Digital factory intelligent management method and system based on machine vision

Abstract

The invention relates to the technical field of digital factory management, and particularly discloses a digital factory intelligent management method and system based on machine vision, and the method comprises the following steps: S1, recognizing key equipment in a production line, and setting a visual collection point to obtain a video stream; s2, analyzing the video stream to calculate the real-time production takt of the key equipment, and comparing the real-time production takt with a preset theoretical optimal takt range; s3, by calculating the rhythm matching degree, judging that the production line rhythm is mismatched when the rhythm matching degree is continuously lower than a threshold value; and S4, if mismatching occurs, reversely adjusting operation parameters of all upstream equipment of the key equipment, so that the takt of the key equipment returns to the optimal range, and therefore, whole-line collaborative energy saving is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital factory management, in particular to a digital factory intelligent management method and system based on machine vision. BACKGROUND

[0002] In today's manufacturing practice, the phenomenon of energy waste in factories is quite prominent, and one of the deep-seated problems is the implicit waste caused by the mismatch of production rhythm. Specifically, downstream equipment often frequently starts and stops due to waiting for materials or instructions, which not only increases the wear and tear cost of the equipment, but also causes a large amount of invalid consumption of resources such as electric energy. At the same time, in order to ensure the overall production efficiency, the entire production line often has to run at high speed according to the rhythm of the slowest equipment, causing other efficient equipment to be in a non-optimal working state, further exacerbating energy waste.

[0003] The traditional management mode is difficult to accurately monitor the production rhythm difference of each link in real time, and cannot make dynamic adjustment in time. With the increasing demand for industrial automation and intelligentization, an advanced technical means is urgently needed to solve the above problems. Based on this, the present application proposes a digital factory intelligent management method and system based on machine vision, aiming to realize accurate matching and dynamic optimization of production rhythm through high-precision visual perception and intelligent algorithm analysis, thereby effectively reducing energy waste and improving production efficiency. SUMMARY

[0004] The purpose of the present application is to provide a digital factory intelligent management method and system based on machine vision, which solves the following technical problems.

[0005] The purpose of the present application can be achieved by the following technical solutions: A digital factory intelligent management method based on machine vision, comprising the following steps: Step S1: determining a production line in the factory, the production line being composed of a plurality of production equipment; determining a key equipment in each production equipment of the production line, and setting a visual acquisition point on the key equipment, the setting of the visual acquisition point being used to acquire a video stream of the key equipment in real time; Step S2: analyzing the video stream based on an image recognition algorithm, obtaining the number of materials completing a production process per unit time through the key equipment in real time, and obtaining a real-time production rhythm index of the key equipment; and presetting a theoretical optimal rhythm range of the key equipment based on historical data; Step S3: comparing the real-time production rhythm index with the theoretical optimal rhythm range to obtain a rhythm matching degree; if the rhythm matching degree is lower than a preset threshold within a preset time window, it is determined that the production line has a rhythm mismatch; Step S4: If the production line has a beat mismatch, obtain the production flow direction of the production line, and obtain all upstream devices in the reverse direction of the production flow direction from the key device; adjust the running parameters of each upstream device in turn, so that the real-time production beat index of the key device is in the range of the theoretical optimal beat.

[0006] As a further scheme of the present application: the determination process of the key device comprises: Obtain the process file of the production line, determine the importance level of each production process based on the process file, record the production process with the highest importance level as the key process, and record each production device in the production line that performs the key process as the candidate device; Obtain the topological connection relationship between each production device of the production line, for any candidate device, record all production devices except the candidate device as the remaining devices, when the candidate device fails, determine all remaining devices that will inevitably stop running according to the topological connection relationship, record as the associated device; obtain the number of associated devices of each candidate device, record the candidate device with the most associated devices as the key device.

[0007] As a further scheme of the present application: the setting process of the visual collection point comprises: Obtain the two-dimensional plane layout of the area where the key device is located, mark the outline of the key device and the flow path of the material in the two-dimensional plane layout, and record the area set where the outline and the flow path are located in the two-dimensional plane layout as the collection area; Obtain the shooting range of the visual collection point, adjust the shooting range so that the shooting range contains the collection area, and make the center point of the shooting range coincide with the center point of the collection area.

[0008] As a further scheme of the present application: the process of the real-time production beat index of the key device comprises: The video stream is divided into a plurality of image frames based on a preset sampling frequency, and Gaussian filtering processing is performed on each image frame to eliminate noise interference of the image frame; for any image frame, a gray image of the image frame is obtained, denoted as a current gray image, and a gray image of a previous frame of the image frame is obtained, denoted as a previous gray image; based on the inter-frame difference method, the current gray image and the previous gray image are subjected to pixel-level subtraction operation to generate a difference image; the absolute value of the pixel change of each pixel is obtained according to the difference image, a difference threshold is set, the pixels with an absolute value of pixel change greater than the difference threshold are recorded as foreground pixels, and the pixels with an absolute value of pixel change less than or equal to the difference threshold are recorded as static pixels; a region composed of all foreground pixels in the difference image is extracted and recorded as a material region; the center point of the material region in each image frame is obtained and recorded as a material center, and the actual flow path of the material is obtained according to the positions of the material centers in each image frame. According to the flow path of the material of the key equipment, a virtual detection line is marked in the shooting range, the overlap degree of the actual flow path and the virtual detection line is obtained, and if the overlap degree exceeds a preset overlap threshold, it is recorded that the material completes a production process; the total number of materials that complete the production process per unit time through the key equipment is obtained and recorded as N, and the production cycle index Pci=N / t is obtained, wherein t is the length of the unit time.

[0009] As a further scheme of the present application: the process of obtaining the overlap degree of the actual flow path and the virtual detection line comprises: A two-dimensional coordinate system is established in the shooting range, an expression of the actual flow path is determined in the two-dimensional coordinate system and recorded as a first expression, and an expression of the virtual detection line is determined in the two-dimensional coordinate system and recorded as a second expression; all coordinate points contained in the first expression in the two-dimensional coordinate system are obtained to obtain a first point set, and the second expression is obtained in the two-dimensional coordinate system, recorded as a second point set; the intersection of the first point set and the second point set is obtained, the number n1 of coordinate points in the intersection is obtained, and the number n2 of all coordinate points in the second point set is obtained to obtain the overlap degree CI=n1 / n2*100%.

[0010] As a further scheme of the present application: the process of setting the theoretical optimal cycle range comprises: The historical data includes the total energy consumption of the production line when producing various materials, and the real-time production cycle index of each piece of equipment. Based on the historical data, the real-time production cycle index of each key piece of equipment and its corresponding total energy consumption are obtained. A coordinate system is established with the production cycle index as the abscissa and the total energy consumption as the ordinate. The real-time production cycle index of each key piece of equipment and its corresponding total energy consumption are fitted in the coordinate system to generate a cycle energy consumption curve. An energy consumption critical value E is set, and a critical horizontal line y=E is obtained in the coordinate system, where y is the abscissa of the coordinate system. All curve segments below the critical horizontal line are extracted from the cycle energy consumption curve, and the abscissa intervals corresponding to each curve segment are obtained. The union of the abscissa intervals is obtained to obtain the theoretical optimal cycle range.

[0011] As a further aspect of the present invention: the process of obtaining the beat matching degree includes: Let the theoretical optimal beat range be denoted as {[L1, U1], [L2, U2], ..., [L...} m U m ]}, where L m U represents the starting point of the m-th x-coordinate interval. m Let [L] denote the endpoint of the m-th horizontal coordinate interval, where m is the total number of horizontal coordinate intervals; any horizontal coordinate interval within the theoretically optimal beat range is denoted as [L]. i U i ], where i∈[1,m] and i is a positive integer, determine whether the production cycle index Pci falls within [L i U i If the judgment result is yes, then the beat matching degree is directly recorded as 100%; if the judgment result is no, then the minimum normalized distance D of the theoretical optimal beat range is obtained. min The beat matching degree is obtained as M = (1 - D) min )×100%.

[0012] As a further aspect of the present invention: the process of obtaining the minimum normalized distance of the theoretically optimal beat range includes: Obtain the production cycle index of each production equipment on the production line, select the maximum and minimum production cycle index, and obtain the difference between the maximum and minimum production cycle index, denoted as the normalization coefficient S; obtain the horizontal coordinate interval closest to the production cycle index of the key equipment, denoted as the optimal sub-interval, and denote the optimal sub-interval as [L]. i U i Then the minimum normalized distance D is obtained. min =min(|Pci-L i | / S,|Pci-U i|Pci-U i |Pci-U i |Pci-U i |Pci-U i |Pci-U

[0013] A digital factory intelligent management system based on machine vision, comprising: A configuration module: determining a production line in a factory, the production line being composed of a plurality of production devices; determining a key device in each production device of the production line, and setting a vision acquisition point on the key device, the vision acquisition point being used for real-time acquisition of a video stream of the key device; A beat monitoring module: based on an image recognition algorithm, analyzing the video stream, obtaining the number of materials completing a production process through the key device per unit time in real time, and obtaining a real-time production beat index of the key device; and based on historical data, presetting a theoretical optimal beat range of the key device; A beat comparison module: comparing the real-time production beat index with the theoretical optimal beat range to obtain a beat matching degree; if the beat matching degree is lower than a preset threshold within a preset time window, it is determined that the production line has a beat mismatch; An adjustment module: if the production line has a beat mismatch, obtaining a production flow direction of the production line, and obtaining all upstream devices in the reverse direction of the production flow direction from the key device; adjusting the operating parameters of each upstream device in turn, so that the real-time production beat index of the key device is within the theoretical optimal beat range.

[0014] The beneficial effects of the present application are: The present application accurately perceives the dynamic operation efficiency of a production system through non-contact machine vision technology, and realizes an intelligent regulation and control mechanism based thereon. Compared with the traditional method relying on a single sensor such as an electric meter, the method of the present application can effectively identify the system performance and energy waste caused by the beat mismatch of each device of the production line, rather than only focusing on the on-off state of a single device. By dynamically adjusting the operating parameters of the upstream devices, the production rhythm of the entire line is synchronized and optimized based on the key bottleneck process, thereby significantly reducing the invalid energy consumption caused by the states of device idling, waiting or overloading under the premise of ensuring production continuity, realizing the improvement from single-point energy saving to system-level energy efficiency optimization, and the method has the advantages of strong self-adaptability, low implementation cost and easy integration with existing production lines. BRIEF DESCRIPTION OF DRAWINGS

[0015] The present application will be further described below with reference to the accompanying drawings.

[0016] Figure 1It is a structural schematic view of a digital factory intelligent management method and system based on machine vision. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0018] Please refer to Figure 1 The present application is a digital factory intelligent management method based on machine vision, which comprises the following steps: Step S1: determining a production line in a factory, wherein the production line is composed of a plurality of production devices; determining a key device in each production device of the production line, and setting a vision acquisition point on the key device, wherein the vision acquisition point is used for real-time acquisition of a video stream of the key device; As a preferred embodiment of the present application, the production line is an automatic production line for discrete manufacturing, which comprises a plurality of production devices, wherein the production device is a device for performing a production process, the types of the production process include assembly, machining, detection and packaging, and each production device is connected through a conveying device to transfer materials or semi-finished products; As a preferred embodiment of the present application, the determination process of the key device comprises: Obtaining a process file of the production line, determining the importance level of each production process based on the process file, recording the production process with the highest importance level as a key process, and recording each production device performing the key process in the production line as a candidate device; Obtaining a topological connection relationship between each production device of the production line, for any candidate device, recording all the production devices except the candidate device as remaining devices, when the candidate device fails, determining all the remaining devices that will inevitably stop running according to the topological connection relationship, recording the remaining devices as associated devices, obtaining the number of associated devices of each candidate device, and recording the candidate device with the largest number of associated devices as a key device; The obtaining process of the topological connection relationship comprises: Extract the process piping and instrument diagram (P&ID) and factory layout diagram from the digital delivery materials of the factory, automatically analyze the equipment symbols and connecting pipelines in the diagram through image recognition algorithm, and preliminarily construct the logical connection relationship between the equipment rooms; and match and associate with the equipment code table stored in the manufacturing execution system (MES), to establish the basic data including equipment ID, model and spatial coordinates; In order to further verify the accuracy of the topological relationship, radio frequency identification (RFID) technology can also be used to scan and locate the actual production line, that is, passive RFID tags are installed at key positions of each device, and the relative position data between devices is collected along the material flow path through a mobile reader, and the actual collected physical position relationship is fused with the diagram analysis result; Finally, a device topology network stored in the form of adjacency matrix is generated, wherein the node weight contains the equipment energy consumption level, and the edge weight represents the material transmission distance and functional dependence strength between devices; As a preferred embodiment of the present application, the setting process of the visual collection point includes: Obtain a two-dimensional plane layout of the area where the key equipment is located, mark the outline of the key equipment and the flow path of the material in the two-dimensional plane layout, and record the area where the outline and flow path are located in the two-dimensional plane layout as a collection area; Obtain the shooting range of the visual collection point, adjust the shooting range so that the shooting range contains the collection area, and make the center point of the shooting range coincide with the center point of the collection area; Specifically, based on the historical energy consumption data and production cycle data of each device recorded in the factory manufacturing execution system (MES), the device with the highest energy consumption proportion or the slowest average cycle in the whole production line is selected as the key equipment candidate; Combined with the device topology relationship diagram, the influence range of the candidate device on the downstream process when it fails is analyzed, and finally a production device with high energy consumption characteristics and key bottleneck position is determined as the key equipment; The setting of the visual collection point needs to be combined with the actual working condition on site, for example, for assembly process equipment, an industrial camera with more than 2 million pixels is fixed on the equipment 1.5 meters above through a magnetic universal support, and the lens pitch angle is adjusted to make the field of view completely cover the material import and export area; For machining center type equipment, a shockproof camera needs to be installed at the observation window of the protective cover to ensure that the main shaft start-stop state and tool exchange action can be clearly captured; The camera sampling frequency is set to 25 frames per second, and a gigabit Ethernet is established with an industrial computer, and the video stream is transmitted to the analysis terminal in real time through H.265 encoding; Step S2: based on the image recognition algorithm, analyze the video stream to obtain the number of materials that complete the production process through the key equipment per unit time in real time, and obtain the real-time production cycle index of the key equipment; and preset the theoretical optimal cycle range of the key equipment based on historical data; Specifically, the video stream first passes through the preprocessing module, adopts the adaptive histogram equalization algorithm to eliminate the influence of light abrupt change, and enhances the material profile clarity through the Gaussian-Laplacian edge enhancement algorithm; for the occasional shielding situation in the production process, combined with the YOLOv5 target detection and DeepSORT multi-target tracking algorithm, a multi-hypothesis tracking model based on space-time context is established, specifically, when the material is temporarily shielded, its appearance position is predicted according to the motion trajectory, and when it reappears, identity re-identification is carried out through SIFT feature matching to ensure the continuity of counting; for the material adhesion situation, a segmentation strategy combining watershed algorithm and concave point detection is adopted, and when an abnormally large profile is detected, the segmentation program is automatically triggered; the beat index calculation introduces a sliding time window mechanism, the window length is dynamically adjusted according to the material size, 10 seconds for large parts and 30 seconds for small parts, and the weighted moving average method is used in each window to eliminate transient fluctuations; As a preferred embodiment of the present application, the process of real-time production beat index of the key equipment includes: The video stream is divided into a plurality of image frames based on a preset sampling frequency, and each image frame is subjected to Gaussian filtering processing to eliminate noise interference of the image frame; for any image frame, the gray image of the image frame is obtained, denoted as the current gray image, and the gray image of the previous frame of the image frame is obtained, denoted as the previous gray image; based on the inter-frame difference method, the current gray image and the previous gray image are subjected to pixel-level subtraction operation to generate a difference image; the pixel change absolute value of each pixel is obtained according to the difference image, a difference threshold is set, the pixels with pixel change absolute value greater than the difference threshold are recorded as foreground pixels, and the pixels with pixel change absolute value less than or equal to the difference threshold are recorded as static pixels; the region composed of all foreground pixels in the difference image is extracted and recorded as a material region; the center point of the material region in each image frame is obtained and recorded as a material center, and the actual flow path of the material is obtained according to the position of the material center in each image frame; According to the flow path of the material of the key equipment, a virtual detection line is marked in the shooting range, the overlap degree of the actual flow path and the virtual detection line is obtained, if the overlap degree exceeds a preset overlap threshold, the material is recorded as completing a production process; the total number of materials completing the production process per unit time through the key equipment is obtained and recorded as N, and the production beat index Pci=N / t is obtained, wherein t is the length of the unit time; The process of obtaining the overlap degree of the actual flow path and the virtual detection line includes: A two-dimensional coordinate system is established in the shooting range, an expression of the actual flow path is determined in the two-dimensional coordinate system, denoted as a first expression, and an expression of the virtual detection line is determined, denoted as a second expression; all coordinate points contained in the first expression in the two-dimensional coordinate system are obtained, to obtain a first point set, and all coordinate points contained in the second expression in the two-dimensional coordinate system are obtained, denoted as a second point set; an intersection of the first point set and the second point set is obtained, the number n1 of coordinate points in the intersection is obtained, and the number n2 of all coordinate points in the second point set is obtained, to obtain an overlap degree CI = n1 / n2 x 100%; The setting range of the overlap threshold value is [60%, 100%]; As a preferred embodiment of the present application, the setting process of the theoretical optimal beat range comprises: The historical data include total energy consumption of the production line when the production line produces each material, and real-time production beat indexes of each device; according to the historical data, each real-time production beat index of the key device and the corresponding total energy consumption are obtained; a coordinate system is established with the implementation production beat index as the abscissa and the total energy consumption as the ordinate, each real-time production beat index of the key device and the corresponding total energy consumption are fitted in the coordinate system, and a beat energy consumption curve is generated; an energy consumption critical value E is set, a critical horizontal line y = E is obtained in the coordinate system, wherein y is the abscissa of the coordinate system; all curve segments below the critical horizontal line on the beat energy consumption curve are intercepted, and the corresponding abscissa intervals of each curve segment are obtained, the union of each abscissa interval is obtained, and a theoretical optimal beat range is obtained; The setting process of the energy consumption critical value E comprises: The total energy consumption data of the production line for 90 consecutive natural days is collected, the energy consumption peak value and the valley value are recorded in units of hours, the kernel density estimation method is used to fit the energy consumption probability distribution curve; the 75th percentile of the distribution curve is taken as the basic critical value E0, which can ensure that about 25% of the high energy consumption conditions are included in the optimization range; and a production plan completion rate correction factor a is introduced, the value range of a is [0.9, 1.1]; if the daily plan completion rate is lower than 95%, a is taken as 1.05 to increase the critical value, so as to avoid excessive limitation of the beat due to the demand for production; when the completion rate is more than 105%, a is taken as 0.95 to reduce the critical value, so as to strengthen the energy saving effect; and dynamic adjustment is combined with the real-time energy price signal, the electricity price weight coefficient b is set, the value range of b is [0.8, 1.2]; the electricity price weight coefficient is taken as 1.15 in the peak electricity period (8:00-12:00, 18:00-22:00) and 0.85 in the valley electricity period (0:00-6:00), and the calculation formula of the final energy consumption critical value E is E = E0 x a x b x g, wherein g is a device aging correction factor, which increases by 0.5% per month. Step S3: Compare the real-time production cycle index with the theoretical optimal cycle range to obtain the cycle matching degree; if the cycle matching degree is lower than the preset threshold within the preset time window, it is determined that the production line has a cycle mismatch. In a preferred embodiment of the present invention, the process of obtaining the beat matching degree includes: Let the theoretical optimal beat range be denoted as {[L1, U1], [L2, U2], ..., [L...} m U m ]}, where L m U represents the starting point of the m-th x-coordinate interval. m Let [L] denote the endpoint of the m-th horizontal coordinate interval, where m is the total number of horizontal coordinate intervals; any horizontal coordinate interval within the theoretically optimal beat range is denoted as [L]. i U i ], where i∈[1,m] and i is a positive integer, determine whether the production cycle index Pci falls within [L i U i If the judgment result is yes, then the beat matching degree is directly recorded as 100%; if the judgment result is no, then the minimum normalized distance D of the theoretical optimal beat range is obtained. min The beat matching degree is obtained as M = (1 - D) min )×100%; The process of obtaining the minimum normalized distance of the theoretically optimal beat range includes: Obtain the production cycle index of each production equipment on the production line, select the maximum and minimum production cycle index, and obtain the difference between the maximum and minimum production cycle index, denoted as the normalization coefficient S; obtain the horizontal coordinate interval closest to the production cycle index of the key equipment, denoted as the optimal sub-interval, and denote the optimal sub-interval as [L]. i U i Then the minimum normalized distance D is obtained. min =min(|Pci-L i | / S,|Pci-U i | / S), where min(|Pci-L) i | / S,|Pci-U i | / S) represents |Pci-L i | / S and |Pci-U i The minimum value between | and S; In a preferred embodiment of the present invention, the process of setting the time window includes: The time required for the production line to complete all production processes for one material is obtained and denoted as the production cycle T. The time window is set to be greater than or equal to 2T. In a preferred embodiment of the present invention, if the beat matching degree is greater than or equal to a preset threshold within a preset time window, the real-time production beat index of the key equipment continues to be monitored. Step S4: If the production line experiences a cycle time mismatch, obtain the production flow direction of the production line, and obtain all upstream equipment in the reverse direction of the production flow direction starting from the key equipment; adjust the operating parameters of each upstream equipment in sequence so that the real-time production cycle time index of the key equipment is within the theoretical optimal cycle time range. In a preferred embodiment of the present invention, the process of adjusting the operating parameters of each upstream device includes: The operating parameters include the operating frequency and rotation speed of the production equipment, the conveying speed of the conveying device, and the cyclic operation speed of the robotic arm or actuator. S401, Priority is assigned to each upstream device in the reverse order of the production process starting from the critical device, with the upstream device having a higher priority the closer it is to the critical device. S402, set a proportional threshold K, 0 < K < 1, determine the adjustment deviation KM according to the cycle matching degree, and adjust the operating parameters of the upstream equipment with the highest priority according to the adjustment deviation; S403. After completing one adjustment, monitor the real-time production cycle index of the key equipment within the time window again and record it as the new real-time production cycle index. If the new real-time production cycle index still does not fall into the theoretical optimal cycle range, repeat the adjustment process of S402 for the upstream equipment with the second highest priority according to the priority queue order until the real-time production cycle index of the key equipment is within the theoretical optimal cycle range.

[0019] A machine vision-based intelligent management system for digital factories includes: Configuration module: Identify the production line within the factory, which consists of several production devices; identify a key device within each production device of the production line, and set up a visual acquisition point on the key device, wherein the visual acquisition point is used to acquire the video stream of the key device in real time; The cycle time monitoring module analyzes the video stream based on image recognition algorithms to obtain the quantity of materials that complete the production process through the key equipment per unit time in real time, and obtains the real-time production cycle time index of the key equipment; and presets the theoretical optimal cycle time range of the key equipment based on historical data. The cycle time comparison module compares the real-time production cycle time index with the theoretical optimal cycle time range to obtain the cycle time matching degree; if the cycle time matching degree is lower than a preset threshold within a preset time window, it is determined that the production line has a cycle time mismatch. Adjustment module: If the production line experiences a cycle time mismatch, the module obtains the production flow direction of the production line and, starting from the key equipment, obtains all upstream equipment in the reverse direction of the production flow direction; and sequentially adjusts the operating parameters of each upstream equipment so that the real-time production cycle time index of the key equipment is within the theoretically optimal cycle time range.

[0020] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the invention.

Claims

1. A method for intelligent management of digital factories based on machine vision, characterized in that, Includes the following steps: Step S1: Determine the production line within the factory, which consists of several production devices; identify a key device within each production device of the production line, and set up a visual acquisition point on the key device, wherein the visual acquisition point is used to acquire the video stream of the key device in real time; Step S2: Analyze the video stream based on the image recognition algorithm to obtain the quantity of materials that complete the production process through the key equipment per unit time in real time, and obtain the real-time production cycle index of the key equipment; And based on historical data, the theoretical optimal cycle range of the key equipment is preset; Step S3: Compare the real-time production cycle index with the theoretical optimal cycle range to obtain the cycle matching degree; if the cycle matching degree is lower than the preset threshold within the preset time window, it is determined that the production line has a cycle mismatch. Step S4: If the production line experiences a cycle time mismatch, obtain the production flow direction of the production line, and obtain all upstream equipment in the reverse direction of the production flow direction starting from the key equipment; adjust the operating parameters of each upstream equipment in sequence so that the real-time production cycle time index of the key equipment is within the theoretically optimal cycle time range.

2. The intelligent management method for digital factories based on machine vision according to claim 1, characterized in that, In step S1, the process of determining the key equipment includes: Obtain the process documents of the production line, and determine the importance level of each production process manually based on the process documents. Record the production process with the highest importance level as the key process, and record each production equipment in the production line that performs the key process as candidate equipment. Obtain the topological connection relationship between each production equipment in the production line. For any candidate equipment, all production equipment other than the candidate equipment are recorded as the remaining equipment. When the candidate equipment fails, all remaining equipment that will inevitably stop operating are determined according to the topological connection relationship and recorded as associated equipment. Obtain the number of associated equipment for each candidate equipment and record the candidate equipment with the most associated equipment as the critical equipment.

3. The intelligent management method for digital factories based on machine vision according to claim 1, characterized in that, In step S1, the process of setting the visual acquisition points includes: Obtain a two-dimensional plan layout diagram of the area where the key equipment is located, mark the outline of the key equipment and the material flow path in the two-dimensional plan layout diagram, and record the set of areas where the outline and flow path are located in the two-dimensional plan layout diagram as the collection area. Obtain the shooting range of the visual acquisition point, adjust the shooting range so that the shooting range includes the acquisition area, and make the center point of the shooting range coincide with the center point of the acquisition area.

4. The intelligent management method for a digital factory based on machine vision according to claim 3, characterized in that, In step S2, the process of determining the real-time production cycle index of the key equipment includes: The video stream is divided into several image frames based on a preset sampling frequency, and each image frame is subjected to Gaussian filtering to eliminate noise interference. For any image frame, the grayscale image of the image frame is acquired and recorded as the current grayscale image, and the grayscale image of the previous frame of the image frame is acquired and recorded as the previous grayscale image. Based on the inter-frame difference method, the current grayscale image and the previous grayscale image are subtracted at the pixel level to generate a difference image. The absolute value of pixel change of each pixel is obtained according to the difference image, and a difference threshold is set. Pixels with an absolute value of pixel change greater than the difference threshold are recorded as foreground pixels, and pixels with an absolute value of pixel change less than or equal to the difference threshold are recorded as static pixels. The region composed of all foreground pixels is extracted from the difference image and recorded as the material region. The center point of the material region in each image frame is obtained and recorded as the material center. The actual flow path of the material is obtained according to the position of the material center in each image frame. Based on the material flow path of the key equipment, a virtual detection line is marked within the shooting range. The overlap between the actual flow path and the virtual detection line is obtained. If the overlap exceeds a preset overlap threshold, the material is recorded as having completed one production process. The total number of materials that complete a production process through the key equipment per unit time is obtained and denoted as N. Then, the production cycle index Pci = N / t is obtained, where t is the length of a unit time.

5. The intelligent management method for a digital factory based on machine vision according to claim 4, characterized in that, In step S2, the process of obtaining the overlap between the actual flow path and the virtual detection line includes: A two-dimensional coordinate system is established within the shooting range. An expression for the actual flow path is determined within this system, denoted as the first expression. An expression for the virtual detection line is also determined, denoted as the second expression. All coordinate points contained in the first expression within the two-dimensional coordinate system are obtained, forming a first point set. All coordinate points contained in the second expression within the two-dimensional coordinate system are also obtained, denoted as the second point set. The intersection of the first and second point sets is obtained. The number of coordinate points n1 within the intersection is obtained, and the total number of coordinate points n2 within the second point set is obtained. The overlap CI = n1 / n2 × 100% is then calculated.

6. The intelligent management method for a digital factory based on machine vision according to claim 1, characterized in that, In step S2, the process of setting the theoretical optimal beat range includes: The historical data includes the total energy consumption of the production line when producing various materials, and the real-time production cycle index of each piece of equipment. Based on the historical data, the real-time production cycle index of each key piece of equipment and its corresponding total energy consumption are obtained. A coordinate system is established with the production cycle index as the abscissa and the total energy consumption as the ordinate. The real-time production cycle index of each key piece of equipment and its corresponding total energy consumption are fitted in the coordinate system to generate a cycle energy consumption curve. An energy consumption critical value E is set, and a critical horizontal line y=E is obtained in the coordinate system, where y is the abscissa of the coordinate system. All curve segments below the critical horizontal line are extracted from the cycle energy consumption curve, and the abscissa intervals corresponding to each curve segment are obtained. The union of the abscissa intervals is obtained to obtain the theoretical optimal cycle range.

7. The intelligent management method for a digital factory based on machine vision according to claim 1, characterized in that, In step S3, the process of obtaining the beat matching degree includes: Let the theoretical optimal beat range be denoted as {[L1, U1], [L2, U2], ..., [L...} m U m ]}, where L m U represents the starting point of the m-th x-coordinate interval. m Let [L] denote the endpoint of the m-th horizontal coordinate interval, where m is the total number of horizontal coordinate intervals; any horizontal coordinate interval within the theoretically optimal beat range is denoted as [L]. i U i ], where i∈[1,m] and i is a positive integer, determine whether the production cycle index Pci falls within [L i U i If the judgment result is yes, then the beat matching degree is directly recorded as 100%; if the judgment result is no, then the minimum normalized distance D of the theoretical optimal beat range is obtained. min The beat matching degree is obtained as M = (1 - D) min )×100%.

8. The intelligent management method for a digital factory based on machine vision according to claim 1, characterized in that, In step S3, the process of obtaining the minimum normalized distance of the theoretically optimal beat range includes: Obtain the production cycle index of each production equipment on the production line, select the maximum and minimum production cycle index, and obtain the difference between the maximum and minimum production cycle index, denoted as the normalization coefficient S; obtain the horizontal coordinate interval closest to the production cycle index of the key equipment, denoted as the optimal sub-interval, and denote the optimal sub-interval as [L]. i U i Then the minimum normalized distance D is obtained. min =min(|Pci-L i | / S,|Pci-U i | / S), where min(|Pci-L) i | / S,|Pci-U i | / S) represents |Pci-L i | / S and |Pci-U i The minimum value between | and S.

9. A machine vision-based intelligent management system for digital factories, characterized in that, include: Configuration module: Identify the production line within the factory, which consists of several production devices; identify a key device within each production device of the production line, and set up a visual acquisition point on the key device, wherein the visual acquisition point is used to acquire the video stream of the key device in real time; The cycle time monitoring module analyzes the video stream based on image recognition algorithms to obtain the quantity of materials that complete the production process through the key equipment per unit time in real time, and obtains the real-time production cycle time index of the key equipment. And based on historical data, the theoretical optimal cycle range of the key equipment is preset; Beat Comparison Module: Compares the real-time production beat index with the theoretical optimal beat range to obtain the beat matching degree; If the beat matching degree is lower than the preset threshold within the preset time window, it is determined that the production line has a beat mismatch. Adjustment module: If the production line experiences a mismatch in cycle time, the module obtains the production flow direction of the production line and, starting from the key equipment, obtains all upstream equipment in the reverse direction of the production flow direction; and sequentially adjusts the operating parameters of each upstream equipment so that the real-time production cycle time index of the key equipment is within the theoretically optimal cycle time range.

Images