KR102314599B1 - quality inspection apparatus for gas injection molding products and method thereof - Google Patents

Abstract

The present invention relates to an apparatus and method for quality inspection of gas injection molded products.
According to the present invention, a thermal image acquisition unit that receives a thermal image generated by shooting an injection-molded target product by injecting gas from a thermal imaging camera, analyzes the acquired thermal image to detect the target product, and detects the detected target product It includes a temperature data extraction unit for extracting surface temperature data within the area of , and a determination unit for determining whether the target product is defective by inputting the extracted temperature data into a machine learning algorithm that has been trained.
As described above, according to the present invention, a non-destructive inspection is possible because a determination result for a thermal image of a gas injection-molded product is automatically provided using a machine learning algorithm, and an increase in the work speed for detecting product defects and a reduction in labor costs can do it

Description

Quality inspection apparatus for gas injection molding products and method thereof

The present invention relates to a quality inspection apparatus and method for a gas injection-molded product, and more particularly, to determine whether a product is defective by using a thermal image of the injection-molded product by injecting gas, but using a machine learning algorithm. It relates to a quality inspection apparatus and method for gas injection molded products capable of non-destructive inspection.

In general, various parts are used to produce one automobile, and as the physical and chemical properties of plastics are improved due to the purpose of lowering the weight of the vehicle body and the development of technology to increase the fuel efficiency of the automobile, the use of plastic parts in automobile manufacturing usage is increasing.

Injection molding is a method in which a molten resin is injected into the inside of a mold and cooled to form a product. Compared to other molding methods such as compression molding and extrusion molding, the shape of the product is It is widely used in the molding of plastic products because of its limited size and excellent productivity and work efficiency. The injection molding mold includes an upper mold and a lower mold that are spaced apart from or in close contact with each other, and a resin injection machine is connected to one of the upper mold and the lower mold. The resin injection machine injects the molten resin into the mold connected to the resin injection machine when the upper mold and the lower mold are closed, that is, when they are in close contact with each other. The molten resin injected into the mold is injected and cured into a cavity corresponding to the shape of the injection molded product through a sprue, a runner, and a gate, and the upper mold and the lower mold are formed. When the mold is opened, that is, when the resin is in a state spaced apart from each other, an injection-molded product formed by curing the resin is taken out.

On the other hand, injection-molded products generate defective products due to the thickness, injection pressure, and cavity of the molded product.

In order to detect the generated defective products, conventionally, a visual inspection method has been mainly used. In the visual inspection method, the operator visually checks the product one by one and judges whether it is defective. When a defect is detected using the visual inspection method, the operation speed is slow, the efficiency is low because the re-inspection is performed according to the subjective judgment of the operator, and the precision of the defect inspection is low.

The technology underlying the present invention is disclosed in Korean Patent Application Laid-Open No. 10-2019-0119475 (published on October 22, 2019).

The technical problem to be achieved by the present invention is to determine whether the product is defective by using a thermal image of the injection-molded product by injecting gas, but by using a machine learning algorithm, a quality inspection device for gas injection-molded products capable of non-destructive inspection and The purpose is to provide a method.

According to an embodiment of the present invention for achieving this technical problem, in the apparatus for inspecting the quality of a gas injection-molded product, a thermal image that receives a thermal image generated by photographing a target product injection-molded by injecting gas is transmitted from a thermal imaging camera An acquisition unit, a temperature data extraction unit that analyzes the acquired thermal image to detect a target product, and a temperature data extractor that extracts surface temperature data within the area of the detected target product, and applies the extracted surface temperature data to a pre-learned machine learning algorithm. and a determination unit for determining whether the target product is defective by input.

The quality inspection device extracts the surface temperature distribution in the region corresponding to the product from the thermal images of the pre-determined normal product and the abnormal product, and a learning unit for learning by inputting the extracted surface temperature distribution into a pre-established machine learning algorithm. may include

The temperature data extractor may detect a region above a preset temperature as the current position of the target product by using a distribution of temperature data included in the thermal image.

The temperature data extractor may generate a one-dimensional or two-dimensional grid in an area corresponding to the detected target product, and sample a temperature distribution of the generated grid to generate a surface temperature distribution diagram of the target product.

The temperature data extraction unit may generate a surface temperature average value using the surface temperature distribution of the generated target product.

The learning unit generates a dataset of a normal product and a dataset of a defective product by using the temperature data extracted from the thermal image of the normal product and the temperature data extracted from the thermal image of the abnormal product, and uses each of the generated datasets. It can be used to train machine learning algorithms.

In addition, in the quality inspection method of a gas injection-molded product using the quality inspection apparatus according to an embodiment of the present invention, the step of receiving a thermal image generated by photographing the injection-molded product by injecting gas from the thermal imaging camera, obtaining Analyze a thermal image to detect the position of the product, extract the surface temperature data within the detected product area, and input the surface temperature data of the target product to be judged to the machine learning algorithm previously learned. Including the step of determining whether the defect.

As described above, according to the present invention, a non-destructive inspection is possible because a determination result for a thermal image of a gas injection molded product is automatically provided using a machine learning algorithm, and an increase in the work speed for detecting product defects and a reduction in labor costs can do it

1 is a block diagram of a quality inspection apparatus according to an embodiment of the present invention.
2 is a flowchart illustrating a quality inspection method of a gas injection molded product using a quality inspection apparatus according to an embodiment of the present invention.
FIG. 3 is a flowchart for explaining step S230 shown in FIG. 2 .

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings. In this process, the thickness of the lines or the size of the components shown in the drawings may be exaggerated for clarity and convenience of explanation.

In addition, the terms described below are terms defined in consideration of functions in the present invention, which may vary according to the intention or custom of the user or operator. Therefore, definitions of these terms should be made based on the content throughout this specification.

Hereinafter, a quality inspection apparatus for a gas injection molded product according to an embodiment of the present invention will be described with reference to FIG. 1 .

1 is a block diagram of a quality inspection apparatus according to an embodiment of the present invention.

As shown in FIG. 1 , the quality inspection apparatus 100 according to an embodiment of the present invention includes a thermal image acquisition unit 110 , a temperature data extraction unit 120 , a learning unit 130 , and a determination unit 140 . include

First, the thermal image acquisition unit 110 acquires a thermal image from the thermal imaging camera. The thermal imaging camera photographs the product immediately after gas molding and injection, and transmits the photographed image to the thermal image acquisition unit 110 .

The temperature data extraction unit 120 extracts temperature data of the obtained thermal image.

In other words, the temperature data extraction unit 120 detects the position of the product in the thermal image by using the obtained distribution of the temperature data. Then, the temperature data extraction unit 120 generates a distribution map of the surface temperature inside the product by using the location information of the product. Then, the average value is calculated using the distribution for the surface temperature inside the product.

The learning unit 130 builds a machine learning algorithm, inputs the surface temperature distribution diagram of the extracted product to the built machine learning algorithm, and learns to determine whether the product is defective.

Finally, the determination unit 140 determines whether the target product is defective by inputting the surface temperature distribution diagram of the target product into a previously built machine learning algorithm.

In other words, the determination unit 140 inputs the surface temperature distribution of the target product to the machine learning algorithm previously learned. Then, the machine learning algorithm determines whether the target product is defective using the surface temperature distribution of the product. Then, the determination unit 140 receives and outputs the result determined through the machine learning algorithm.

Hereinafter, a quality inspection method of a gas injection molded product using a quality inspection apparatus according to an embodiment of the present invention will be described in more detail with reference to FIGS. 2 and 3 .

2 is a flowchart illustrating a quality inspection method of a gas injection molded product using a quality inspection apparatus according to an embodiment of the present invention.

The method for inspecting the quality of a gas injection molded product according to an embodiment of the present invention includes a step of learning a machine learning algorithm and a test step of determining whether a target product is defective through the built machine learning algorithm. As shown in FIG. 2 , the learning step includes steps S210 to S230, and the test step includes steps S240 to S260.

First, the thermal image acquisition unit 110 acquires a thermal image of the injection-molded product (S210). In other words, gas injection molding products are injection-formed by injecting gas into a synthetic resin melted at a high temperature.

Accordingly, the thermal image acquisition unit 110 acquires a thermal image obtained by photographing the product from the thermal imaging camera. In this case, the acquired thermal image represents a thermal image in which the user has determined that the product is a normal product or a defective product.

Then, the temperature data extraction unit 120 extracts data on the surface temperature of the product by using the obtained thermal image (S220).

In other words, in gas injection molding, the space size inside the product is differently molded according to the amount of injected gas. In addition, when the injection-molded product is photographed with a thermal imaging camera, different surface temperatures are measured according to the size of the space inside the product.

Accordingly, the temperature data extraction unit 120 acquires a temperature distribution diagram of the thermal image in which determination as a normal product is completed and a temperature distribution diagram in a thermal image in which determination as an abnormal product is completed.

Next, the learning unit 130 generates each data set by using the temperature data distribution extracted for each normal product and the abnormal product, and the results classified by the user as the normal product and the abnormal product.

Then, the learning unit 130 learns to determine whether the product is defective by inputting the generated data set into a pre-established machine learning algorithm (S230).

In other words, the learning unit 130 obtains each temperature data distribution diagram from a plurality of thermal images determined as normal products, and generates a normal product training dataset using the obtained temperature data distribution diagram. Then, the learning unit 130 obtains each temperature data distribution diagram from a plurality of thermal images read as defective products, and generates a defective product training dataset by using the obtained temperature data distribution diagram. Then, the learning unit 130 inputs the generated normal product training dataset and defective product training dataset, respectively, to a pre-established machine learning algorithm. Then, the machine learning algorithm is trained to determine whether the product is defective from the input temperature data distribution using the input normal product training dataset and defective product training dataset.

When the learning of the machine learning algorithm is completed, the thermal image acquisition unit 110 acquires a thermal image to be determined from the thermal imaging camera ( S240 ).

In other words, since the molten resin is injected under high pressure, thermal images with different surface temperatures are obtained for each product. And the obtained thermal image is transmitted to the temperature data extraction unit 120 .

Then, the temperature data extraction unit 120 extracts the surface temperature data of the target product using the received thermal image (S250).

Hereinafter, step S250 will be described in more detail with reference to FIG. 3 .

FIG. 3 is a flowchart for explaining step S250 shown in FIG. 2 .

As shown in FIG. 3 , the temperature data extraction unit 120 detects the current position of the target product using the received thermal image ( S251 ).

The temperature data extraction unit 120 sets a temperature reference value that can be determined as a product, extracts a region having a temperature value higher than the set temperature reference value, and estimates the location of the target product.

Then, the temperature data extraction unit 120 creates a grid in the area of the target product (S252).

In other words, the temperature data extraction unit 120 sets a grid in the area of the detected target product. In this case, the set grid may be implemented in the form of a one-dimensional grid or a two-dimensional grid.

When the grid setting is completed in step S2352, the temperature data extraction unit 120 extracts the temperature of the grid area to generate a temperature distribution diagram (S253).

That is, the temperature data extractor 120 acquires the temperature data of the grid area from the image. In addition, the temperature data extraction unit 120 generates a temperature distribution diagram using the obtained temperature value, and calculates an average value of the generated temperature distribution diagram. The calculated average value is transmitted to the determination unit 140 .

When step S250 is completed, the determination unit 140 receives the result of determining whether the target product is defective by inputting the received surface temperature distribution of the target product into the machine learning algorithm ( S260 ).

That is, the determination unit 140 inputs the surface temperature distribution of the extracted target product to the machine learning algorithm. Then, the machine learning algorithm outputs the determination result for the thermal image obtained by using the input surface temperature distribution map of the target product. Accordingly, the machine learning algorithm determines whether the product is defective according to the input surface temperature distribution of the target product, and transmits the determined result to the determination unit 140 . Then, the determination unit 140 outputs the received determination result.

As described above, according to the present invention, a non-destructive inspection is possible because a determination result for a thermal image of a gas injection molded product is automatically provided using a machine learning algorithm, and an increase in the work speed for detecting product defects and a reduction in labor costs can do it

Although the present invention has been described with reference to the embodiment shown in the drawings, which is merely exemplary, those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom. will be. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the following claims.

100: quality inspection device
110: thermal image acquisition unit
120: temperature data extraction unit
130: study department
140: judgment unit

Claims (12)

A quality inspection apparatus for gas injection molding products, comprising:
A thermal image acquisition unit that receives a thermal image generated by shooting an injection-molded target product by injecting gas from a thermal imaging camera;
A temperature data extraction unit that analyzes the acquired thermal image to detect a target product and extracts surface temperature data within the area of the detected target product, and
A determination unit for determining whether the target product is defective by inputting the extracted surface temperature data into a machine learning algorithm that has been trained,
The temperature data extraction unit,
By using the distribution of temperature data included in the thermal image, an area above a preset temperature is detected as the current location of the target product,
generating a one-dimensional or two-dimensional grid in an area corresponding to the detected target product, and then sampling the temperature distribution of the generated grid to generate a surface temperature distribution diagram of the target product;
A quality inspection device for generating an average surface temperature value by using the surface temperature distribution diagram of the target product. According to claim 1,
Quality inspection apparatus further comprising a learning unit for extracting the surface temperature distribution within the region corresponding to the product from the thermal images of the pre-determined normal product and the abnormal product, and inputting the extracted surface temperature distribution into a pre-established machine learning algorithm to learn . delete delete delete 3. The method of claim 2,
The learning unit,
Using the temperature data extracted from the thermal image of the normal product and the temperature data extracted from the thermal image of the abnormal product, a dataset of a normal product and a dataset of a defective product are generated,
A quality inspection device that trains a machine learning algorithm using each generated dataset. In the quality inspection method of a gas injection molding product using a quality inspection device,
Receiving a thermal image generated by shooting an injection-molded product by injecting gas from a thermal imaging camera,
Detecting the product by analyzing the acquired thermal image, extracting surface temperature data within the area of the detected product, and
Including the step of determining whether the product is defective by inputting the surface temperature data of the target product to be determined to the machine learning algorithm that has been trained,
The step of extracting the temperature data,
Using the distribution of temperature data included in the thermal image to detect an area above a preset temperature as the current location of the target product, and generating a one-dimensional or two-dimensional grid within the area corresponding to the detected target product;
generating a surface temperature distribution diagram of the target product by sampling the temperature distribution of the generated grid, and
A quality inspection method that generates an average surface temperature value using the surface temperature distribution diagram of the generated target product. 8. The method of claim 7,
Extracting the surface temperature distribution within the region corresponding to the product from the thermal images of the pre-determined normal product and the abnormal product, and inputting the extracted surface temperature distribution into a pre-established machine learning algorithm for learning. . delete delete delete 9. The method of claim 8,
The step of learning the machine learning algorithm comprises:
Generating a dataset of a normal product and a dataset of a defective product using the temperature data extracted from the thermal image of the normal product and the temperature data extracted from the thermal image of the abnormal product; and
A quality inspection method comprising the step of learning a machine learning algorithm using each generated dataset.

Images