TWI316652B - - Google Patents

Description

1316652

[Technical Field of the Invention] The present invention is a visual guidance system applied to an NC punching machine, which uses an X γ working platform on which a single CCD (Charge Coupled Device) is mounted. The soft and physical integration of the machine vision work platform system, combined with the optimal punching path planning, into the drilling of the flexible circuit board, because the flexible circuit board material is flexible, it is improved by visual aiding guidance correction. The accuracy and speed requirements, the workflow is firstly mechanically positioned by the tension control system to the fixed point of the image range, and then the image is captured by the CCD for visual feedback to the ρ^ controller for positioning and fine adjustment compensation. Then use ccd as the base-quasi-hole position detection, and then use GA to do the best path drilling of the remaining hole positions to complete the high-precision positioning drilling requirements. [Prior Art] At present, in the flexible printed circuit board manufacturing process, the shrinkage of the substrate material is generally difficult to control, and the dimensional tolerance of the single product of the flexible circuit board is relatively reduced; especially in the high-tech industry LCD related product combination method brings Considerable impact, and equipment manufacturers are limited by the limitations of mold forming technology, which results in the inaccuracy of positioning of flexible circuit boards (such as the effect of etching holes, mold forming holes), which often fails to improve product quality. Therefore, in order to improve market competitiveness, reduce the cost of manufacturers' costs, and start to develop a set of visual aided positioning equipment and apply this technology to other soft board production-related equipment, the soft board industry can be upgraded quickly. Manufacturers can replace foreign products of the same grade with lower prices, and further expand the application level of machine and Taiwan manufacturers in the field of electromechanicals to localize technology and upgrade the country.

I *·—^- age willow 4 q i;|conversion page 1316652

M, m?A (\9Q V. Invention description (2) The competitive advantage of the inner soft board manufacturer. In the face of the trend of stricter quality requirements in the industry, a small and diversified market is rampant, 丨, 里且夕扬If I ask for the city's reforms in response to China's accession to the World Trade Organization, the promotion of the soft board equipment manufacturing industry to the automation direction: Text 1, has been an urgent task. In the past, the traditional positioning control equipment to do positioning teaching Record the pulse wave application of each moving position. This - the positioning method has been verified by actual verification, and it is easy to: after the metal fatigue, the wear of the mechanism' causes the mechanical error to increase, so that the machine calibration work becomes a regular work, so It does not reduce the efficiency of work and increase the cost, and it will delay the customer's delivery period, especially the variability difference, which will cause a very loud noise to the business opportunity. β [Invention content] ψ In order to reduce equipment construction costs and master relevant technologies, research and development A visual aided positioning device, and the technology is applied to related equipment, the industry can be quickly upgraded, and can replace foreign countries with lower prices and services; The research plan is to use Taiwan's excellent manufacturing capabilities and talents to optimize the equipment of a flexible circuit board to automatically guide the punching system. The main purpose of the equipment creation is: 1. Designing an excellent image processing positioning control program. Use a tension-free sensing element control system to save costs. 3. Use the combination of numerical simulation soft Zhao and GA to simulate the tension control response, which not only saves the time for manually correcting the adjustment parameters, but also = know the response effect, Avoid direct scripting errors and can be used during the measurement phase.

Page 6 1316652 Case No. 94124029

JE_R 曰 Repair - V. Invention Description (3) Modify the data at any time to obtain system parameters. Through the implementation of the above-mentioned technology autonomy and the relevance of localization and root-to-root technology, we will upgrade the domestic industrial technology level, meet the high-precision and high-speed technical needs of the electronic production and tooling industry, and achieve good sales performance. At present, this creation is to use a smashing operation platform, which is equipped with CCD (Charge-Coupled Device), software and hardware integration of machine vision work platform system, and drilling with flexible circuit board with the most diameter planning. operation. Since the soft > quality has a telescopic property, it is required to correct the correction, the degree and the speed by visual aid. The operation flow is firstly positioned by the tension control system to the fixed point of the image range, and then the Ccd is taken back to the PC-Based controller for positioning fine adjustment compensation, and the drilling CCD is used as the reference hole position detection, and then used Do the remaining hole diameter drilling to complete the high precision positioning drilling requirements. By means of the closed loop system control, the usual positioning correction work due to mechanical errors can be eliminated. 【 [Embodiment] 丨f system equipment is mainly used in soft printed circuit boards to achieve high precision, high speed, high 0 phase · Γ λ quality control, (1) mechanical vision assisted correction, (-) Two-parameter adjustment and shortest path 夂) Secondly, the system should be an integrated creation; among them, the key technology industry automation is to create a more single set to improve the precision of the circuit board. Mechanical rough image for visual 'reuse of the best way to carry out the punching, the creation of the needle and (3) three main 1316652 ^ _ case number 94124029 five, invention description (4) (a) Mechanical vision assisted correction

B Revision month Φ.靡: (4) The page is to shorten the development time of the program. 'This creation uses the general image capture card on the market' to match the attached image processing function, and then use the C++ programming language to write the application' via the machine. Positioning to the vicinity of the drilling positioning point, and then taking the eigenvalue of the image 'Temp 1 ate' through the CCD ^, using the Pattern Matching image processing technology, and comparing the image with the captured image, The reference coordinate 'beyond the coordinate conversion, the error correction is performed via the servo system, and the punching operation is performed immediately after the positioning is completed. (Refer to Figure 1 for the system software flow chart, Figure 2 for the visual architecture block diagram, and Figure 3 for the visual system flow chart). The above-mentioned visual servo visual servo refers to the provision of visual feedback information through the camera and the control of closed loop to form a visual servo system. It is necessary to integrate research results in various fields, including image processing and kinematics (kine-matic). ), dynamics and control theory. As far as the camera placement method of the visual feeding system is concerned, it can be divided into: (1) the camera is fixed on the mobile platform (6乂6-111_11311(1) and (2) the camera is fixed on the working platform. How to obtain the positioning coordinate There are three types: (1). Directly transfer the position coordinate _, this method is suitable for a small number of different diameter hole drilling. However, if the quantity is large, Gu Yi generates errors due to manual input, and will cause subsequent error checking to be laborious. (2). Read the drilling file or film file of the circuit design software (such as Pr〇teh pCAD 'OrcadPCB, etc.), analyze its file storage format' and then convert its coordinates and bit size into a format that can be directly processed. (3). It is the teaching input 'which is also the most commonly used method in the industry.

生一月曰1316652 --^S_94124029 V. Description of the invention (5) '匕 is the use of external force to move the axis of motion to record the optical tape feedback coordinates or use the feeding service as the positioning coordinates, or automatically move the CCD lens The position 'automatically finds the coordinates of the drilling position in the visible range of the CCD and records it as follows: 1 · Manual input method 1 is mainly to manually input the positioning coordinate information for the gene to perform the algorithm to make the shortest path planning. In addition, the storage coordinate array is added to the computer file, and the computer can be used to edit the output. The program steps are as follows: jl), first create X, γ each 3 〇 阵列 point array, and will be initialized to zero. W2), set the X, γ coordinate input value limit, and alert the out of range to display the error 'and zero the point coordinate. (3) After recording the input coordinates, replace the corresponding array value and output it to the preview coordinate graph and increment the input point by 1. (4) Repeat with the While-i00p until all coordinate points are entered. (5), slim down the X and Y arrays according to the number of points entered and store them in the file. 2. Read the iron hole file of the circuit design software ___ according to the hole coordinate file provided by the upstream waste merchant, but the soft circuit board will suffer stress and vibration during the process, g = the soft plate is deformed, and the original The designed hole coordinates will also be 3. Automatically teach design - the coordinate design of this coordinate is the same technology that uses the core of image recognition. First, the χ γ axis zeroing operation is performed, and the input search area is converted into the number of loop executions by the image-to-machine proportional carry method. After recording the number of categories, create an array of category names and build options. Input start =

131^2 Case No. 9 V. Invention Description (6) Absolute coordinates of the position, and save in advance. The coordinate conversion of the absolute origin will be the absolute coordinate at the moment. The machine moves to the back row to search for the hole position. If the board data is judged, select the classification hole name to get the recorded points and data class output to the file, so as to calculate the number of subsequent movements as the xinm of the search area. The visual movement distance is • 'and unconditional carry, ensuring the rest. The number of Y-axis movements is similar. Most of them rely on manual visual inspection and manual sorting. If you can develop optical texts via XY automatic shift, identify the hole position and automatically teach the search range in the future. The pre-processing workflow of the image is very wide. From the beginning of the light source to the CCW storm, the 'cpu processor, the system operation of the human-machine operating mediator, etc., all involve the fact that each link is crucial. As the position is reset to the zero start position, the search for possible bits. And write down the hole. Finally, the Xminl^ of the domain that has been sorted by the path plan is also verified by the search area of the image pair. The current stage teaches or semi-automatically identifies the OCR symbol, each of which is divided. (In the figure, the digital video camera, the video camera, the motion card, the slave hardware, the soft, but if the time is converted into a pair, because the mechanical movement moves to the first hole position, the sample is recorded. The number of points and the recorded XY coordinates are divided into visual shifting and mechanical ratio conversion, and the input position of the teaching and searching program can be performed by the manual selection processing function, and the class is recorded, and the 11th figure shows the processing of the included card. In terms of the overall mechanical to theoretical consumption, the pre-processing of the image has the greatest impact, but the latest trend is to solve the problem of avoiding too many pre-processing programs by using physical characteristics. Implementing an image processing task There are many basic steps required, and the first step in processing is shadowing.

Page 10 1316652 Ref. 94124029 V. Invention Description (7) Image acquisition (image acquis image. To complete this step, the signal generated needs to be digitized, every 1/30th of a second it generates a line scan camera, In this case, the motion of the object is transmitted through the image, such as CCD. If the signal in the form of a digital image is taken, it needs to be digitized. The type of the sensor is determined by the application method.

Align), that is, to obtain an analogy or digital image of the image that can be detected by the sensor = can be a monochrome or color television image. The imaging sensor also produces an image line at a time. In this type of scanning camera, the output of a two-dimensional camera or other imaging sensor is not obtained by an analog-to-digital converter A/D and the image obtained by the sensor. After the image, the second step is to pre-process the image. The primary purpose of the pre-processing is to improve the image of the defect in order to increase the chances of success in subsequent processing. Typical pre-processing includes enhancing contrast, filtering out noise, and separating areas whose texture is similar to character information. (Figure 12 in the figure shows the basic steps of image processing). The third step is to segment the segmentation. According to the broad definition, segmentation is the separation of its target and background from an input image. Usually, in digital image processing, automatic segmentation is one of the most difficult tasks. • If the original image is not clear. On the one hand, a robust separation method also takes a long process, so that it does not reach a perfect solution. On the other hand, a simple or error-based segmentation algorithm always leads to the final failure. The output of the split step is usually the original pixel data 'which constitutes the boundary of a region or all points within the region. ^ In both cases, it is necessary to convert the data into a form suitable for computer processing.

11th true

1316652 ----- Case No. 94124029 V. Description of invention (8). The first thing to decide is the data area. When we care about whether the outer shape is shown as a boundary or a complete one, the boundary representation is appropriate. If the β of interest is such as a corner and an internal characteristic when bending, the area representation is appropriate, and the texture or skeleton shape is present at the same time. In this application, in the application, the two types of labels need to be based on the shape of the boundary: the domain is the square of the last name of the coordinate β and the device is created, because it is used to calculate the center has been effectively suppressed Some noise products::: The choice of reasonable light source is visible, so in the aspect of image segmentation: no::: The clear choice between the target and the target is simply to convert the original data into a suitable one to describe this. The difference between the ridges and the ridges is to determine a method of Dtin.: data in order to highlight the features of interest. Description (deScri - This is called feature selection (featUre Selecti〇n), which extracts the characteristics of the quantitative information of interest or the features required to open a class of objects and other objects. The final step includes identification and interpretation. [e^〇gniti〇n) is the process of marking an object based on the information provided by its description. Interpretation is the meaning given to the object being identified. For example, to identify a character C, it is required to associate the descriptor of the word 70 with the marker C. The purpose of the explanation is to give a meaning to a set of marked characters. For example, a 5-digit numeric string or a 5-digit numeric string followed by a dash and then a number of 4 or more digits can be interpreted as a zip code. The interaction between the knowledge base and the processing module shown in Figure 12 in the figure. The knowledge of problem images is compiled into the image processing system in the form of a knowledge base. For example pointing out an image of interest

HH ____ — page 12 This can be the resolution of the problem, knowing the meaning of the machine. The number of doubts in each module is two steps. 1316652 5. The information of the invention is located in Sofan Park. All of the possibilities in the % question are covered, and each area guides each processing model. The communication basis between the processing module groups of the processing module *· For example, • must be given to the other components of the system 'also through the correct position of the string but may cause the solution library to be returned to request the current image processing as 94192029 Areas, libraries, and high-level groups of work and knowledge of the main domain are usually used to make the location of the library of knowledge in the only four-module distribution of the sub-services to limit the search for complex, The relational table or satellite imagery library also controls the relationship of the modules connected by a double head. What is the result of this out of a string is that it can recognize the word. This knowledge not only refers to the implementation of the feedback number (the number of which is connected to a number of times to let it do another one. The search for this information) For example, in the material inspection, the change detection application is like a database. In addition to the interrogation of the group, the prior knowledge of the processing pattern is the postal code of the postal code number relative to the address to block the function of each module. For example, the -" bit was not recognized.) Through knowledge processing, it is true that not all image processing applications require interaction. A lot of practical applications are done by the outside one. In fact, in some cases, I also need it. For example, a visual interpretation of a person requires only a pre-processing module. In general, the processing functions are related to image analysis applications. Search for image boundaries: In the complex circle of the above picture, the images provided by the road are not all of these pieces of chess are often reluctant to identify and interpret

Page 13 1316652 Case No. 94124029 曰修· V. Invention Description (10) At the boundary of the reference graph, gray ep), as shown in Fig. 13, ideal state 'in the ideal state, distributed after imaging, should also be presented The position where a step distribution is located, that is, the boundary to. However, the general image capture is such that when the image is processed by a CCD camera, an image, etc., the resulting image > often fails to achieve an ideal step distribution, as shown in Fig. 14. In other words, the resulting image, the situation 'causes, there will be several pixels and lose the correctness, and it is more difficult to judge the truth. The following is the distribution of the order value of a gray gradient value as a gray scale value at the boundary of a step (the image of the gray-scale value distribution diagram of the image boundary), and it can be clearly distinguished that the boundary is just a single pixel point. The system is a digital system. Because the image capture card is sampled, quantized, and formed, the gray-scale value distribution at the boundary is passed, and the slope-like gray-scale value distribution of the image boundary is indicated. Points that are ambiguous at the boundary are likely to be identified as the location of the boundary of the boundary. Relatively speaking, the gray gradient is described as follows: 1. Record the grayscale value of a pixel on the image coordinate as g ( x, y), that is, a pixel whose coordinate value is (x, y), whose grayscale value is g(xy). 2. For a pixel whose coordinate value is (x, y), it is in the X direction. The gray gradient value is:

Gx C^c, = 1 / ligdx + 1, χ yy> and the gray gradient value in the Υ direction is: >») = !/ 2C^C^, y + vy — g(jK, y —1) )

Page 14 1316652

V. INSTRUCTIONS (11) In the case of images, the large value of the grayscale value at the true boundary can be determined by the concept of gray gradient value, using the XY platform to microstep or Searching near the boundary of the automatically fed party and finding the absolute value of the gray gradient value can be used to find the position of the real boundary. The coordinate design process of the center of the image boundary: The main purpose of image processing of this device is to find out the coordinates 'and then calculate the difference value based on the actual image and the CCD center image block' and use the servo motor to correct the position. Design the process for the center of the image boundary. Detailed Process 1. Command the image capture card to capture the image. 2. Using the minimum error threshold method, the image obtained after binarization and binarization, each image has only 〇 and 255, that is, only white and all black, and the boundary characteristics are more obvious. 3. For the binarized image, search for the coordinate value of the boundary. Next: _( 1 ) Search from the origin of the image coordinates. If it is horizontal and right, search for each column down. If the pixel is vertically corrected, then the top-down, column-by-column (r〇w) grayscale value from the left pixel. (2) In the search, in each block (column), if the continuous search for 1^ grayscale values are all 0, then the change in these consecutive N pixels is the most boundary, in the largest possible way. The difference between the center marks of the generated images, as shown in Figure 15, is as follows: The grayscale value of the captured pixels can be such that the edge search step is positive, then the left grayscale value is searched for each pixel to the right. Point, the first one

Page 15 94142029 Λη 1316652 ----- V. Invention Description (12) — Search for the pixel points, for the downside of the energy boundary point, the coordinate value of the record is too small 'may misidentify the noise point If the edge is large, it may be possible to ask for the temperature of the 7th and before the completion of the unfinished, the reference point of the circle j is positive. I am correcting the boundary of f, so the boundary of f obtained at this time is not the same as the column parallel to the coordinate axis of t ( bar). All of the initial boundary points are not necessarily all in the area where the optimal boundary threshold is drawn and the target value is drawn. In order to get a complete image of the image, we are in the process. Adjust the optimal boundary threshold selection switch, which can be used to see if the image taken is correct or not. It can be used according to the image to determine whether the captured image meets the requirements. In this way, if the source image is not blurred, it is necessary to find an image that meets the accuracy. However, there is a precondition for the above-mentioned part of the image boundary to be obtained, and the difference point of each source image is not too large, thereby affecting the accuracy of the area recognition. 5. Obtain the coordinates of the four corners of the boundary and calculate the center value. Generally, before the binarization process is performed, 'if the center coordinates of the target are to be obtained, then after the image is segmented, the objects in the image have been separated. At this time, we The geometric properties of the object can be analyzed. The calculation method of the center point is as follows: (1) The coordinates of the center point of the image object (X〇, y〇) before the binarization process can be used to show the position of the object's drop point. The mathematical formula is where N is the object. The number of points 'X is the horizontal coordinate value of the object's point, and y is the ordinate value of the object's point. The above equation can easily find the center point coordinate (x0, y〇). Replacement page

(4) 2 cable number __ five 'invention description (13) (2) The image center point coordinates (x0, y0) obtained by binarization are mainly obtained by using the above method = taking out the image 'from left to right, Scan each ··: from top to bottom to find the same number of thresholds as the set. As the maximum benefit of the boundary, draw the horizontal and vertical lines to form the target area, and calculate the coordinates of the center point. (x0, y0). (As shown in Figure 16 is the boundary area map after edge finding). (2) Fixed tension control system The feedback control structure with tension sensor can be designed directly for high-precision tension control. But ί::: the quality of the winding, so this control structure is not suitable for high speed! The method used in this creation is mainly through the estimation of the amount of unwinding radius 'calculation of the tension value from the fixed tension value The torque feedback control is used to indirectly achieve the constant tension control. Two of them: Obtain, basically, the constant tension control is to read the rotational displacement angle of the motor by the encoder of the motor ^, and obtain the value of the shaft diameter after estimation. By calculating the tension set value, the motor torque reference command corresponding to the tension can be obtained. The torque control moment estimator 'estimates the winding tension during the winding process, the purpose of the control, so that it is not necessary to directly rely on the tension = measurement = the tension control (see Figure 6 for the tension) System control block

Age 4泠6 U;::

—------ V. Description of the invention (14) Round, Figure 7 is a schematic diagram of the fixed tension control hardware). The case is wound by numerical simulation software, and according to a physical relationship, the complete system is analyzed and the dynamic response of the computer simulation analysis system is established. The cost is adopted, and the closed-down system of the tension-free sensing component is adopted. The feedback signal to achieve the closed back: turn = system and two as the winding system loop tension control to turn control 'and then indirectly to achieve closed, the door Π :::: adjust the speed of the take-up motor, $ - At the same time, the control is controlled, in which the winding circuit is controlled, and the closing circuit of the Jinxi speed constant force controls the change of the rotational speed of the winding motor and the torque radius of the unwinding motor (refer to Fig. 4, the force control system) Architecture diagram). The mathematics of the tension between the winding motor and the unwinding motor in the creation of the case, the mathematical mode between the winding motor and the unwinding motor is pure, due to: the connection of the belt, between the tapes Whether the winding motor or the unwinding motor can be used as an external interference source (the friction between the tape and the axle), if the length between the two axes is constant, and the two axes are In the case where the tape tension value I is evenly distributed, the tape tension value may be F = Edx, where Ε is the elastic coefficient of the tape, and dx is the minute change length of the tape after stretching. Combined with the above fixed speed control system and constant tension control system, the overall control system architecture circle can be drawn, and then the winding motor can achieve the fixed speed control under the tension limit. The unwinding motor can be under the speed limit. Complete the control of the constant tension, and finally reach the control system of the fixed speed and the constant tension

Page 18 1316652

941 瓣---~土月 a 俨τ V. Description of invention (15) - The 8th circle of Guang is shown as the overall control system square circle type (1) is the measured tension feedback value '(2) is estimated The tension feedback value). (b) = A optimizes the PI parameter adjustment and the shortest path planning PI, :J is a special case of the hysteresis_advance controller has been successfully used: in the multi-industrial control system, including position servo system. Since the position servo system usually has high requirements for control performance, it is more difficult to artificially complete. This creation has been developed in the past two decades - the stochastic optimization algorithm of the evolutionary mechanism of biological imitation, called the genetic algorithm (GA), which is characterized by the structural information that does not require problems, and is not caused by local Optimal and lose the possibility of finding the global optimum, and the algorithm itself should be simple (the basic flow chart of the genetic algorithm is not shown in Figure 9), making it an efficient parameter adjustment for the controller. algorithm. (And the optimal parameter search procedure for the ?1 controller is shown in Figure 1). The parameter adjustment of the GA controller for the GA can not only avoid the problem of setting the initial value of the parameter of the Ι controller, but also obtain a good dynamic response. This creation uses a numerical simulation software coupled with a computer language program to 'find the optimal PI controller's parameter optimization solution (see Figure 5 for a system architecture diagram of GA for PI parameter adjustment). According to the design of the case, the simulation results and analysis are as follows: 1. The image seeking center coordinates are designed according to each step of the central design process of the image boundary. After the program simulation, the following results can be obtained: 1. Capture the image In terms of the first part of the lighting design, it must be considered that the light source is easy to obtain and the installation is simple. Generally, the positive light source is used, but the experimental knot

Page 19 1316652

V. Description of invention (16) Since the object to be tested is a soft printed circuit board with copper falling, the reflective intensity is too high to be effectively imaged, so this creation is changed to a backlight design to improve *, ', this method is applied in the light source. The contour of the object to be measured can provide a contrast of the shadows' and is easy to install. The color of the light source also affects the experimental results; the red LED is used to prove that it can be compared with the soft printed circuit board to separate the target and have a long life and low price. 你2 After your creation is finished Clear images (such as the sharp image f-finding shown in Figure 17), skewed images (such as the skewed image search shown in Figure 18) and the more blurred images (such as the fuzzy image search shown in Figure 19) Looking at the comparison of the centers, we can see that the center point coordinates are better when the threshold is 97. Second, PI tension control system simulation results This equipment is a flexible printed circuit board tension control system, its main purpose is to be able to achieve constant tension control in the case of tension-free sensors, = actual predictive control system, The numerical simulation software pre-simulation is used, and the parameters of the PI controller are adjusted to achieve the best simulation effect; finally, the computer language is used to generate the c language code, and combined with the image control system to be the system control of the device. (As shown in Figure 20P, the initial simulation data of the tension system) ^ PI intelligent GA parameter adjustment efficiency evaluation: In order to solve the traditional PID controller parameter adjustment insufficiency, the f-response curve is used to obtain the control performance. The indicator, that is, the maximum overrun, the stable interval between $ and the rise time, and then use the GA establishment rules to adjust the number of PI controllers to find the best solution. In this paper, the mother algebra is 15〇, the child

Page 20 1316652 . | (f and March 修正 4th revised replacement page - number 94ΐ24η洲__年月日日. Correction...一~. V. Invention description (17) Number 130 and algebra 1〇 for control simulation Based on, and use the multi-objective weighting rule 'to find the best solution ^ (as shown in Figure 21 is the fixed speed adjustment method response curve comparison [MOGA is multi-objective optimization genetic algorithm in the figure], 22nd The graph shows the response curve comparison of the constant tension adjustment method, and the 23rd circle is the controller parameter adjustment comparison. mov,mof: the maximum overshoot of speed and tension. tsv,tsf: the steady state time of speed and tension. trv,trf : The rise time of speed and tension. On the basis of analyzing the development of PID parameters, it is pointed out that the traditional P1D parameter tuning has the advantages of clear physical meaning, easy to grasp and the shortcomings of its parameter optimization. However, with the development of intelligent control theory, fuzzy control, Gene algorithm and neuron control benefits are valued by the control community. In this paper, by adjusting the PI parameter tuning, it can be better than the traditional P1D parameter tuning method. For the current P1 parameter self-tuning method Commenting, pointing out the characteristics of P1 parameter optimization. Revealing the advanced control theory ^ Tuning method to integrate the new PI controller. The effectiveness evaluation of the tension control system simulation: Let the line speed be 1500 (mm / s), according to the formula The initial speed of the take-up motor _ should be 100 (rad/s), and the initial speed of the unwinding motor should be 33.3 (rad/s). The speed of the take-up motor in the simulation phase will decrease as the radius value increases. The speed of the unwinding motor increases as the radius value decreases. The speed change condition ' can be verified by the simulation results. It is better to use the GA adjustment rule than the trial and error adjustment according to the following comparison (as shown in Figure 24). The results of the speed response comparison shown in Figure 25 and the comparison of the tension response shown in Figure 25) ^ M0GA tension control simulation results (as shown in Figure 26):

Page 21 1316652 ^ Case No. 94124029 V. Description of invention (18) Considering the case where the line speed command is 1500 (nini/js), the result state of the simulation from the above figure is maintained at 500 (mm/s), tension The value is in the range of 0.5 (kg). In the tension controller, the disturbance torque and the estimated state estimator can obtain the required isolinear velocity and the equal tension control theory by the estimation method. The author successfully applies the gene transfer algorithm i theory application. The punching of the flexible printed circuit board is controlled by the PC-Based vision-assisted positioning system control card, and with the friendly human-machine interface, punching achieves the purpose of mechanical vision control. For the small 'this article uses the torque estimator, the unwinding is used as the feedback signal of the winding system, in order to indirectly achieve the feasibility of the radius estimation method of the closed loop constant tension control, which can be pre-calculated under the tension condition The unwinding motor is matched with the intelligent PI controller parameter adjustment. The soft and powerful simulation mechanism, according to the set value, and the idea of multi-target-GA, can make the system stabilize. Benefits of the visual guidance system of this creation: Using the creation of this case to successfully refer to the gene transfer date correction replacement page, the force command is 0·5 (kg) 'The line speed value can be maintained after reaching steady state. In terms of value, the torque value of the open horse is known from the above figure, and the state variable is known from the simulation. The results show sexuality. , image positioning and tension control system development. System, image capture card and motion can effectively use the image to locate the actual control output tension. The motor performs torque estimation and achieves closed loop torque control. From the simulation results, the radius of the reel is estimated and the torque value is used. Based on the GA method, the numerical simulation results of the performance index curve are used to retrieve the target simulation results. It is known that the following industrial performance algorithms, image localization and Zhang are indeed calculated.

year

]31όό^2 Case No. 9419 Xian 9〇5, invention description (19) Force control theory applied to the flexible printing circuit board punching system research by the auxiliary positioning system, system, video operation card and motion control card to the surface, can Effectively use the image to locate the punching hole, and the skewer controls the output tension. This creation uses the torque estimator j to open the coil to perform the torque estimation' and use it as the feedback system of the winding system to achieve the closed loop torque. Controlling '胄 indirectly to meet the requirements of closed loop constant tension control. ^ ^ The result shows the feasibility of the radius estimation method, which can estimate i ’ immediately and calculate the matching rotation of the unwinding motor under the constant tension condition. Based on the wisdom of GA? 1 The basic parameter of the controller parameter adjustment is the powerful simulation mechanism of the numerical simulation software, and then the target value is fed back according to the set M = curve, and the idea of multi-objective _GA is combined, and „ . 〇 knows that 'the system can be made It tends to be stable. The following are the specific results of this study. r Lighting knife' uses backlight and red light source 'after measurement' to FP binary::, the clearest image can be obtained. ^ ^ Control system: can be in no Under the sensing component, the fixed speed and constant tension control can be controlled by the soft 鳢 program, and the high speed is also 3Hi and the cost of the sensing component is reduced. 'The combination of the numerical simulation software and the GA, It can save the manual correction and adjustment, avoid the error of directly writing the program, and modify the data at any time during the actual measurement to obtain the system parameters. 4. This equipment is used in the research of the mechanism, using the simple design in the punch mechanism _ a 23rd Page 1316652 窠号.94124029 曰Revision date correction replacement page V. Invention description (20) In terms of design and automatic monolith design, significant results have been obtained in terms of punching speed and material deviation. Adjust the parameters of the PI controller to obtain the parameters of the optimal or suboptimal controller. Using the multi-target-GA tuning and simulation, a very good Li Ying curve can be obtained. In the case, the creation of the case is applied to NC punching. The machine vision guidance system not only has a high degree of ideological creation technology, but also has excellent industrial utilization and economic efficiency in the practical application of the industry.

Page 24 Amendment

year month day

1316652 _ Case No. 94124029 Simple description of the circle Simple description of the diagram] Figure 1: Flow chart of the system software processing of this creation. Figure 2: Block diagram of the visual system architecture of this creation. Figure 3: Flow chart of the visual system of this creation. Figure 4: Architecture diagram of the tension control system of this creation. Figure 5: The system architecture diagram of the GA's PI parameter adjustment. Figure 6: Block diagram of the fixed tension system control. Figure 7: Schematic diagram of the control of the tension for the tension. Figure 8: Block diagram of the overall control system. Figure 9: Basic flow chart of the genetic algorithm. Figure 10: PI controller optimal parameter search steps. Figure 11: Schematic diagram of the automatic teaching search range. Figure 12: Basic steps of image processing. Figure 1 3: Schematic diagram of grayscale value distribution of image boundary in ideal state. Figure 14: Schematic diagram of the grayscale value distribution of the actual image boundary. Figure 15: The central design flow of the image boundary. Figure 16: Diagram of the boundary area after edge finding. Figure 17: Clear and clear image search. Figure 18: Skew image search. Figure 19: Searching for more blurred images. Figure 20: Initial simulation data for the tension system. Figure 21: Comparison of the response curve of the fixed speed adjustment method. Figure 22: Comparison of the response curve of the fixed tension adjustment method. Figure 23: Comparison of controller parameter adjustments.

Figure 25: Comparison of tension response results. Figure 26: MOGA (Multi-Objective Optimized Genetic Algorithm) * y] 33⁄4 system simulation results. [Main component symbol description] CCDCharge-Coupled Device Charge-combined component PC-Based computer-based • GA genetic algorithm LCDLiquidu-Crystal - Display Liquid crystal display Protel computer-aided design software PCAD computer-aided design software OrcadPCB computer-aided design software while- loopWhile loop A/D analog/digital minimum error threshold LEDLight Emiting Doide LEDs • M0GA multi-objective optimization genetic algorithm FPCFlexible Print Circuit Soft printed circuit board mov speed maximal excess mof tension Maximum overrun tsv speed steady state time tsf tension steady state time

Page 26 1316652 Case No. 94124029 Date of the month Amendment

The diagram simply illustrates the rise time of the trv speed. The rise time of trf tension.

Training Phase Object Model Image Model Image Image Image Image Processing

Object ImageModel Building Target Image Model Creation Object Recognition Target Recognition Robot Visual Serving Robotic Arm Visual Servo Fd Tension _ Ru Torque

Vd speed Rr radius Pixel pixel SIMULINK software name C0UT times

Gray Leve 1 Gray scale Pos iti on Position Kp proportional control parameter g K i Integral control parameter K2p Second group proportional control parameter Κ 2 i Second group integral control parameter K3p Third group proportional control parameter K3i Third group integral control parameter Web Dynamic Roll substrate dynamic characteristics

Page 27 1316652 Case No. 94124029 Year Month Day_Revision ^^ Month 4th Revision Replacement Page i Schematic Description

On-line Phase Actual Online Status Page 28

Claims (1)

1316652 Sixth, the scope of application for patents 1. A visual guidance system applied to the NC punching machine is to locate the vicinity of the drilling positioning point by mechanical positioning, and then capture the image through the CCD, extract the characteristic value of the template, and use the graphic ratio. (Pattern worry about "calling" image processing technology, and the captured image is compared, get the reference coordinates and then coordinate conversion 'error correction through the servo system to be positioned 11 completed, then punching operation; The XY working platform is equipped with a single set of CCD (Charge-Coupled Device), which is integrated with the soft and hard 由此 of the machine vision working platform system, and is matched with visual aid guiding correction to improve the accuracy and Speed requirements; the system process includes: column f kill tension control system to do mechanical coarse positioning to the fixed range of the image; CCD capture image for visual feedback to the PC-Based controller for positioning fine adjustment compensation after drilling; Then use the CCD to make the reference hole position detection, and then use GA to do the best path drilling of the remaining hole positions to complete the high-precision positioning drilling. The visual guidance system applied to the NC punching machine according to Item 1 of the scope, wherein the tension control system performs modeling and analysis of various parts of the winding system by numerical simulation software, and according to a physical relationship, the complete system is • The mathematical model is established, and the dynamic response of the system is analyzed by computer simulation. 3. The visual guidance system applied to the NC punching machine according to the scope of the patent application section [where the tension control system is operated by tension-free sensing The closed loop control method of the component and the torque estimator are used to estimate the torque of the unwinding motor, and use this as a feedback signal of the winding system to achieve closed loop torque control. Page 29 1316652 fem 94124029 VI. Applying for a patent specification and indirectly achieving a closed loop constant tension control. The bow 1 system is the visual guide for the punching machine described in the first item, wherein the tension control system adjusts the winding motor by one side:: on the other hand, the torque of the unwinding motor is simultaneously adjusted. In order to achieve = closed-loop synchronous control of tension, where the speed of the take-up motor and the torque of the opening 'change with the radius of the winding. 5. The visual guidance system applied to the Nc punching machine according to the item i of the patent application scope, wherein the mechanical vision assisting correction system uses an image capturing function and an attached image processing function, and then uses c + + The programming language compose application, which is mechanically positioned near the drilling anchor point, then captures the image through the CCD, extracts the feature value of the TemPlate, and uses the Pattern Matching image processing technology and the captured image. The comparison obtains the reference coordinate, and then the coordinate conversion is performed through the coordinate conversion of the coordinate system, and the punching operation is performed immediately after the positioning is completed. 6. The visual guidance system applied to the Nc punching machine according to the first paragraph of the patent application scope. The optimization program design of the GA for PI parameter adjustment and shortest path planning utilizes numerical simulation software, supplemented by computer language. The program is linked to find the optimal solution for the parameters of the best PI controller. 7. According to the scope of claim 1 of the patent application, the visual guidance system applied to the NC punching machine, wherein the image processing of the image boundary center coordinate design process of the captured image is to find the image center coordinate, and then according to The difference between the actual image and the CCD center image coordinate, calculate the difference value and use the servo motor to make the position correction. 8. Apply the visual guide to the nc punching machine according to the first item of the patent application scope. Page 30 1316652 \ffmm -- Case No. 94124029_年月日日_^正L—TM—---———— VI. Applying for the patent range introduction system, the design process of the image boundary center coordinates of the captured image is: 1 · Command the image capture card to capture the image; 2. Use the minimum error threshold method to binarize the captured image; 3. For the binarized image, search for the coordinates of the boundary; 4. Adjust the most The boundary value of the good boundary and the area where the target value is drawn. 5. Obtain the coordinates of the four corners of the boundary and calculate the center value. 9 'Applicable to the visual guide system of the NC punching machine according to item 1 of the patent application scope, wherein the image obtained after binarization in the image processing is 'the gray scale value of each pixel is only 〇 And 255, that is, only two whites and all blacks, so that the boundary characteristics can be made more obvious. 10. The visual guidance system applied to the NC punching machine according to the first application of the patent scope, wherein the captured image The process of searching for the coordinate values of the boundary using the binarized image in the process is as follows: 〇) Searching from the origin of the image coordinates, if it is horizontal correction, then searching from left to right The grayscale value of a pixel, if it is vertical correction 'from top to bottom, search for the grayscale value of each pixel from left to right column by column; * (2) in the search, in each column (column) If the grayscale values of N pixels & are consecutively searched for 〇, then the pixels of the consecutive N pixels that are searched for, the possible boundary points, and their coordinates The value recorded is 'too small' may cause miscellaneous noise Boundary point, ν Shang take excessive removal might leak true boundary points; is to be noted that, before the correction is not completed JL 'boundary of the reference pattern is not necessarily the image coordinate axes Page 31 1316652 a The first month. On the 6th, the replacement page I is amended. VI. The scope of the patent application is parallel. Therefore, all the initial boundary points obtained at this time may not all be in the same column (column). 11. The visual guidance system applied to the NC punching machine according to the first aspect of the patent application scope, wherein the captured image processing can obtain a complete image capturing range for a less clear image, and set a program in the program. Manually adjust the optimal boundary threshold selection switch, which can be used to make the initial calibration to see if the image taken meets the requirements. It can use the human eye to distinguish the image according to the source image. Meet the requirements, so as long as the source image is not blurred, you will be able to find images that match the accuracy. According to the first aspect of the patent application scope, the vision guiding system applied to the NC punching machine, wherein the center point calculation of the center value of the image boundary quadrangular coordinate before the binarization processing is performed in the captured image processing Method: The coordinates of the center point of the object (X〇, y〇) can be used to indicate the position of the falling point of the object. The mathematical formula is weeping, where N is the number of points in the object, and the value of the abscissa of the object is added. The ordinate values of the object points are added up. 'The center point coordinates (χ〇, y〇) can be easily obtained by using the above formula. 13. The visual guidance system applied to the NC punching machine according to the scope of the patent application scope, wherein the center point of the central value of the art image boundary before the binarization processing is calculated in the captured image processing Calculation method: first extract the image, then scan each pixel from top to bottom from left to right to find the same number of threshold points as the maximum change point of the boundary. Use this point to draw the level. And the vertical line area reuse geometry to calculate the center point coordinates (x〇y〇). Phosphorus dare 14. The visual guide applied to the punching machine as described in item i of the patent of Yizhongqing Page 32 1316652 Cable No. 94142029 Λ__a. B Correction Correction Replacement Page Sixth, apply for the patent range introduction system, in which the image is captured first, the light design part is first, because the object to be tested is a soft printed circuit board copper foil, so the backlight design is used to light the light, according to the indexing system, 16. According to the indexing system, when the skewed image is 97, it is obtained by the system, and the system is applied to the NC. In terms of image, the light source is a red LED. Please refer to the visual guide of the NC punching machine mentioned in the first paragraph of the patent, in which the image is captured, and the clear image after the light is compared with the fuzzy image to find the center. Center point coordinates. The visual guide for the NC punching machine mentioned in the first item of the benefit range uses the numerical simulation software to perform the simulation and control the parameters of the controller to achieve the best simulation effect; C language code, and image control are the system control of the creation. Page 33