JP7053366B2 - Inspection equipment and inspection method - Google Patents

Description

The present invention relates to an inspection object formed by melting a material with heat, an inspection object manufactured by polishing the surface, or an inspection object manufactured by cutting, and the surface shape of the curved surface of the inspection object is determined. Regarding the inspection device and inspection method to be inspected.

Inspection of products with complicated curved surface shapes such as impellers of large pumps and their parts (hereinafter referred to as parts) is one of the tasks that is difficult to quantify, requires skill, and requires a large number of man-hours. In particular, such parts are often manufactured by casting, etc., but parts manufactured by casting, etc. may be polished with a grinder or the like in order to remove surface roughness and waviness (unevenness). It is common. Further, even in the case of manufacturing by machining or the like, polishing may be performed in order to erase so-called cutting marks after machining. Alternatively, "warping" or "waviness" of parts or the like may occur due to thermal deformation or the like during processing. Further, in recent years, there is also a technique of melting and laminating metal wires and the like to mold parts (fused metal laminating method, etc.), but in this case as well, the "step" generated by the laminating is polished and removed after molding. Sometimes. As described above, the shape inspection after molding or polishing by melting is difficult to quantify and requires skill in judgment as described later.

Conventionally, an impeller inspection method has been developed. For example, in Patent Document 1, the step of imaging the front surface of the impeller from the direction of the axis of rotation of the impeller placed at the installation location and the front of the impeller from the direction of the axis of rotation of the impeller imaged in the above imaging step. The process of binarizing the captured image related to the above and obtaining the binarized image, and the tip-corresponding bright part of all the blade-compatible bright parts provided around the impeller-compatible bright part based on the binarized image. The step of detecting the position, the step of calculating the positional relationship between each tip-corresponding bright portion and the predetermined reference portion in the impeller-corresponding bright portion for all the detected tip-corresponding bright portions, and the above calculation. A method for inspecting the blade shape of an impeller is disclosed, which comprises a step of determining the quality of the blade shape of the impeller by comparing the obtained positional relationship with a predetermined specified value.

Japanese Unexamined Patent Publication No. 2008-51664

However, unintended irregularities occur on the curved surface during casting, and the surface shape during casting (so-called "casting surface") is rough as a part of a fluid machine such as a pump, so the surface is polished and smoothed. Because it is necessary to do so, sometimes the surface may be shaved too much and unevenness may occur. Alternatively, even in the case of machining, unevenness is generated by polishing the surface in order to erase the cutting marks generated after machining, and "waviness" or "warp" is caused by thermal deformation during machining. May cause unevenness. It is difficult to set an allowable value for such unevenness.

Specifically, as an inspection method for such products, the dimensions of the product are precisely measured by various dimensional measuring instruments or so-called three-dimensional measuring instruments, and it is inspected whether the dimensions are within the permissible value. There is a way to do it.

FIG. 1 is a first schematic diagram of a design shape (reference shape), an upper limit and a lower limit of an allowable value, and a product shape. FIG. 1 shows a one-dot chain line L1 representing a design shape (reference shape), a broken line L2 representing the lower limit of the allowable value of dimensions, a broken line L3 representing the upper limit of the allowable value of dimensions, and a solid line L4 representing the product shape. There is. For example, as shown in FIG. 1, when the upper limit and the lower limit of the allowable value of dimensions are set for the design shape (reference shape) within the range that does not affect the performance of the product, the product shape has the upper limit and the lower limit of the allowable value. If it is within the range (that is, within the allowable range), it can be regarded as a pass.

On the other hand, in the case of parts of a fluid machine such as an impeller of a pump, this alone may not be enough to consider the product as acceptable. FIG. 2 is a second schematic diagram of a design shape (reference shape), an upper limit and a lower limit of an allowable value, and a product shape. FIG. 2 shows a one-dot chain line L11 representing the design shape (reference shape), a broken line L12 representing the lower limit of the allowable value of the dimension, a broken line L13 representing the upper limit of the allowable value of the dimension, and a solid line L14 representing the product shape. There is. For example, when there is "waviness" on the surface of the product as in the product shape of FIG. 2, the product shape (measured value at each point) is between the upper limit and the lower limit of the allowable value (that is, within the allowable range). However, such "waviness" has a great influence on the performance of the inspection device, and therefore cannot be passed. As the parts of such a fluid machine, most of the parts in the flow path are targeted, and in particular, the following parts can be mentioned.
○ Impeller ○ Diffuser (including pressure recovery flow path, swirl casing, guide blades, etc.)
○ Suction pipe / Discharge pipe ○ Underwater casing of bearing / shaft seal, especially its wetted surface ○ Suction bell ○ and parts that compose these

When trying to determine such a defect by numerical measurement, it is necessary to quantify such "waviness". There are many possible methods, but in any case, complicated processing such as measuring, recording, and statistically processing a large number of points is required, and it is difficult to set a reference value.

FIG. 3 is a schematic diagram showing an inspection method according to a comparative example. In FIG. 3, the alternate long and short dash line L21 representing the design shape (reference shape), the broken line L22 representing the lower limit of the allowable value of the dimension, the broken line L23 representing the upper limit of the allowable value of the dimension, the solid line L24 representing the product shape, and sampling data are shown. A stepped solid line L25 is shown. On the other hand, for example, as shown in FIG. 3, a method is conceivable in which shape data is sampled at regular intervals, the difference between adjacent measured values (sampling data) is calculated, and the quality is determined based on the difference. In this case, there is a method such as making a defect when this difference is larger than a certain value or smaller than a certain value with respect to the difference of the design shape. However, in this case, it is necessary to measure a large number of points, and since the period (wavelength) and magnitude of the waviness are different each time, it is difficult to determine the threshold value.

Therefore, in practice, such "waviness" is often performed not by measuring the dimensions (numerical values) but by a visual inspection by an inspector or a so-called "sensory inspection" by the tactile sensation of a hand or the like. In the case of a sensory test, the quality of a product can be judged by the sense of an inspector without setting a standard value each time. In addition, it is possible to reduce the risk of judging a defective product as a good product by judging whether it is good or bad based on the sensibility of a human being (inspector), rather than setting a reference value or the like.

However, in the sensory test, (1) the judgment varies depending on the measurer (the judgment of pass / fail differs depending on the measurer), (2) the measurement can only be performed by humans, so it cannot be automated, and (3) the numerical value. It is difficult to convert and record, etc. (4) It is difficult to determine the "allowable range" because it is difficult to quantify, and there is a tendency to obtain "excessive quality" (it takes more correction man-hours than necessary). There is a problem such as that. In particular, in recent years, it has become difficult to secure skilled inspectors due to the aging of inspectors and the difficulty of passing on technology, and the development of inspection equipment and inspection methods that can objectively and automatically inspect these. Is an urgent task.

Also, among the parts of fluid machinery, so-called custom products such as large pumps, for example, the shape of the impeller is optimized according to the customer's specifications, so the standard shape differs for each product and is inspected each time. It is necessary to create the data used for. This also hinders automation.

The present invention has been made in view of the above problems, and provides an inspection device and an inspection method capable of objectively and automatically inspecting the surface shape of a curved surface of an object, which is difficult to determine only by dimensional tolerances. The purpose is.

The inspection device according to the first aspect of the present invention relates to an inspection object formed by melting a material with heat, an inspection object manufactured by polishing the surface, or an inspection object manufactured by cutting. An inspection device for inspecting the surface shape of a curved surface of an inspection object, the projection device for projecting a specific pattern on the inspection object, an imaging device for imaging the inspection object on which the pattern is projected, and the inspection. A learning object that is the same type as the object and whose surface shape of the curved surface is known to be good or bad. The learning object is photographed with the same specific pattern as the specific pattern projected on the inspection object. It has artificial intelligence that has learned the set of the image and the sensory test result of the surface shape of the curved surface of the object for learning as teacher data, and has already learned the image captured by the image pickup device. It is provided with a determination circuit that is applied to the artificial intelligence and determines whether the surface shape of the curved surface of the inspection object is good or bad.

According to this configuration, in order to judge the quality of the surface shape of the curved surface of the inspection object by applying it to the learned artificial intelligence, the surface shape of the curved surface of the inspection object, which is difficult to judge only by the dimensional tolerance, is objectively and. It can be inspected automatically. That is, defects such as "waviness", which were difficult to determine by the conventional numerical method, can be mechanically determined. Further, the inspection can be performed unattended, the determination result can be recorded numerically, and the determination can be eliminated from the inspector's variation as in the conventional case.

The inspection device according to the second aspect of the present invention is the inspection device according to the first aspect, and the specific pattern is a striped pattern or a lattice pattern.

According to this configuration, if there are waviness, steps, or cracks in the surface shape, waviness, corners, or part of the pattern disappears in a part of the striped pattern or lattice pattern. By learning the changes by artificial intelligence, it is possible to determine defects such as waviness, steps, and cracks in the surface shape.

The inspection device according to the third aspect of the present invention is the inspection device according to the first or second aspect, and there are two projection devices, and each projection device is from two directions in which the projection directions are substantially orthogonal to each other. By projecting the striped pattern, the lattice pattern is projected as the specific pattern.

According to this configuration, by learning the change of the lattice pattern by artificial intelligence, it is possible to determine defects such as waviness, step, and crack in the surface shape.

The inspection device according to the fourth aspect of the present invention is the inspection device according to any one of the first to third aspects, and the artificial intelligence is an inspection object known to have a good surface shape of a curved surface. The pair of the image of the inspection target and the identification information for identifying the non-defective product, and the identification of the inspection target whose surface shape of the curved surface is known to be defective, to identify the image of the inspection object and the cause of the defect. It is learned to output the certainty of a good product and the certainty of each factor of a defect by using a set with information as teacher data, and the determination circuit uses the certainty of a good product and the certainty of a factor of a defect to output a good product. Or, output identification information that identifies the cause of the defect.

According to this configuration, the inspector can grasp not only the quality of the inspection target but also the cause of the defect in the case of a defect.

The inspection device according to the fifth aspect of the present invention is the inspection device according to any one of the first to third aspects, and the determination circuit is provided with artificial intelligence for each cause of failure. For each of the artificial intelligences, a set of an image of the learning object and identification information for identifying a good product for a learning object which is the same type as the inspection object and is known to have a good curved surface shape, and the above-mentioned For a learning object that is the same type as the inspection object and is known to have a defect factor targeted by the artificial intelligence in the surface shape of the curved surface, the image of the learning object and the defect factor of the target are identified. It is learned to output the certainty of a good product and the certainty of the cause of the defect to be the target by using the set with the identification information as the teacher data, and each of the artificial intelligences uses the captured image of the inspection object. , The certainty of good products and the certainty of different factors of defects for the inspection object are output, and the determination circuit outputs the certainty of each of the good products and the certainty of the factors of the defects output from each of the artificial intelligences. Each of them is used to output identification information for identifying the cause of good or bad for the inspection object.

According to this configuration, the inspector can grasp not only the quality of the inspection target but also the cause of the defect in the case of a defect.

The inspection device according to the sixth aspect of the present invention is the inspection device according to any one of the first to fifth aspects, and the projection device can switch between projection and non-projection. The image pickup apparatus captures the inspection object in a non-projected state to acquire a first image, and images the inspection object in a state where the pattern is projected to acquire a second image. The determination circuit applies the difference image between the first image and the second image to the trained artificial intelligence learned by using the teacher data of the difference image similarly created. The quality of the surface shape of the curved surface of the inspection object is determined.

According to this configuration, only the projected pattern is emphasized in the difference image, so that the determination accuracy by learning can be improved and the accuracy of the determination result can be improved. In particular, in the case of a part or the like formed by laminating molten metal, it is possible to reduce the influence of the striped pattern or the like generated by the laminating.

The inspection device according to the seventh aspect of the present invention is the inspection device according to any one of the first to sixth aspects, and is a part manufactured by polishing the surface.

According to this configuration, in order to judge the quality of the surface shape of the curved surface of the part manufactured by polishing the surface by applying it to the learned artificial intelligence, the surface shape of the curved surface of the part that is difficult to judge only by the dimensional tolerance is determined. It can be inspected objectively and automatically.

The inspection device according to the eighth aspect of the present invention is the inspection device according to any one of the first to seventh aspects, and the inspection object is a part of a fluid machine.

According to this configuration, in order to judge the quality of the surface shape of the curved surface of the fluid machine part by applying it to the learned artificial intelligence, the surface shape of the curved surface of the fluid machine part, which is difficult to judge only by the dimensional tolerance, is objectively determined. It can be inspected objectively and automatically.

The inspection device according to the ninth aspect of the present invention is the inspection device according to any one of the first to eighth aspects, and the inspection object is a part manufactured by a fused metal lamination method or polishing.

According to this configuration, in order to judge whether the surface shape of the curved surface of a part manufactured by the molten metal lamination method or polishing is good or bad by applying it to the learned artificial intelligence, it is difficult to judge the surface of the curved surface of the part only by the dimensional tolerance. The shape can be inspected objectively and automatically.

In the inspection method according to the tenth aspect of the present invention, the surface shape of the curved surface of the inspection object is inspected for the inspection object formed by melting the material with heat or the inspection object manufactured by polishing the surface. In the inspection method, the procedure of projecting a specific pattern on the inspection object, the procedure of imaging the inspection object on which the pattern is projected, and the captured image are applied to the learned artificial intelligence. The artificial intelligence has a procedure for determining the quality of the surface shape of the curved surface, and the artificial intelligence is applied to the inspection target for a learning object which is the same type as the inspection object and whose surface shape of the curved surface is known to be good or bad. Teacher data is a set of the image of the learning object taken with the same specific pattern as the projected specific pattern and the sensory test result of the quality of the surface shape of the curved surface of the learning object. It is an inspection method that was learned as.

According to this configuration, in order to judge the quality of the surface shape of the curved surface of the inspection object by applying it to the learned artificial intelligence, the surface shape of the curved surface of the inspection object, which is difficult to judge only by the dimensional tolerance, is objectively and. It can be inspected automatically. That is, defects such as "waviness", which were difficult to determine by the conventional numerical method, can be mechanically determined. Further, the inspection can be performed unattended, the determination result can be recorded numerically, and the determination can be eliminated from the inspector's variation as in the conventional case.

According to one aspect of the present invention, in order to determine the quality of the surface shape of the curved surface of the inspection object by applying it to the learned artificial intelligence, the surface shape of the curved surface of the inspection object, which is difficult to determine only by the dimensional tolerance, is determined. It can be inspected objectively and automatically. That is, defects such as "waviness", which were difficult to determine by the conventional numerical method, can be mechanically determined. Further, the inspection can be performed unattended, the determination result can be recorded numerically, and the determination can be eliminated from the inspector's variation as in the conventional case.

It is the first schematic diagram of the design shape (reference shape), the upper limit and the lower limit of the permissible value, and the product shape. It is the second schematic diagram of the design shape (reference shape), the upper limit and the lower limit of the permissible value, and the product shape. It is a schematic diagram which shows the inspection method which concerns on the comparative example. It is a schematic block diagram which shows the structure of the inspection apparatus which concerns on this embodiment. It is a schematic diagram when a lattice pattern is projected on a good impeller. It is a schematic diagram when the lattice pattern is projected on the impeller of a defective product. It is a schematic diagram explaining the structure of the determination circuit which concerns on this embodiment. It is a flowchart which shows an example of the determination process which concerns on this embodiment. It is a schematic diagram explaining the structure of the determination circuit which concerns on the modification.

Hereinafter, each embodiment will be described with reference to the drawings. However, more detailed explanation than necessary may be omitted. For example, detailed explanations of already well-known matters and duplicate explanations for substantially the same configuration may be omitted. This is to avoid unnecessary redundancy of the following description and to facilitate the understanding of those skilled in the art.

The inspection device and inspection method according to the present embodiment are inspected for an inspection object formed by melting a material with heat, an inspection object manufactured by polishing the surface, or an inspection object manufactured by cutting. It inspects the surface shape of the curved surface of an object. In particular, it is suitable for surface inspection of parts of fluid machines such as large pumps and compressors, and is suitable for inspection of products having a complicated three-dimensional shape such as impellers among parts of fluid machines. Here, the parts of the fluid machine include, for example, an impeller, a diffuser (including a pressure recovery flow path, a swirl casing / guide vane, etc.), a suction pipe / discharge pipe, an underwater casing for a bearing / shaft seal, and particularly a wetted surface thereof. The suction bell and the parts that make up these. Further, examples of the material formed by melting the material by heat include casting, powder metallurgy, and laminating of molten metal. In this way, the inspection object formed by melting the material with heat needs to be inspected because the surface shape of the curved surface may not be as designed because the material shrinks during the cooling process. .. Further, when the surface is polished and manufactured, unevenness may occur due to excessive scraping of the surface. In this case as well, inspection is required.

FIG. 4 is a schematic configuration diagram showing the configuration of the inspection device according to the present embodiment. The inspection device includes a projection device that projects a specific pattern onto an inspection object (here, an impeller 2 as an example). The pattern is preferably a striped pattern or a checkered pattern. As shown in FIG. 4, the inspection device 1 according to the present embodiment includes a projection device 11 and a projection device 12, and as an example, from two projection devices whose projection directions are substantially orthogonal to each other, that is, the projection device 11 and the projection device 12. By projecting a striped pattern, a grid pattern may be projected on the impeller 2 which is an inspection object.

As a result, a plurality of curves W1 to W10 appear on the surface of the product due to the curved surface thereof. At this time, if there is waviness or the like as described above, the projected pattern is also distorted due to the waviness.

Here, the inspection device 1 according to the present embodiment applies the learned artificial intelligence to the image pickup device 13 that captures the impeller 2 that is the inspection target on which the pattern is projected and the image captured by the image pickup device 13. A determination circuit 14 for determining the quality of the surface shape of the curved surface is provided.

FIG. 5A is a schematic view when a grid pattern is projected on a non-defective impeller. FIG. 5A is a schematic view when a grid pattern is projected on a defective impeller. For example, in the case of a non-defective product, a smooth curve appears on the impeller as shown in FIG. 5A. On the other hand, if there is "waviness" as described above, the projected pattern also undulates as shown in FIG. 5B. Further, if there is a "step" that causes an obtuse angle on the surface, a "corner" is generated in the projected pattern. Further, if "cracks" or the like are generated on the surface, a part of the projected pattern disappears or corners are generated.

For example, when this "wavy" is set as the first defect factor and the first defect factor is present, the classification of the impeller which is the first defect and has the first defect factor is set to the first defect class. There is. For example, when this "step" is set as the second defect factor and the second defect factor is present, the classification of the impeller which is the second defect and has the second defect factor is set to the second defect class. There is. For example, when this "crack" is set as the third defect factor and the third defect factor is present, the classification of the impeller which is the third defect and has the third defect factor is set to the third defect class. There is.

FIG. 6 is a schematic diagram illustrating the configuration of the determination circuit according to the present embodiment. A person (for example, a skilled inspector) determines in advance whether a known impeller is good or bad, for example, a good product, a first defect, a second defect, or a third defect. Then, the set of the known impeller image acquired by the image pickup device 13 in the state where the same specific pattern as the specific pattern projected on the inspection object is projected and the quality thereof becomes the teacher data, and the determination is made. The artificial intelligence in the circuit 14 is learned using this teacher data.
In this way, the determination circuit 14 using artificial intelligence is preliminarily trained using a predetermined number of pairs of image data of good and defective products and good or bad determined by a person. back. At that time, in the case of a defective product, by dividing the defective product into several defective factors and learning the imaging data, it is possible to identify the defective factor at the same time as determining the quality.

In short, the artificial intelligence 32 is a set of an image of the inspection object and identification information for identifying a good product for a learning object known to be of the same type as the inspection object and having a good surface shape of the curved surface, and a curved surface. For a learning object whose surface shape is known to be defective, an image of the learning object taken with the same specific pattern as the specific pattern projected on the inspection object and the defect The pair with the identification information (here, for example, the first defect, the second defect, or the third defect) that identifies the factor is learned as teacher data. The artificial intelligence 32 outputs the certainty of a good product and the certainty of each defect factor for the inspection object by using the captured image of the inspection object. Then, the determination circuit 14 uses, for example, the certainty of the good product and the certainty of the cause of the defect to identify the good product or the cause of the defect of the inspection object (here, for example, the first defect). , Second defect, or third defect) is output. With this configuration, the inspector can identify not only whether the inspection target is a good product or a defective product, but also, in the case of a defective product, the cause of the defect.

Here, the artificial intelligence used in the determination circuit will be described. The artificial intelligence used in this embodiment is similar to so-called "image recognition", and a neural network, particularly one using a deep neural network (hereinafter, also referred to as DNN) is suitable, and therefore, a deep neural network is used here. Will be described as an example.

Generally, in DNN, a necessary number of images of "good products" and a plurality of "defective products" are prepared in advance, and these are learned by a method called "deep learning". In the present embodiment, the image of the product is acquired by the method described so far, and the same product is subjected to the conventional sensory test to determine whether the product is good or bad, and the good product and a plurality of defective products. Prepare the required number of images (for example, about several tens to several hundreds) and let the DNN learn. As a result, DNN generally does not appear in "good products", but shows a strong reaction to the "characteristics" that appear in each "defective product", and the "score" of the corresponding "class" is highly evaluated. Become. In this way, the determination circuit 14 takes a picture in a state where the same specific pattern as the specific pattern projected on the inspection object is projected on the learning object which is the same type as the inspection object and whose surface shape of the curved surface is known. It has artificial intelligence that has learned a set of the image of the learning object and the sensory test result of the quality of the surface shape of the curved surface of the learning object as teacher data.

What is important here is that there is no reference value for these "features", and among the "neurons" that are logical elements in the DNN, those corresponding to those features respond more strongly to "learning". Is to return. In general, in DNN used for image processing, by learning "defects" in various states, a strong reaction can be obtained regardless of the position and size in which those characteristics appear.

In addition, when learning, it is important to create learning data using products that have many variations (types) for the products to be used. For example, in the case of a pump impeller, there are large and small ones, different Ns values (axial flow pump and mixed flow pump, etc.), number of impellers, two-dimensional impeller and three-dimensional impeller, and the like. As a result, the DNN can learn that the change in the image due to these differences is "not related" to the quality of the product, so that the DNN can judge the quality of the product even if there is no standard shape. ..

Therefore, in the case of the present embodiment, unlike the case of performing the conventional shape measurement, the threshold value as the dimensional value is unnecessary, and the quality of the product can be judged even if the design (reference) shape is not known. Is possible. Therefore, once the learning is completed, it is easy to automate the inspection and the like.

As the artificial intelligence used in the determination circuit, a neural network (particularly a deep neural network) is used as an example in this embodiment, but in some cases, another artificial intelligence algorithm is used (for example, KNN method, decision tree). The method, MT method, etc.) may be used.

The judgment of good or bad is made by the following method. The output of the artificial intelligence 32 in FIG. 6 is a matrix of a plurality of output items (classes) including "non-defective products" and their certainty (score). Specifically, the output items (classes) are a good product class and a defective class corresponding to one or more defective factors. In addition, here, it is described as having a defect class of a plurality of factors (2 or more).

Specifically, in the present embodiment, as shown in FIG. 6, as an example, an captured image obtained by capturing an image of an inspection object is input to the artificial intelligence 32 of the determination circuit 14 as an input image 31. The output matrix 33 output from the artificial intelligence 32 is a good product class and its certainty (score), a first bad class and its certainty (score), a second bad class and its certainty (score), and a third. Includes bad class and its certainty (score). Here, the certainty (score) indicates the certainty to the corresponding class, and the larger the score value, the higher the certainty. This output matrix 33 is input to the final determination unit 34 of the determination circuit 14.

After that, the final determination unit 34 makes a determination according to FIG. 7. FIG. 7 is a flowchart showing an example of the determination process according to the present embodiment.
(Step S101) First, the final determination unit 34 confirms the score of the “good product” class. For example, assuming that the reference score of a non-defective product is 0.8, the final determination unit 34 determines whether or not the score of the “non-defective product” class is 0.8 or more.

(Step S102) When the score of the non-defective product class is 0.8 or more in step S101, the final determination unit 34 determines that the inspection target is a non-defective product.

(Step S103) When the score of the non-defective product class is less than 0.8 in step S101, the final determination unit 34 then sequentially confirms the scores of a plurality of defective items (defective class). Here, among the scores of each defective item, the one with the highest score is likely to be a defective factor. Therefore, the final determination unit 34 determines the class having the highest score among the first to third defective classes, and the identification information for identifying the defect factor corresponding to the class having the highest score (for example, the first defect). May be output. Here, as an example, first, the final determination unit 34 determines whether or not the score of the first defective class among the first to third defective classes is the maximum.

(Step S104) When the score of the first defect class among the first to third defect classes is the maximum in step S103, the final determination unit 34 determines that the first defect.

(Step S105) When the score of the first defective class among the first to third defective classes is not the maximum in step S103, the final determination unit 34 determines the score of the second defective class among the first to third defective classes. Is the maximum.

(Step S106) When the score of the second defect class among the first to third defect classes is the maximum in step S105, the final determination unit 34 determines that the second defect.

(Step S107) When the score of the second defect class among the first to third defect classes is the maximum in step S105, the final determination unit 34 determines that the third defect.

Then, the final determination unit 34 determines identification information (specifically, a first defect, a second defect, or a third defect) that identifies the cause of the defect as a result of determining the quality of the inspection object. Defective) is output. As a result, it is possible to grasp the quality of the inspection target, and in the case of a defect, it is possible to grasp the cause of the defect.

As described above, in the present embodiment, the artificial intelligence 32 outputs the certainty of the good product and the certainty of each factor of the defect. The determination circuit 14 determines whether the surface shape of the curved surface is good or bad by using the certainty of the good product and the certainty of the cause of the defect. As a result, the inspector can grasp not only the quality of the inspection object but also the cause of the defect in the case of a defect.

When there are two classes of defect factors, a first defect class and a second defect class, the determination circuit 14 may output either a non-defective product, a first defect, or a second defect. When the class of the defect factor is four or more, the determination circuit 14 may output either a good product or each defect.

The image pickup apparatus 13 takes an image of an inspection object (here, an impeller as an example) on which the pattern is projected. Since the image pickup apparatus 13 finally requires digital data, a so-called digital camera or a similar one is desirable.

The projection device 11 and the projection device 12 may be capable of switching between projection and non-projection of the pattern. In that case, at the time of measurement, the image pickup apparatus 13 first takes an image of the inspection object in a state where the pattern is not projected and acquires a first image, and then takes an image of the inspection object in a state where the pattern is projected. And get the second image. Then, the image pickup apparatus 13 may obtain the difference between the first image and the second image and output the difference image as image data to the determination circuit 14. The determination circuit 14 applies the trained artificial intelligence learned by using the teacher data of the difference image created in the same manner to the difference image between the first image and the second image, and applies the surface shape of the curved surface. Judge the quality of. As a result, only the projected pattern is emphasized in the difference image, so that the determination accuracy by learning can be improved and the accuracy of the determination result can be improved. In particular, in the case of a part or the like formed by laminating molten metal, it is possible to reduce the influence of the striped pattern or the like generated by the laminating. That is, even if the surface of a part or the like molded by the fused metal lamination method or the like is polished, the joint surface (“step” of the lamination) at the time of lamination tends to remain as a striped pattern. Such a striped pattern is confusing with a grid pattern or the like projected for inspection, and affects the judgment. Therefore, if the difference between the images is used, these striped patterns are almost erased, and the determination is less likely to be affected. When further improving the accuracy, image quality adjustment or the like may be added to the first image and the second image in order to suppress the influence of the reflected light of the grid pattern to be projected on the first image and the second image before taking the difference.

As described above, the inspection device 1 according to the present embodiment inspects the surface shape of the curved surface of the inspection target, which is formed by melting the material with heat or the inspection target manufactured by polishing the surface. It is a device. The inspection device 1 includes projection devices 11 and 12 that project a specific pattern on the inspection object, and an image pickup device 13 that captures an image of the inspection object on which the pattern is projected. Further, the inspection device 1 was photographed in a state where the same specific pattern as the specific pattern projected on the inspection object is projected on the learning object which is the same type as the inspection object and whose surface shape of the curved surface is known. It has artificial intelligence that learns the set of the image of the learning object and the sensory test result of the surface shape of the curved surface of the learning object as teacher data, and is imaged by the image pickup device 13. The image is applied to the trained artificial intelligence, and the determination circuit 14 for determining the quality of the surface shape of the curved surface of the object to be inspected is provided.

According to this configuration, in order to judge the quality of the surface shape of the curved surface of the inspection object by applying it to the learned artificial intelligence, the surface shape of the curved surface of the inspection object, which is difficult to judge only by the dimensional tolerance, is objectively and. It can be inspected automatically. That is, defects such as "waviness", which were difficult to determine by the conventional numerical method, can be mechanically determined. Further, the inspection can be performed unattended, the determination result can be recorded numerically, and the determination can be eliminated from the inspector's variation as in the conventional case.

<Modification example>
As a modification, a determination circuit using artificial intelligence that individually determines whether to pass or fail may be used for a plurality of defective factors, and the captured images may be determined simultaneously or sequentially.

FIG. 8 is a schematic diagram illustrating the configuration of the determination circuit according to the modified example. As shown in FIG. 8, the determination circuit 14b according to the modified example includes an artificial intelligence 41 for a first defect factor, an artificial intelligence 42 for a second defect factor, and an artificial intelligence 43 for a third defect factor. And prepare.

<Processing during learning>
At the time of learning the artificial intelligence 41, the artificial intelligence 41 for the first defect factor is an image of the learning object which is the same type as the inspection object and is known to have a good curved surface shape. An inspection object for a learning object that is known to have the first defect factor (here, waviness as an example), which is the same type as the inspection object and has the surface shape of the curved surface. An image of the learning object taken with the same specific pattern projected on the object and identification information (here, for example, the first defect) for identifying the first defect factor. The set of is learned in advance as teacher data.

Further, at the time of learning the artificial intelligence 42, the artificial intelligence 42 for the second defect factor is the same type as the inspection object and the surface shape of the curved surface is known to be good for the learning object. A learning target that is known to have a second defect factor (here, as an example, a step) that is the same type as the inspection target and has a curved surface shape that is the same as the combination of the image and the identification information that identifies the non-defective product. Identification information that identifies the image of the learning object taken with the same specific pattern projected on the object and the second defect factor (here, for example, the second defect). The pair with and is learned in advance as teacher data.

Further, at the time of learning the artificial intelligence 43, the artificial intelligence 43 for the third defect factor is the same type as the inspection object and the surface shape of the curved surface is known to be good. An inspection target for a learning object that is known to have a third defect factor (here, as an example, cracking) in the combination of the image and the identification information that identifies the non-defective product, and the surface shape of the curved surface that is the same as the inspection object. Identification information that identifies the third defect factor from the image of the learning object taken with the same specific pattern projected on the object (here, for example, the third defect). The pair with and is learned in advance as teacher data.

<Processing at the time of judgment>
At the time of determination of the artificial intelligence 41, the captured image of the inspection target is input as an input image to the artificial intelligence 41 for the first defect factor, and the good product class and its score set, and the first defective class and its score set. An output matrix 51 including and is output.
At the time of determination of the artificial intelligence 42, the captured image of the inspection target is input as an input image to the artificial intelligence 42 for the second defect factor, and the good product class and its score set, the second defect class and its score. The output matrix 52 including the set is output.
At the time of determination of the artificial intelligence 43, the captured image of the inspection target is input as an input image to the artificial intelligence 43 for the third defective factor, and the good product class and its score set, and the third defective class and its score. The output matrix 53 including the set is output.

These three output matrices 51 to 53 are input to the final determination unit 61. In the final determination unit 61, for example, the score of the good product class of the output matrix 51 is higher than the score of the first defective class, the score of the good product class of the output matrix 52 is higher than the score of the second defective class, and the score of the output matrix 53 is higher. If the score of the good product class is higher than the score of the third bad class, it is judged as a good product. On the other hand, for example, when the score of the first defective class is higher than the score of the good product class of the output matrix 51, the final determination unit 61 determines that it is the first defective. Further, for example, when the score of the second defective class is higher than the score of the good product class of the output matrix 52, the final determination unit 61 determines that the score is the second defect. Further, for example, when the score of the third defective class is higher than the score of the good product class of the output matrix 53, the final determination unit 61 determines that the third defect is defective. In this case, if the score of the first defective class is higher than the score of the good product class of the output matrix 51 and the score of the second defective class is higher than the score of the good product class of the output matrix 52, the final determination unit 61 is to be inspected. It is determined that the object is the first defect and the second defect. In this way, it can be determined that the inspection target has a plurality of defects.

As described above, the determination circuit 14b is provided with artificial intelligence for each cause of failure. Each of the artificial intelligences 41 to 43 is a set of an image of the learning object and identification information for identifying a good product for the learning object which is the same type as the inspection object and is known to have a good curved surface shape. And the same specific pattern as the specific pattern projected on the inspection object is projected for the learning object that is the same type as the inspection object and that the surface shape of the curved surface is known to have the cause of the defect targeted by the artificial intelligence. The set of the image of the learning object taken in the state of being taken and the identification information for identifying the cause of the target defect is used as teacher data to output the certainty of the good product and the certainty of the target defect factor. I'm learning. Each of the artificial intelligences 41 to 43 outputs the certainty of a non-defective product and the certainty of different defective factors for the inspection target by using the captured image of the inspection target. The determination circuit 14b outputs identification information for identifying whether the inspection target is a good product or a defective factor by using the certainty of the good product and the certainty of the cause of the defect output from each artificial intelligence. As a result, the inspector can grasp not only the quality of the inspection object but also the cause of the defect in the case of a defect.

Here, comparing the present embodiment with the modified example, the present embodiment has an advantage that the determination can be made in a short time without consuming much computer resources. On the other hand, in the case of the modified example, it is an advantage that individual defect factors can be appropriately determined even when a plurality of defect factors are compounded.
In most cases, a plurality of defective factors rarely occur at the same time, and even if they do occur, the method of the present embodiment can also determine the score to some extent. In addition, there is no particular inconvenience if it is only a judgment of good or bad. Therefore, although it is more convenient to use this embodiment as compared with the modified example, it is preferable to use it properly according to the intended use.

As described above, the present invention is not limited to the above embodiment as it is, and at the implementation stage, the components can be modified and embodied within a range that does not deviate from the gist thereof. In addition, various inventions can be formed by an appropriate combination of the plurality of components disclosed in the above-described embodiment. For example, some components may be removed from all the components shown in the embodiments. Further, components over different embodiments may be combined as appropriate.

1 Inspection device 11, 12 Projection device 13 Imaging device 14, 14b Judgment circuit 2 Impeller 31 Input image 32 Artificial intelligence 33 Output matrix 34 Final judgment unit 41, 42, 43 Artificial intelligence 51, 52, 53 Output matrix 61 Final judgment unit

Claims (10)

An inspection device that inspects the surface shape of the curved surface of an inspection object that is formed by melting the material with heat, an inspection object that is manufactured by polishing the surface, or an inspection object that is manufactured by cutting. And
A projection device that projects a specific pattern on the inspection object, and
An image pickup device that captures an image of an inspection object on which the pattern is projected, and an image pickup device.
A learning object that is the same as the inspection object and whose surface shape of the curved surface is known to be good or bad. The learning object is photographed with the same specific pattern as the specific pattern projected on the inspection object. It has artificial intelligence that learns the set of the image of the object and the sensory test result of the surface shape of the curved surface of the object for learning as teacher data, and learns the image captured by the image pickup device. A determination circuit that is applied to the artificial intelligence that has already been applied to determine the quality of the surface shape of the curved surface of the inspection object, and
Inspection device equipped with. The inspection device according to claim 1, wherein the specific pattern is a striped pattern or a grid pattern. The inspection according to claim 1 or 2, wherein there are two projection devices, and each projection device projects a grid pattern as the specific pattern by projecting a striped pattern from two directions whose projection directions are substantially orthogonal to each other. Device. The artificial intelligence is a set of an image of a learning object and identification information for identifying a good product for a learning object which is the same type as the inspection object and is known to have a good surface shape of a curved surface, and the above-mentioned. For a learning object that is the same type as the inspection object and is known to have a defective surface shape on the curved surface, a set of an image of the learning object and identification information that identifies the cause of the defect is used as teacher data for a good product. I am learning to output the degree of certainty and the degree of certainty for each cause of failure,
The artificial intelligence uses the captured image of the inspection object to output the certainty of a good product and the certainty of each defect factor for the inspection object.
The determination circuit uses the certainty of the good product and the certainty of the cause of the defect to output identification information for identifying the good product or the cause of the defect for the inspection target, whichever is claimed 1 to 3. The inspection device according to paragraph 1. The determination circuit is provided with artificial intelligence for each cause of failure.
Each of the artificial intelligences is a set of an image of the learning object and identification information for identifying a good product for the learning object which is the same type as the inspection object and is known to have a good curved surface shape, and For a learning object that is the same type as the inspection object and is known to have a defect factor targeted by the artificial intelligence in the surface shape of the curved surface, the image of the learning object and the defect factor targeted by the target are We are learning to output the certainty of good products and the certainty of the cause of the defect to be targeted by using the pair with the identification information to be identified as teacher data.
Each of the artificial intelligences uses the captured image of the inspection object to output the certainty of a non-defective product and the certainty of different defective factors for the inspection object.
The determination circuit uses the certainty of the good product and the certainty of the cause of the defect output from each of the artificial intelligences to obtain identification information for identifying whether the inspection object is a good product or the cause of the defect. The inspection device according to any one of claims 1 to 3 to be output. The projection device can switch between projection and non-projection.
The image pickup apparatus captures the inspection object in a non-projected state to acquire a first image, and images the inspection object in a state where the pattern is projected to acquire a second image. death,
The determination circuit applies the difference image between the first image and the second image to the learned artificial intelligence learned by using the teacher data of the difference image similarly created, and said. The inspection device according to any one of claims 1 to 5, which determines whether the surface shape of the curved surface of the inspection object is good or bad. The inspection device according to any one of claims 1 to 6, wherein the inspection object is a part manufactured by polishing the surface. The inspection device according to any one of claims 1 to 7, wherein the inspection object is a component of a fluid machine. The inspection device according to any one of claims 1 to 8, wherein the inspection object is a part manufactured by a fused metal lamination method or polishing. It is an inspection method for inspecting the surface shape of the curved surface of the inspection object, which is formed by melting the material with heat or the inspection object manufactured by polishing the surface.
The procedure for projecting a specific pattern on the inspection object and
A procedure for imaging an inspection object on which the pattern is projected, a procedure for applying the captured image to learned artificial intelligence, and a procedure for determining the quality of the surface shape of the curved surface.
Have,
The artificial intelligence is photographed in a state where the same specific pattern as the specific pattern projected on the inspection object is projected on the learning object which is the same type as the inspection object and whose surface shape of the curved surface is known. An inspection method in which a set of an image of the learning object and a sensory test result of the quality of the surface shape of the curved surface of the learning object is learned as teacher data.

Images