A casting process control method and device based on artificial intelligence technology

Abstract

The present application provides a casting process control method and device based on artificial intelligence technology, the method acquires a plurality of first casting images with a first defect. Input a plurality of first casting images as samples into a denoising model, and train the denoising model. Receive a casting surface image, and input the casting surface image into a denoising model. Determine the demarcated area of ​​the casting surface image by the denoising model. Determine the second defect in the non-demarcated area of ​​the casting surface image through a pre-trained second defect recognition model. Determine the first matching degree corresponding to the second defect and a plurality of casting processing equipment respectively according to the defect type of the second defect and the position information of the second defect in the casting. When at least one first matching degree meets the preset conditions, generate a process control instruction, and send the process control instruction and the information of each casting processing equipment corresponding to each first matching degree to the corresponding process control terminal.

Description

Technical Field

[0001] The present application relates to the field of casting process technology, and in particular to a casting process control method and equipment based on artificial intelligence technology. Background Art

[0002] The foundry industry plays a very important role in people's lives. Various foundry products, such as wheel hubs, pots, etc., are produced by casting. At present, science and technology are constantly developing, and people's requirements for the tools they need are getting higher and higher. People are beginning to transition from pursuing useful products to pursuing high-quality products. The traditional casting process is mainly carried out manually, which is too wasteful of manpower. Today's large-scale foundry industry has begun to use robots to achieve continuous production of casting processing equipment and save manpower.

[0003] However, once casting processing equipment has produced product surface defects, such as surface cracks that affect consumer experience, subsequent casting processing equipment may also have historical product problems. If the above problems cannot be discovered in time during the casting process and the process flow that causes product surface defects cannot be adjusted in time, this will seriously affect the casting production efficiency and casting production quality.

[0004] Based on this, there is an urgent need for a technical solution that can adjust and control casting processing equipment in a timely manner according to casting problems that have occurred in the past, so as to ensure casting efficiency and casting quality. Summary of the invention

[0005] The embodiments of the present application provide a casting process control method and equipment based on artificial intelligence technology, which are used to timely adjust and control casting processing equipment, ensure casting efficiency and casting quality, and improve consumer experience.

[0006] On the one hand, an embodiment of the present application provides a casting process control method based on artificial intelligence technology, the method comprising:

[0007] Acquire several first casting images with a first defect. Input the multiple first casting images as samples into a denoising model to train the denoising model. Receive a casting surface image and input the casting surface image into a denoising model. Determine the demarcated area of ​​the casting surface image by the denoising model. The demarcated area includes the first defect. Determine the second defect in the non-demarcated area of ​​the casting surface image through a pre-trained second defect recognition model. According to the defect type of the second defect and the position information of the second defect in the casting, determine the first matching degree corresponding to the second defect and several casting processing equipment respectively. Among them, the defect type, the position information of the second defect in the casting and the casting processing equipment information are correspondingly stored in a defect process comparison table. When at least one first matching degree meets the preset conditions, generate a process control instruction, and send the process control instruction and each casting processing equipment corresponding to each first matching degree to the corresponding process control terminal.

[0008] In one implementation of the present application, a plurality of second casting images are input into a second defect recognition model to determine whether a second defect exists in each second casting image. From the plurality of second casting images, each second casting image with a second defect is eliminated to obtain a plurality of first sample images. Each first sample image is screened, and a preset number of first sample images after screening are sent to a user terminal to determine a noise retention result of each first sample image based on an operation of the user on the user terminal. The noise retention result is a result of whether the first defect is included in the first sample image. When each noise retention result matches a preset result, each first sample image is used as a first casting image.

[0009] In one implementation of the present application, the casting type corresponding to each first casting image is determined. It is determined whether the casting types are consistent. When the casting types are consistent, a denoising model corresponding to the casting type is determined. When the casting types are inconsistent, the casting types are classified according to the casting types. And the denoising models matched by the classified casting types are determined.

[0010] In one implementation of the present application, the type of casting corresponding to the casting surface image is determined by an image recognition model. Alternatively, a casting type acquisition instruction is sent to a sending terminal of the casting surface image to determine the casting type of the casting surface image. The denoising model corresponding to the casting type of the casting surface image is determined by the casting type. Determine the demarcation area of ​​the casting surface image by the denoising model, specifically including: determining the edge feature points of the first defect in the casting surface image by the denoising model. Connect the edge feature points in a preset manner to obtain the edge line of the first defect. According to the denoising model and the edge line, determine the defect area located inside the edge line as the demarcated area.

[0011] In one implementation of the present application, according to the denoising model, the defect area of ​​the first defect is determined, and the area of ​​the defect area is determined. The closed area area of ​​the closed area of ​​the edge line is determined. The area corresponding to the maximum value of the area coverage of the closed area area and the defect area area is used as the demarcated area. The surface image of the casting is uploaded to a preset database, and the casting parameters of the casting corresponding to the casting surface image stored in the preset database are determined. Among them, the casting parameters include: casting size, casting density. According to the casting size, the ratio of the area of ​​the defect area inside the edge line to the surface area of ​​the casting is determined. Determine whether the ratio of the area of ​​the defect area inside the edge line to the surface area of ​​the casting is greater than a preset threshold. When it is determined that the ratio of the area of ​​the defect area inside the edge line to the surface area of ​​the casting is greater than the preset threshold, the image corresponding to the defect area is sent to the production terminal.

[0012] In one implementation of the present application, a plurality of process defect images are obtained from a preset defect image database. The process defect image includes a second defect generated by a casting processing device. Through the model input layer, each process defect image is input into the convolution layer of the second defect recognition model to perform convolution processing on each process defect image to obtain a feature image of each process defect image. Each feature image is input into the pooling layer for pooling processing. Through the output layer of the second defect recognition model, the model recognition image, the defect type and the location information of the second defect in the casting are output. Among them, the model recognition image includes an image of the second defect. Each model recognition image and the corresponding defect type and the location information of the second defect in the casting are uploaded to the image comparison database to determine whether the second defect in each model image, the corresponding defect type and the location information of the second defect in the casting match, and the number of matches is accumulated. According to the accumulated number of matches and the corresponding matching results, the second matching degree is calculated. When the second matching degree is less than the second preset threshold, a plurality of process defect images are obtained, and the second defect recognition model is retrained until the second matching degree is greater than or equal to the second preset threshold, and the training of the second defect recognition model is completed.

[0013] In one implementation of the present application, a defect process comparison table of the previous time period in a preset database is determined. The defect process comparison table records the matching degree of each casting processing equipment with each defect type, the location information of the second defect in the casting, and each casting processing equipment. The defect process comparison table is generated according to the historical records of each casting processing equipment and the defect type and the location information of the second defect in the casting. The defect type and the location information of the second defect in the casting are compared with the first sequence group in the defect process comparison table to determine whether the defect type and the location information of the second defect in the casting exist in the defect process comparison table. The first sequence group includes several defect types and defect location information. If so, the matching sequence in the second sequence group corresponding to the defect type and the location information of the second defect in the casting is determined, so as to determine several first matching degrees according to the matching sequence. The second sequence group includes each casting processing equipment matched with each defect type and the location information of the second defect in the casting in the first sequence group and the corresponding matching degree. In one implementation of the present application, at least one first matching degree corresponding to the maximum value of several first matching degrees is determined, and it is determined whether at least one first matching degree is greater than a preset value. When at least one first matching degree is greater than a preset value, it is determined that the proportion of at least one first matching degree meets a preset condition, and a process control instruction is generated, wherein the process control instruction is used to adjust the equipment operation of the casting processing equipment corresponding to the process control terminal.

[0014] In one implementation of the present application, when there is no first matching degree that satisfies the preset condition among the first matching degrees, a number of process defect images are acquired to retrain the second defect recognition model, and the regional image corresponding to the second defect is sent to each process control terminal of each casting processing equipment.

[0015] On the other hand, the embodiment of the present application also provides a casting process control device based on artificial intelligence technology, the device comprising:

[0016] At least one processor; and a memory in communication with the at least one processor. The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0017] Acquire several first casting images with a first defect. Input the multiple first casting images as samples into a denoising model to train the denoising model. Receive a casting surface image and input the casting surface image into a denoising model. Determine the demarcated area of ​​the casting surface image by the denoising model. The demarcated area includes the first defect. Determine the second defect in the non-demarcated area of ​​the casting surface image through a pre-trained second defect recognition model. According to the defect type of the second defect and the position information of the second defect in the casting, the first matching degree corresponding to the second defect and several casting processing equipment is obtained. Among them, the defect type, the position information of the second defect in the casting, and the casting processing equipment information are correspondingly stored in a defect process comparison table. When at least one first matching degree meets the preset conditions, a process control instruction is generated, and the process control instruction and the casting processing equipment information corresponding to each first matching degree are sent to the corresponding process control terminal.

[0018] Through the above scheme, it is possible to accurately identify casting problems caused by casting processing equipment, and according to the casting problems that have occurred in the past, timely adjust and control the casting processing equipment that caused the casting problems to ensure casting efficiency and casting quality. At the same time, this application can avoid the situation where the subsequent production of casting products continues the previous problems after the casting processing equipment has problems, and produce batches of inferior products. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0020] Figure 1 A schematic diagram of a flow chart of a casting process control method based on artificial intelligence technology in an embodiment of the present application;

[0021] Figure 2 This is another flow chart of a casting process control method based on artificial intelligence technology in an embodiment of the present application;

[0022] Figure 3 This is a schematic diagram of the structure of a casting process control device based on artificial intelligence technology in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0024] The embodiments of the present application provide a casting process control method and equipment based on artificial intelligence technology, which are used to adjust and control casting processing equipment in a timely manner according to casting problems that have occurred in the past, so as to ensure casting efficiency and casting quality.

[0025] The following describes in detail various embodiments of the present application in conjunction with the accompanying drawings.

[0026] The present application embodiment provides a casting process control method based on artificial intelligence technology, such as Figure 1 As shown, the method may include steps S101-S107:

[0027] S101, the server obtains a plurality of first casting images having a first defect.

[0028] In the embodiment of the present application, before the server obtains a plurality of first casting images having a first defect, the process includes:

[0029] First, the server inputs a plurality of second casting images into a second defect recognition model to determine whether a second defect exists in each of the second casting images.

[0030] The second defect recognition model is a convolutional neural network model, which is trained by a number of process defect images and is used to identify the second defect.

[0031] The second defect is a defect caused by the casting processing equipment. The second defect and the first defect are obtained by comparing the defect positions and defect types of several casting images. For the comparison difference, the server can compare the first defect with the second defect through a neural network model, or obtain the characteristics of the second defect and the characteristics of the first defect by manually marking features.

[0032] Then, the server removes the second casting images having the second defect from the plurality of second casting images to obtain a plurality of first sample images.

[0033] When the second defect recognition model recognizes that the second casting image has the second defect, the server may remove the second casting image from the plurality of second casting images and use the remaining second casting images excluding the second casting images with the second curve as the first sample images.

[0034] Next, the server screens the first sample images and sends a preset number of the screened first sample images to the user terminal, so as to determine the noise retention result of each first sample image based on the operation of the user on the user terminal.

[0035] The noise retention result is a result of whether the first sample image includes the first defect.

[0036] The number of first sample images may be too many, and the server may select a preset number of first sample images from the first sample images by screening. The preset number is set during actual use, and this application does not specifically limit this. The user may select the first sample image as a result including the first defect or a result excluding the first defect by clicking on each first sample image displayed on the display interface of the user terminal.

[0037] Finally, the server uses each first sample image as the first casting image when each noise retention result matches the preset result.

[0038] The preset result may be a result including the first defect. If all the noise retention results of the first sample images after screening are results including the first defect, the server uses each first sample image as the first casting image.

[0039] Through the above scheme, the present application can obtain a first sample image containing only the first defect, thereby avoiding the existence of the second defect and affecting the training result of the denoising model when training the denoising model, thereby ensuring the accuracy of the denoising model in identifying the first defect.

[0040] It should be noted that the server, as the executor of the casting process control method based on artificial intelligence technology, exists only for exemplary purposes. The executor is not limited to the server, and this application does not make any specific limitations on this.

[0041] S102: The server inputs a plurality of first casting images as samples into a denoising model to train the denoising model.

[0042] In the embodiment of the present application, the denoising model needs to be trained before it can be used normally. Therefore, before the server inputs the plurality of first casting images as samples into the denoising model to train the denoising model, the method further includes:

[0043] First, the server determines the casting type corresponding to each first casting image.

[0044] The casting types may include at least pump castings, valve castings, hydraulic castings, pneumatic component castings, metallurgical machinery castings, mining machinery castings, etc. The server obtains the casting type by image recognition or by determining it to the sending terminal.

[0045] The server then determines whether the casting types are consistent.

[0046] The sending terminal may send several first casting images, and the casting types corresponding to the first casting images may be inconsistent. Therefore, in order to ensure the accuracy of defect identification and the accuracy of casting processing equipment regulation, the server needs to first determine the consistency of the casting types.

[0047] Next, when the casting types are consistent, the server determines a denoising model that matches the name of the casting type.

[0048] When the casting types are inconsistent, the server classifies the casting types according to the names of the casting types.

[0049] Finally, the server determines each denoising model that matches the name of each classified casting type.

[0050] Through the above scheme, the denoising model corresponding to each casting type can be determined, and the first defect of the casting surface image can be identified more accurately through classification.

[0051] S103, the server receives the casting surface image, and inputs the casting surface image into a denoising model.

[0052] In the embodiment of the present application, before the server inputs the casting surface image into the denoising model, Figure 2 As shown, the following steps are also included:

[0053] S201, the server determines the casting type corresponding to the casting surface image through an image recognition model. Alternatively, the server sends a casting type acquisition instruction to a terminal sending the casting surface image to determine the casting type of the casting surface image.

[0054] The server can obtain the casting type of the casting surface image through an image recognition model or by obtaining it from a sending terminal. The casting surface image can be an image collected by an image acquisition device during the casting processing equipment, or it can be an image collected by an image acquisition device after the casting production is completed. Among them, the image recognition model includes but is not limited to: AlexNet, VGG19, ResNet_152, InceptionV4, DenseNet. The image acquisition device can be a mobile phone, a camera or other equipment, and this application does not specifically limit this.

[0055] S202: The server determines a denoising model corresponding to the casting type of the casting surface image according to the casting type.

[0056] After obtaining the casting type, the server selects a denoising model that matches the name of the casting type from a number of denoising models according to the name of the casting type. For example, if the name of the casting type is: wheel hub, then the name of the denoising model is wheel hub denoising model.

[0057] S203: The server determines edge feature points of the first defect in the casting surface image through a denoising model.

[0058] After the server inputs the casting surface image into the denoising model, the denoising model determines the area of ​​the first defect in the casting surface image and marks the edge feature points of the area of ​​the first defect. The area may be a regular shape, but the edge feature points can mark the edge points of the first defect in the regular image.

[0059] S204, the server connects the edge feature points in a preset manner to obtain an edge line of the first defect.

[0060] The server may connect adjacent edge feature points in a clockwise or counterclockwise manner in sequence, and the connected edge feature points form an edge line, which may form a regular shape or an irregular shape.

[0061] S205: The server determines the defect area inside the edge line as the demarcated area based on the denoising model and the edge line.

[0062] The server determines the defect region of the first defect and the area of ​​the defect region according to the denoising model. The closed region area of ​​the closed region of the edge line is determined. The area corresponding to the maximum value of the area coverage ratio of the closed region area and the defect region area is used as the demarcated area.

[0063] That is, the server obtains the area of ​​the first defect and the edge line of the first defect according to the denoising model, and takes the area with the largest coverage ratio of the closed area of ​​the edge line and the area coverage area of ​​the first defect as the demarcated area.

[0064] S206, the server uploads the casting surface image to a preset database, and determines casting parameters of the casting corresponding to the casting surface image stored in the preset database.

[0065] Among them, casting parameters include: casting size and casting density.

[0066] The server may send the casting surface image from the image acquisition device to a preset database, which summarizes and stores casting parameters of each casting corresponding to the casting surface image, such as casting parameters of hub A.

[0067] In addition, the preset database can actively identify the casting surface image and obtain the name of the casting. Or the server recognizes the casting surface image, obtains the name of the casting corresponding to the casting surface image, and sends the casting surface image and the casting name to the preset database, and the preset database compares the casting surface image with the image in the database, and compares the casting name with the name in the preset database, and determines the casting parameters of the matched casting after comparison.

[0068] S207, the server determines the ratio of the area of ​​the defect region inside the edge line to the surface area of ​​the casting according to the size of the casting.

[0069] The server can calculate the area of ​​the defective region inside the edge line of the casting corresponding to the casting surface image according to the casting size, and calculate the ratio of the area to the surface area of ​​the casting. For example, the area of ​​the defective region inside the edge line of the casting is S1, the surface area of ​​the casting is S2, and the ratio is S1 / S2.

[0070] S208, the server determines whether the ratio of the area of ​​the defective region inside the edge line to the surface area of ​​the casting is greater than a preset threshold.

[0071] In the embodiment of the present application, the preset threshold value can be set according to the actual use process, such as 0.5, 0.4, and the specific data of the preset threshold value of the present application is not limited.

[0072] S209, when the server determines that the ratio of the area of ​​the defective area inside the edge line to the surface area of ​​the casting is greater than a preset threshold, the server sends an image corresponding to the defective area to the production terminal.

[0073] In an embodiment of the present application, if the area of ​​the first defect on the surface of the casting is too large, although the first defect is not caused by the casting processing equipment, its defect will affect the appearance or usage experience. Therefore, when the area of ​​the first defect is too large, the image corresponding to the defective area is collected and the image is sent to the production terminal, so that the production terminal selects to perform production operations on the casting corresponding to the first defect, such as recasting, or repairing the first defect by other means.

[0074] Through the above scheme, the area corresponding to the first defect can be accurately identified, so that the production terminal can discover the problem of the first defect, so that the casting can be promptly repaired or recast, ensuring the quality of the casting products and improving the quality of casting production. At the same time, when the first defect can be identified between casting processing equipment, it can be ensured that the subsequent casting processing equipment is not affected by the existence of the first defect, thereby ensuring the normal production of the casting processing equipment.

[0075] S104, the server determines the demarcation area of ​​the casting surface image by the denoising model.

[0076] Wherein, the demarcated area includes the first defect.

[0077] An example of determining the demarcated area is the above-mentioned step S205, which will not be described in detail here.

[0078] S105, the server determines the second defect in the non-demarcated area of ​​the casting surface image by using a pre-trained second defect recognition model.

[0079] In the embodiment of the present application, before the server determines the second defect in the non-demarcated area of ​​the casting surface image by using the pre-trained second defect recognition model, the server also includes:

[0080] First, the server obtains a number of process defect images from a preset defect image database.

[0081] The process defect image includes a second defect caused by the casting processing equipment.

[0082] The server can establish a connection with a preset defect image database and train the first defect recognition model through process defect images in the defect image database. Several process defect images correspond to different defect information, and the defect information at least includes: a defect type protrusion at the defect position, b defect type crack at the defect position.

[0083] Secondly, the server inputs each process defect image into the convolution layer of the second defect recognition model through the model input layer to perform convolution processing on each process defect image to obtain a feature image of each process defect image.

[0084] The characteristic image includes the characteristics of the second defect.

[0085] Again, the server inputs each feature image into the pooling layer for pooling processing.

[0086] Next, the server outputs the model recognition image, the defect type, and the location information of the second defect in the casting through the output layer of the second defect recognition model.

[0087] The model recognizes an image including the second defect in the image.

[0088] Subsequently, the server uploads each model identification image, defect type and location information of the second defect in the casting to the image comparison database to determine whether the second defect in each model image matches the corresponding defect type and location information of the second defect in the casting, and accumulates the number of matches.

[0089] The server can match the defect type and defect location with the second defect of the model image in the image comparison database by determining whether the defect type and defect location are consistent, and accumulate the total number of matches.

[0090] Subsequently, the server calculates a second matching degree according to the accumulated matching times and the corresponding matching results.

[0091] The second matching degree is calculated by determining the cumulative number of matches X, the number of matching coincidences is T, and the second matching degree Y=T / X.

[0092] Finally, when the second matching degree is less than the second preset threshold, the server obtains several process defect images and retrains the second defect recognition model until the second matching degree is greater than or equal to the second preset threshold, thereby completing the training of the second defect recognition model.

[0093] When the second matching degree is less than the second preset threshold, the server determines that the accuracy of the second defect recognition model is insufficient. In actual use, the setting of the second preset threshold affects the accuracy of the second defect recognition model. Therefore, the second preset threshold can be set according to actual needs.

[0094] S106: The server determines first matching degrees corresponding to the second defect and a plurality of casting processing equipment respectively according to the defect type of the second defect and the position information of the second defect in the casting.

[0095] Among them, the defect type, the location information of the second defect in the casting, and the casting processing equipment information are correspondingly stored in the defect process comparison table.

[0096] In the embodiment of the present application, the server determines a number of first matching degrees according to the defect information of the second defect, specifically including:

[0097] First, the server determines the defective process comparison table of the previous time period in the preset database.

[0098] Among them, the defect process comparison table records the types of defects generated by each casting processing equipment, the location information of the second defect in the casting and the matching degree of each casting processing equipment. The defect process comparison table is generated based on the historical records of each casting processing equipment and the defect type, and the location information of the second defect in the casting.

[0099] The defect process comparison table is generated before the current casting surface image is sent. The defect process comparison table contains historical data recorded after defect information occurs in the casting processing equipment. The defect information includes the defect type and the location information of the second defect in the casting. For example, the casting processing equipment N1 corresponds to the casting information m1, and the matching degree is n1. The casting processing equipment N2 corresponds to the casting information m1, and the matching degree is n2. The defect process comparison table records the historical records of the second defect that has occurred in the casting processing equipment of the casting corresponding to the casting surface image in the enterprise or manufacturer in the past.

[0100] Then, the server compares the defect type in the defect information and the location information of the second defect in the casting with the first sequence group in the defect process comparison table to determine whether the defect type and the location information of the second defect in the casting are in the defect process comparison table.

[0101] The first sequence group includes information on several defect types and defect locations.

[0102] Next, when it is determined that there is a defect process comparison table for the defect type and the location information of the second defect in the casting, the server determines the matching sequence in the second sequence group corresponding to the defect type and the location information of the second defect in the casting, so as to determine a number of first matching degrees based on the matching sequence.

[0103] The second sequence group includes each casting processing equipment that matches each defect type in the first sequence group and the position information of the second defect in the casting and the corresponding matching degree.

[0104] The second sequence group records the matching degree of each casting processing equipment and the defect information corresponding to each casting processing equipment.

[0105] When the server determines that the defect information does not exist in the defect process comparison table, it sends the image corresponding to the defect information to the management terminal, and the experts at the management terminal identify the matching degree of the defect information corresponding to each casting processing equipment.

[0106] Through the above scheme, the casting processing equipment for the casting problems that have occurred in the past can be generated in time, so that the casting processing equipment can be adjusted and controlled in time.

[0107] S107, when at least one first matching degree meets a preset condition, the server generates a process control instruction, and sends the process control instruction and each casting processing equipment corresponding to each first matching degree to a corresponding process control terminal.

[0108] The preset condition is used to determine at least one first matching degree ratio.

[0109] The process control terminal may correspond to a casting processing device, for example, each casting processing device corresponds to a process control terminal, or multiple casting processing devices correspond to a process control terminal. The process control terminal may be a computer, a mobile phone, a server, etc., and this application does not specifically limit this.

[0110] In the embodiment of the present application, when at least one first matching degree satisfies a preset condition, the server generates a process control instruction, specifically including:

[0111] The server determines at least one first matching degree corresponding to a maximum value among the plurality of first matching degrees, and determines whether the at least one first matching degree is greater than a preset value.

[0112] When at least one first matching degree is greater than a preset value, the server determines that a proportion of at least one first matching degree meets a preset condition and generates a process control instruction.

[0113] Among them, the process control instruction is used to adjust the equipment operation of the corresponding casting processing equipment of the process control terminal, for example: stop operation, continue operation, etc.

[0114] In another embodiment of the present application, when there is no first matching degree satisfying the preset condition among the first matching degrees, the server obtains a number of process defect images, retrains the second defect recognition model, and sends the regional image corresponding to the second defect to each process control terminal of each casting processing equipment.

[0115] Through the above scheme, the present application can accurately identify casting problems caused by casting processing equipment, and according to the casting problems that have occurred in the past, timely adjust and control the casting processing equipment that caused the casting problems, so as to ensure the casting efficiency and casting quality, and to ensure the number of qualified products produced. While improving the casting quality and the number of qualified products, it can provide consumers with a better purchasing and use experience.

[0116] Figure 3 A schematic diagram of a casting process control device based on artificial intelligence technology provided in an embodiment of the present application, the device comprises:

[0117] At least one processor; and a memory in communication with the at least one processor. The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0118] Acquire several first casting images with a first defect. Input the multiple first casting images as samples into a denoising model to train the denoising model. Receive a casting surface image and input the casting surface image into a denoising model. Determine the demarcated area of ​​the casting surface image by the denoising model. The demarcated area includes the first defect. Determine the second defect in the non-demarcated area of ​​the casting surface image through a pre-trained second defect recognition model. According to the defect type of the second defect and the position information of the second defect in the casting, determine the first matching degree corresponding to the second defect and several casting processing equipment respectively. Among them, the defect type, the position information of the second defect in the casting and the casting processing equipment information are correspondingly stored in a defect process comparison table. When at least one first matching degree meets the preset conditions, generate a process control instruction, and send the process control instruction and the casting processing equipment information corresponding to each first matching degree to the corresponding process control terminal.

[0119] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0120] The device and method provided in the embodiments of the present application correspond one to one, and therefore, the device also has similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device will not be repeated here.

[0121] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0122] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A casting process control method based on artificial intelligence technology, characterized in that: The method comprises: Acquire a plurality of first casting images having a first defect; Inputting a plurality of the first casting images as samples into a denoising model to train the denoising model; Receiving a casting surface image, and inputting the casting surface image into the denoising model; Determining a region defined by the denoising model for the casting surface image; the region defined includes the first defect; Determining a second defect in a non-defined area of ​​the casting surface image by using a pre-trained second defect recognition model; Determine, according to the defect type of the second defect and the position information of the second defect in the casting, the first matching degree corresponding to the second defect and a plurality of casting processing equipment respectively; wherein the defect type, the position information of the second defect in the casting and the casting processing equipment information are correspondingly stored in the defect process comparison table; When at least one first matching degree satisfies a preset condition, a process control instruction is generated, and the process control instruction and the casting processing equipment information corresponding to each first matching degree are sent to a corresponding process control terminal; Wherein, according to the defect type of the second defect and the position information of the second defect in the casting, determining the first matching degree of the second defect of the casting processing equipment and the first matching degree corresponding to the plurality of casting processing equipment respectively includes: Determine a defect process comparison table for the previous time period in a preset database; wherein the defect process comparison table records the matching degree of each casting processing equipment with respect to each defect type, the location information of the second defect in the casting, and each casting processing equipment, and the defect process comparison table is generated according to the historical records of each casting processing equipment, the defect type, and the location information of the second defect in the casting; Compare the defect type and the position information of the second defect in the casting with the first sequence group in the defect-process comparison table to determine whether the defect type and the position information of the second defect in the casting are in the defect-process comparison table; wherein the first sequence group includes a plurality of defect types and defect position information; If so, determine the matching sequence in the second sequence group corresponding to the defect type and the position information of the second defect in the casting, and determine a number of the first matching degrees based on the matching sequence; wherein the second sequence group includes each of the casting processing equipment that matches each of the defect types and the position information of the second defect in the casting in the first sequence group and the corresponding matching degrees.

2. The method according to claim 1, characterized in that: Before acquiring a plurality of first casting images having a first defect, the method further includes: Inputting a plurality of second casting images into the second defect recognition model to determine whether the second defect exists in each of the second casting images; From a plurality of second casting images, remove the second casting images having the second defect to obtain a plurality of first sample images; Screening the first sample images, and sending a preset number of the screened first sample images to a user terminal, so as to determine a noise retention result of the first sample images based on an operation of a user on the user terminal; wherein the noise retention result is a result of whether the first defect is included in the first sample image; In the case where each of the noise retention results matches a preset result, each of the first sample images is used as the first casting image.

3. The method according to claim 1, characterized in that: Before inputting the plurality of first casting images as samples into a denoising model to train the denoising model, the method further comprises: determining a casting type corresponding to each of the first casting images; Determining whether the types of the castings are consistent; When the types of the castings are consistent, determining the denoising model corresponding to the casting type; In the case where the types of the castings are inconsistent, classifying the types of the castings according to the types of the castings; And determine the denoising models that match the classified casting types.

4. The method according to claim 1, characterized in that: Before inputting the casting surface image into the denoising model, the method further comprises: Determining the casting type corresponding to the casting surface image by an image recognition model; or Sending a casting type acquisition instruction to a terminal sending the casting surface image to determine the casting type of the casting surface image; Determining the denoising model corresponding to the casting type of the casting surface image according to the casting type; Determining the delimited area of ​​the casting surface image by the denoising model specifically includes: Determining edge feature points of the first defect in the casting surface image by using the denoising model; Connecting the edge feature points in a preset manner to obtain an edge line of the first defect; According to the denoising model and the edge line, a defect area located inside the edge line is determined as the demarcated area.

5. The method according to claim 4, characterized in that: According to the denoising model and the edge line, determining a defect area located inside the edge line as the demarcated area specifically includes: Determining a defect region of the first defect and an area of ​​the defect region according to the denoising model; Determine the closed region area of ​​the closed region of the edge line; The area corresponding to the maximum value of the area coverage ratio of the closed area and the defect area is used as the demarcated area; According to the denoising model and the edge line, a defect area located inside the edge line is determined, and after the area is demarcated, the method further includes: Uploading the casting surface image to a preset database, and determining casting parameters of the casting corresponding to the casting surface image stored in the preset database; wherein the casting parameters include: casting size and casting density; Determine, according to the size of the casting, the ratio of the area of ​​the defect region inside the edge line to the surface area of ​​the casting; Determining whether the ratio of the area of ​​the defective region inside the edge line to the surface area of ​​the casting is greater than a preset threshold; If so, the image corresponding to the defective area is sent to the production terminal.

6. The method according to claim 1, characterized in that: Before determining the second defect in the non-defined area of ​​the casting surface image by using the pre-trained second defect recognition model, the method further includes: Acquire a plurality of process defect images from a preset defect image database; the process defect images include the second defect generated by the casting processing equipment; Inputting each of the process defect images into the convolution layer of the second defect recognition model through the model input layer to perform convolution processing on each of the process defect images to obtain a feature image of each of the process defect images; Input each of the feature images into a pooling layer for pooling processing; Outputting a model recognition image, the defect type, and location information of the second defect in the casting through the output layer of the second defect recognition model; wherein the model recognition image includes an image of the second defect; Uploading each of the model recognition images, the corresponding defect type, and the location information of the second defect in the casting to an image comparison database to determine whether the second defect in each of the model recognition images matches the corresponding defect type and the location information of the second defect in the casting, and accumulating the number of matches; Calculating a second matching degree according to the accumulated number of matching times and the corresponding matching results; When the second matching degree is less than the second preset threshold, a number of process defect images are acquired and the second defect recognition model is retrained until the second matching degree is greater than or equal to the second preset threshold, thereby completing the training of the second defect recognition model.

7. The method according to claim 1, characterized in that: When at least one first matching degree satisfies a preset condition, generating a process control instruction specifically includes: Determine at least one first matching degree corresponding to a maximum value among the plurality of first matching degrees, and determine whether the at least one first matching degree is greater than a preset value; When at least one first matching degree is greater than the preset value, it is determined that the proportion of the at least one first matching degree meets the preset condition, and the process control instruction is generated; wherein, the process control instruction is used to adjust the equipment operation of the casting processing equipment corresponding to the process control terminal.

8. The method according to claim 1, characterized in that: The method further comprises: When there is no first matching degree satisfying the preset condition among the first matching degrees, acquiring a plurality of process defect images and retraining the second defect recognition model; And the regional image corresponding to the second defect is sent to each process control terminal of each casting processing equipment.

9. A casting process control device based on artificial intelligence technology, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a casting process control method based on artificial intelligence technology as described in any one of claims 1-8.

Images