RU2738532C1 - Method of controlling ceramic articles drying - Google Patents

Abstract

FIELD: technological processes.

SUBSTANCE: invention relates to the field of equipment associated with annealing of natural materials and articles made therefrom, and can be used in production of construction materials, in particular, ceramic bricks. Method of controlling drying of ceramic articles, comprising determining heat carrier rate and temperature, determining location of bogies with ceramic articles in a drying furnace, visual detection of defects in articles and transmission of data on bogies with defective articles to computer for subsequent monitoring of movement of such trolleys along drying furnace, wherein when detecting defects in articles rate and/or temperature of heat carrier in drying furnace is changed, and after detecting defects in articles, visual inspection is carried out, in which visual inspection of items is carried out by means of a wireless video camera, which is moved with a bogie, by means of which an image of bricks surfaces is obtained, which is transmitted to a computer, where processing of said image is performed to recognize defect areas thereon, for which the obtained image is divided into fragments of given dimensions, for each fragment a histogram is plotted for distribution of brightness gradient vectors at all points of the fragment, the obtained histogram is transmitted to the input of a pre-trained neural network, by means of which each fragment is classified, relating it to one of two classes: cracks, depending on brick drying modes, or defects, not depending on brick drying modes, wherein detected defects during further drying are automatically continuous monitoring, and in case of dynamic change of detected defects, from a knowledge base, a neural network generates recommendations to an operator, which, in accordance with recommendations received, performs change of drying (humidity and temperature) modes, until termination of growth of detected defects, after which rate, temperature and humidity of heat carrier are maintained in mode established after termination of appearance or growth of defects.

EFFECT: technical result is shorter time for personnel to stay in drying chamber, minimum waste at drying.

1 cl, 8 dwg, 1 tbl, 2 ex

Description

The invention relates to the field of technology associated with the burning of natural materials and products from them, and can be used in the production of building materials, in particular ceramic bricks.

A known method of drying capillary-porous materials by continuous blowing them with a gaseous heat carrier (AS No. 1698224 A1, class C 04 B 33/30, F 26 B 3/04).

A known method of drying raw bricks and ceramic stones (Saybulatov S.Zh. Production of ceramic bricks. - M .: Stroyizdat. 1989.-p .: 200 ill. - (Improving the skills of workers in the construction and building materials industry) - a two-stage drying method, providing for impulse (rhythmic) drying of the heat carrier during the shrinkage period, when the blowing of the products with the coolant alternates with the cessation of the supply of the coolant, and continuous supply during the final drying of the products.

However, each of these methods allows you to control only the completion or intermediate stages of the drying process as the moisture content in the dried brick decreases. The listed methods do not reveal the formation of cracks, spalling and other mechanical damage in bricks, which does not allow timely removal of defective products (bricks) from the further process and change the drying mode or the mode of preparation of raw components and the actual molding when defects are detected in the products.

The closest to the claimed technical solution is a method for controlling the drying of ceramic products (Chaika A.Yu., ed. Certificate RU 2308649 C1, 2007)

Method - a prototype, providing for determining the speed and temperature of the coolant, determining the location of the carts with ceramic products in the drying oven, visually detecting defects in the products and transferring data about the carts with defective products to a computer for subsequent control of the movement of such carts along the drying oven, and when When detecting defects in products, the speed and / or temperature of the coolant in the drying oven is changed, and visual and / or instrumental control is carried out for the identified defects in the products. In this case, in the event of further termination of the growth of initially identified defects, the speed and temperature of the coolant are maintained in the mode established after the detection of defects. In case of further growth of the initially detected defects in the products, the products with such defects are removed after the controlled carriage leaves the drying oven. The seized defective products are subjected to analysis to identify violations in the processes of the previous preparation of raw components and plastic molding of products and make changes to these processes.

The disadvantages of this technical solution are that visual control of the condition of the products is carried out directly by the personnel in the drying chamber, at least in not the hottest zones. The duration of the drying process significantly exceeds the time of one work shift for the operating personnel. If small defects that are not defects are detected in the products, one shift of personnel cannot follow the further condition of the products and the possible development of defects to the state of defects. In addition, in the process of drying, carts with products are gradually moved through the drying oven, and selective removal of defective products from the carts before they leave the drying oven is practically impossible.

The prototype method for product quality control does not allow to reduce the time spent by personnel in the drying chamber and, moreover, to completely eliminate the need for the presence of workers in the drying chamber. Automatic recognition of defects and minimization of rejects due to changes in the parameters of drying products, depending on the results of monitoring the development of detected defects in ceramic products, are not provided.

The objective of the invention is to eliminate the time spent by the service personnel inside the drying chamber when monitoring the quality of drying ceramic products, automatic detection and recognition of defects and issuing information to the operator for making a decision to change the drying parameters, which will improve the quality of drying and work culture.

The problem is solved due to the fact that in the method of monitoring the drying of ceramic products, which provides for determining the speed and temperature of the coolant, determining the location of carts with ceramic products in a drying oven, visually identifying defects in products and transferring data about carts with defective products to a computer for subsequent control over the movement of such carts along the drying oven, and when defects are detected in the products, the speed and / or temperature of the coolant in the drying oven is changed, and the detected defects in the products are visually inspected, while visual control of the products is carried out by means of a wireless video camera moved with the cart, with with the help of which an image of the surfaces of the brick is obtained, which is transferred to a computer, where the said image is processed, to recognize defective areas on it, for which the resulting image is divided into fragments of specified sizes, for each fragment of the order m the histogram of the distribution of the vectors of the brightness gradients at all points of the fragment, the resulting histogram is fed to the input of a pretrained neural network, with the help of which each fragment is classified, referring it to one of two classes: cracks depending on the brick drying modes, or to defects that do not depend from the brick drying modes, while the detected defects in the process of further drying are automatically continuously monitored, and in the event of a dynamic change in the detected defects, the neural network from its knowledge base generates recommendations to the operator, who, in accordance with the recommendations received by him, changes the drying modes ( humidity and temperature), until the growth of the detected defects stops, after which the speed, temperature and humidity of the coolant are maintained in the mode established after the cessation of the appearance or growth of defects.

Figure 1 schematically shows the process of drying bricks. FIG. 1 introduced the following designations: 1 - drying chamber; 2 - trolleys; 3 and 4 - channels for supplying and removing the coolant, respectively; 5 - recirculation channel; 6 - mixing chamber; 7 - channel for supplying water vapor; 8 - control shut-off valve; 9 and 10 - entrance and exit doors, respectively; 11 and 12 - gates; 13 - air duct for intake of atmospheric air; 14 - air duct for hot air supply; 15 - video camera; 16 - controlled bricks; 17 - computer with a neural network. FIG. 2. shows a sketch of a brick with major defects. FIG. 3 shows a block diagram of the developed algorithm for detecting cracks on the image of the brick surface in automatic mode. FIG. 4 shows an image of the surface of a brick with cracks. FIG. 5, 6 represent images of the brick surface after transformation in the image processing unit.

The histogram of the brightness distribution of the brick image for input into the neural network is shown in Fig. 7. In FIG. 8 shows a diagram of image transfer from a camcorder to a computer. The figures serve to explain the essence of the invention.

The essence of the invention is as follows. The drying chamber 1 (Fig. 1) is loaded with trolleys 2 with raw bricks. Heat carrier is supplied to the drying chamber through the inlet channels 3, which is removed from the chamber through the outlet channel 4. Through the recirculation channel 5, the heat carrier enters the mixing chamber 6, which also receives water vapor through the water vapor supply channel 7. The supply of water vapor to the mixing chamber 6 is regulated by the shut-off valve 8. The trolleys with bricks 2 are loaded into the drying chamber 1 and unloaded through the entrance and exit doors 9 and 10, respectively. The supply of the heat carrier, water vapor is regulated by the dampers 11 and 12. The intake of atmospheric air is carried out through the air duct 13 - the air duct for the intake of atmospheric air coming from the annealing furnace to the mixing chamber 6. The air heated by the heat carrier is fed into the chamber through the duct for supplying hot air 14. Hot air are fed into injection channels located on the roof of the drying oven and then through the holes in the roof (or nozzles) distribute air along the length of the channel.

Bricks belong to capillary-porous bodies in which heat and mass transfer between the drying agent and the product proceeds according to rather complex patterns. One of them is that the entire drying period is divided into three sections: a warm-up period, a period of constant and decreasing drying speed. When drying ceramic products, defects may appear on them, arising both when the technological regimes of the drying process itself are violated, and as a result of a violation of the technology of the previous stages of processing. FIG. 2 schematically shows the surface of a brick and possible defects that appear on the brick during its loading and drying. The classification of these defects and the reasons for their occurrence are shown in Table 1.

Table 1

Defect name Defect description Note Crack Breaking the product without destroying it into parts; a crack is a rupture with an opening width of more than 0.5 mm. Cracks in the front faces of front products are not allowed, and 4 cracks are allowed in ordinary products. Through crack A crack that runs through the entire thickness of the product and is more than half the width of the product. Not allowed. Cut Crack with an opening width of less than 0.5 mm. Separate cuts with a total length of no more than: 40 mm for the front product; for an ordinary product, cuts are permissible. Efflorescence The formation on the surface of the product of stains and streaks from water-soluble salts upon contact with moisture. Not allowed on facing and clinker products.

The appearance of cracks should be considered the beginning of improper drying. Also, defects arise not due to a violation of the drying regime, but in the process of loading, moving and unloading bricks. When drying, to control the modes, a trained neural network is used to timely change the drying modes to reduce the growth of the mentioned cracks and their growth to unacceptable sizes. If the number of cracks or their depth has exceeded the permissible values, then the product with such unacceptable defects is marked as a marriage and is subsequently withdrawn from the batch.

A wireless video camera in a thermobox 15 (Fig. 1) is fixed on a cart 2 with a controlled brick. When loading the trolleys into the drying chamber, a frame is taken at the first position at the beginning of the tunnel. In the process of drying, the cart 2 with the video camera 15 moves along the tunnel with the fixation of video frames at each position. Images of the surface of bricks 16 of the first and subsequent frames are transmitted wirelessly from the video camera 15 to the computer 17 with an image processing module and a neural network. The image of the brick at each subsequent position is compared with the image of the brick at the first position. For comparison and analysis of changes on the surface of 15 bricks filmed with a video camera, the resulting image of the brick surface is broken into fragments. On these fragments, the surface texture is studied by examining a set of curves of different brightness and thickness, with different tilt angles that characterize the direction of maximum brightness change. Cracks in the images look like thin curved structures, so the most effective scheme is to combine the detection of a crack line from several cross-sections in different directions. The existing oriented derivative of the mask (filter) is modified to strengthen the fracture line. The algorithm uses an orientation fragmentation scheme to suppress noise and background. For fragmentation, the algorithm uses a multi-fragment model of integration and segmentation of the brick image. The proposed technique improves crack detection by isolating weaker objects and suppressing interference unrelated to drying.

To describe areas of images with cracks, a local binary template was applied to classify brick textures. The choice of this approach is due to the fact that the dimension of the feature vector is significantly reduced, the speed of response is realized and not critical to the invariance of brick lighting.

FIG. 3 shows a block diagram of the developed algorithm for detecting cracks in the image in automatic mode. The algorithm consists of three stages: preprocessing, defect detection using local binary descriptors, and neural network recognition of cracks and non-cracks. Pre-treatment is an important step in crack detection, designed to suppress the noise component, as well as eliminate defects that are independent of drying modes. The original operator in the form of a sliding mask is calculated by comparing each pixel around the central pixel with the central pixel, taken as the threshold, in a local area of 3 × 3 pixels. A modification is applied, which consists in increasing the radius of the pixels, which are compared with the central pixel. Also, 10 homogeneous neighborhoods have been identified, which carry the greatest information content about the texture features of the image, which also makes it possible to reduce the number of little informative bins. Each of the 10 homogeneous neighborhoods correspond to its own textural features of the image, such as linear cracks, circular cracks, spots. Thus, a histogram for the local area is formed, which is a descriptor for analyzing local areas (features) in the image for the presence of cracks. By concatenating these histograms, a general histogram is formed, taking into account both local and global features of the image. Figure 7 presents a general histogram with defects. To classify the crack image, the elements of the histogram are sent to the inputs of the neural network as components of the vector of characteristics of this fragment. Thus, the number of neurons in the input layer of the network should be equal to the number of histogram columns. Moreover, based on the fact that there are two classes for recognition - a crack (signal) and a defect (not a crack), the neural network must contain two output neurons to perform the classification. The number of neurons in the hidden layer is selected experimentally, based on the specific conditions for registering a brick with a video camera.

To train the neural network, the Backpropagation mode was used, in which the propagation of error signals from the outputs of the neural network to its inputs was used, i.e. in the opposite direction to the forward propagation of signals in normal operation. As a result of the research carried out, this learning mode turned out to be the most effective for modifying the coefficients of connections between neurons of individual layers of the neural network.

In the process of iterative learning, the neural network is tuned in such a way that when presented with histograms, which are formed from a large number of fracture images containing fracture and noise images, it can accurately recognize them.

The implementation of the proposed method in this device is as follows. To obtain (register) an image of the surface of a brick 16 (Fig. 1), the video camera 15 is installed on a trolley above the controlled brick and is automatically positioned to its original position. After that, using a digital video camera 15 with an LED backlight source, an image of the brick surface is recorded (captured). Further, the digital signal from the output of the antenna of the wireless video camera passes from the tunnel to the antenna of the ETHERNET 21 router and further to the computer 17. In the computer, the frame code goes to the block 18 for converting the image into a vector of the brightness gradient, where the image is divided into fragments of a given size, and then from the components the vector of the brightness gradient in block 19 is the formation of histograms of the brightness distribution for each fragment of the image. These histograms are fed to the neural network 20, which is pre-trained using a neural network training unit and a memory unit that stores information about pre-processed images of fragments of cracks and other defects (not cracks) obtained in advance using blocks 18, 19. Neural The network 20 classifies the histogram elements arriving at its inputs, which are components of the brightness gradient vector - characteristics of image fragments. As a result of the classification, the neural network 20 classifies the cracks in the image fragments on the brick surface into one of two classes: crack or other defects (not cracks). It should be clarified that the neural network 20 classifies the object under study into one of the classes in accordance with a certain partition of the N-dimensional space, which is called the input space of the neural network, the dimension of which is the number of components of the brightness gradient vector. Those. the inputs of the neural network 20 receive vector components (characteristics of image fragments), and from its two outputs corresponding to the number of classes, a signal corresponding to one of the specified classes is removed. The signal corresponding to the fracture is then fed to the fracture type identification unit, in which the fracture types are identified.

The drying mode is carried out by the operator according to the recommendations from the knowledge base containing the knowledge model of expert technologists. The process of drying bricks is controlled in the prototype method directly in the tunnel, visually, by the operator. In the claimed method, it is performed automatically using a video camera 15, which continuously transmits an image of the surface of bricks 16 to a computer 17 with an image processing module and a trained neural network located in it. Simultaneously with the image of the surface of the bricks 16, information from sensors located in the chamber about the temperature, humidity and location of the trolley 2 is transmitted to the neural network. The neural network recognizes the defects on the bricks and reveals cracks from them that are formed during the drying process.

The above information on the control results is entered into the processor during its shift or during the transfer of the shift, with a note on the location of the ceramic product with a defect that can lead to rejects during further drying. Subsequent work shifts are monitored by a neural network for the previously noted ceramic product with a defect. Defects on the brick can occur for the following reasons. Violation of the technological mode of brick molding, improper selection and mixing of components, poor-quality composition of components, incorrect laying of bricks on the cart. To control the drying process, only cracks arising during the drying process are used as an informative element, which are compared with the characteristics of the defects specified in the regulatory documents (see table 1). Based on the control results, the neural network generalizes and forms a conclusion about the need to change the drying modes. These factors can be too high or low humidity, or drying temperature. Depending on the type of crack and the dynamics of its change, the neural network generates a signal about the need to change the drying mode (increase or decrease in humidity or drying temperature), and the operator issues the appropriate control actions to the actuators of the drying chamber.

In the proposed method, constant video monitoring of the cart with controlled bricks in the drying oven is carried out. This eliminates the time spent by personnel in the drying oven. Defects are detected on the controlled brick and then the development of these defects during the drying process is automatically monitored. The use of this method makes it possible to exclude the development of a defect to an unacceptable size due to incorrect drying regimes, which is a positive effect of this invention. For this, the parameters in the drying oven are changed (temperature, blowing speed, humidity, time spent by ceramic products in the drying oven). By changing the parameters, a decrease in rejects during drying is achieved.

Examples of specific implementation.

Example 1 . In a continuously operating drying oven, trolleys with raw bricks were sequentially fed, for which the technological operations of preparing raw components and plastic molding of products were carried out in three different shifts (A, B and C) of the service personnel. Each of the carts contained products prepared for drying by one shift of workers (products from one batch). The location of each of the carts in the drying oven was tracked by a computer, while carts with items prepared by different shifts of workers (items from different batches) were marked with different designations, such as different colors on the monitor screen. Carts with products from the same batch were marked with the same color. The operating mode of the drying oven was set for the temperature of the heat carrier (hot air) in the middle part of the oven, equal to 90 ° C; coolant speed 5 m / s, which corresponded to the average residence time of the product in the drying chamber 60 hours.

On two carts with products prepared by shifts A and B, at the initial stage of drying, using a video camera 15 and a neural network of a computer 17, surface cracks of minimum linear dimensions that did not belong to the category of defects were detected by measuring the length and depth of cracks. Data on carts with defective products were entered into a computer for subsequent automatic control over the movement of these carts through the drying oven. The operating mode of the drying oven was softened: the coolant temperature was lowered to 80 ° C; blowing speed - up to 4 m / s, which corresponded to the average residence time of the product in the drying chamber 65 hours.

Upon further monitoring of the marked carts with products using an image processing module and a neural network, it was found that the initially detected cracking of products prepared by shift A did not increase subsequently, and this batch of products was dried without rejects. Analysis of the brick surfaces using a video camera and a neural network showed that the dimensions of surface cracks are within acceptable limits.

Defects in items prepared by shift B have grown to unacceptable levels despite the softening of the drying regime. After the cart with such products left the drying oven, they were rejected and sent for analysis. As a result of the analysis, it was found that during the preparation of raw components, the components were not uniformly mixed enough. This led to uneven drying of different parts of the product, destructive deformations and rejection of the final product. Appropriate changes were made to the technological operation of mixing raw components, which made it possible to further reduce the percentage of scrap from 18% to 5%.

Example 2. After the first three hours of staying in a drying oven (coolant temperature 65 ° C; speed 9 m / s), items with crumbled and crumbled items were detected on trolleys with ceramic items prepared by shift G using a video camera 15 and a neural network of a computer 17 areas of the surface (figure 5). FIG. The area and depth of the detected defects were unacceptable, which meant a defect in this batch of products. Products from other batches did not show visible defects during the drying process. The location of the carts with defective products in the drying oven was monitored by a computer, while, to reduce the residence time of defective products in the oven, the drying mode was changed to a tougher one (heat carrier temperature 85 ° C; speed 10 m / s). Carts with products prepared by different shifts of workers, as in the previous example, were marked in different colors on the monitor screen. Immediately after leaving the drying oven, the marked carts with defective products were seized and sent for analysis. It was found that the components of the raw material were crushed unevenly and contained a large amount of too large particles.

Changes were made to the technological operation of grinding raw components, which made it possible to further reduce the percentage of scrap (up to 3%).

The technical result of the claimed invention is to eliminate the time spent by personnel in the drying chamber, as well as to ensure minimum rejects during drying by changing the operational change in the drying parameters of products depending on the results of automatic control of the development of a defect in the marked ceramic product.

The drying process becomes shorter (by about 15-18 percent), because it eliminates the operations of checking defects by personnel inside the tunnel and the need to roll out trolleys with defective products. The number of defective bricks is also reduced (up to 3-5 percent) due to timely and qualitative changes in drying modes.

Claims (1)

A method for controlling the drying of ceramic products, providing for determining the speed and temperature of the coolant, determining the location of carts with ceramic products in a drying oven, visually detecting defects in products and transferring data on carts with defective products to a computer for subsequent control of the movement of such carts through the drying oven, moreover, when defects are detected in the products, the speed and / or temperature of the coolant in the drying oven is changed, and visual control is carried out for the identified defects in the products, characterized in that the visual control of the products is carried out by means of a wireless video camera moved with a trolley, with which an image of the brick surfaces is obtained, which is transferred to the computer, where the said image is processed to recognize defective areas on it, for which the resulting image is divided into fragments of specified sizes, for each fragment, a histogram of the distribution of vectors brightness gradients at all points of the fragment, the resulting histogram is fed to the input of a pre-trained neural network, with the help of which each fragment is classified, referring it to one of two classes: cracks depending on the brick drying modes, or defects that do not depend on the brick drying modes, at the same time, for the identified defects in the process of further drying, automatic continuous monitoring is carried out, and in the case of a dynamic change in the identified defects, the neural network from its knowledge base generates recommendations to the operator, who, in accordance with the recommendations received by him, changes the drying modes (humidity and temperature), until the growth of the detected defects stops, after which the speed, temperature and humidity of the coolant are maintained in the mode established after the cessation of the appearance or growth of defects.

Images