input a defective product dataset relating to at least one of the production equipment and the defective products when the production equipment is producing defective products into a good product model that has learned from a good product dataset when good products are being produced, and identify deviation data in the defective product dataset that deviates from the good product dataset.

Images