Intelligent identification method for abnormal assembly curve of fastener

Abstract

The invention provides an intelligent identification method for an abnormal assembly curve of a fastener, and the method comprises the steps: S101, carrying out the image preprocessing, and carrying out the preprocessing of a fastener assembly curve training graph; step S102, constructing a neural network model, and performing marking and network training on the fastener assembly curve training graph; step S103, using the trained neural network model to collect and identify a fastener assembly curve graph: step S1031, collecting the fastener assembly curve graph, and preprocessing an image; and step S1032, identifying and judging the preprocessed image, and if it is judged that the assembly is abnormal, sending an early warning notification to related personnel. The problem that an abnormal assembly curve is difficult to effectively recognize is solved, the recognition capability of recognizing abnormal equipment is improved, the assembly efficiency is further improved, and the production cost is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of automobile fastener assembly, and more specifically to an intelligent recognition method for abnormal fastener assembly curves. Background Art

[0002] In the field of automobile manufacturing, fasteners are key elements that connect various vehicle components, and their assembly quality is directly related to the safety, durability and performance of the entire vehicle. At present, although tightening guns generally have the function of collecting angles and torques during the assembly process, and set monitoring windows to perform quality control in the final stage, this "point monitoring" method is stretched when dealing with complex and changeable assembly scenarios. In particular, for those abnormal assembly curves that can indicate potential quality problems, such as gun jump curves and over-yield curves, traditional methods are often difficult to effectively identify, thus burying quality risks. Abnormal assembly curves not only reflect atypical behaviors during tightening, but may also indicate major defects in the assembly quality of fasteners. For example, a gun jump curve may mean that there is an unexpected interruption or slippage during the tightening process, while an over-yield curve may indicate that the fastener has exceeded the elastic limit of its material and entered the plastic deformation area. If these abnormal phenomena are not identified and corrected in a timely manner, they are likely to lead to insufficient or over-tightening of fastener assembly torque, which in turn leads to serious consequences such as loosening and breaking of parts, posing a threat to the safety and reliability of the vehicle.

[0003] Given the significant impact of abnormal assembly curves on fastener assembly quality, it is particularly important to develop a technology that can automatically identify these curves. This can not only effectively avoid potential quality problems, but also improve assembly efficiency and reduce production costs. Summary of the invention

[0004] The present invention provides an intelligent recognition method for abnormal assembly curves of fasteners, aiming to solve the problem that abnormal assembly curves are difficult to effectively recognize, improve the recognition ability of abnormal equipment, further improve assembly efficiency, and reduce production costs.

[0005] To achieve the above object, the present invention provides an intelligent identification method for abnormal assembly curve of fasteners, comprising:

[0006] Step S101, image preprocessing, preprocessing the fastener assembly curve training image;

[0007] Step S102, constructing a neural network model, marking the fastener assembly curve training graph and performing network training;

[0008] Step S103, using the trained neural network model to collect and identify the fastener assembly curve graph, including:

[0009] Step S1031, collecting a fastener assembly curve diagram and preprocessing the image;

[0010] Step S1032, identify and judge the pre-processed image, and if it is judged that there is an abnormality in the assembly, issue a warning notification to relevant personnel.

[0011] In one embodiment, preprocessing includes: cropping the image, unifying pixels, and grayscale processing.

[0012] In one embodiment, building a neural network includes configuring an input layer, a convolution layer, a pooling layer, an activation layer, a fully connected layer, and a final classification layer, and setting hyperparameters.

[0013] In one embodiment, the constructed neural network is a convolutional neural network, which automatically extracts and identifies different types of assembly curve features through deep learning and marks them by labeling.

[0014] In one embodiment, the fastener assembly curve training diagram includes a normal assembly curve training diagram, a jump gun assembly curve training diagram, and an over-yield assembly curve training diagram;

[0015] The fastener assembly curve training diagram includes part fitting points and process torque points;

[0016] Among them, the process torque point is the tightening end position, which is determined according to the torque value of the fastener.

[0017] In one embodiment, in step S102, marking includes marking a normal assembly label, a jump gun assembly label, and a yield assembly label;

[0018] When the slope of the fastener assembly curve from the part fitting point to the process torque point remains basically unchanged, a normal assembly label is affixed;

[0019] When the curve of the fastener assembly curve drops sharply during the process from the part fitting point to the process torque point, the jump gun assembly label is affixed;

[0020] When the slope of the curve decreases significantly during the process from the part fitting point of the fastener assembly curve to the process torque point, the yield assembly label is affixed.

[0021] In one embodiment, in step S103:

[0022] When the fastener assembly curve is judged to be normal assembly, it will be displayed as normal assembly and no warning notification will be issued;

[0023] When the fastener assembly curve is judged to be a jump-gun assembly, the abnormal assembly is displayed and an early warning notification is issued;

[0024] When the fastener assembly curve is judged to be an over-yield assembly, abnormal assembly is displayed and an early warning notification is issued.

[0025] In one embodiment, when the neural network makes an incorrect judgment on the fastener assembly curve graph, the incorrectly judged image is put into the neural network, the neural network is retrained, and the training and iteration of the neural network are continued.

[0026] In one embodiment, after the early warning notification is triggered, an email is automatically sent to relevant personnel.

[0027] In one embodiment, the normal assembly curve training graphs include at least 1000 graphs, the gun-jump assembly curve training graphs include at least 1000 graphs, and the over-yield assembly curve training graphs include at least 1000 graphs.

[0028] The present invention has the following beneficial effects:

[0029] 1. Operational convenience: The present invention uses a convolutional neural network artificial intelligence algorithm to automatically collect and identify the tightening curve of the workshop and determine whether there is any abnormality in the tightening curve.

[0030] 2. High accuracy: This method is based on the overall shape of the torque image rather than the value of a single data point, so it has a better recognition effect than the conventional window method. Moreover, through continuous iteration and learning, the accuracy of the neural network can further improve product quality and reduce the risk of abnormal fastener assembly.

[0031] 3. High efficiency: The trained neural network can automatically identify the tightening curve of the workshop without the need for human judgment, thereby improving production efficiency, reducing errors and delays, and significantly improving corporate benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic flow chart of a method for intelligently identifying abnormal assembly curves of fasteners according to an embodiment of the present invention;

[0033] Figure 2 A schematic diagram of a normal assembly curve of an intelligent identification method for abnormal assembly curve of a fastener according to an embodiment of the present invention;

[0034] Figure 3 A schematic diagram of a gun-jumping assembly curve of an intelligent method for identifying abnormal assembly curves of fasteners according to an embodiment of the present invention;

[0035] Figure 4 A schematic diagram of an over-yield assembly curve of an intelligent identification method for abnormal assembly curves of fasteners according to an embodiment of the present invention;

[0036] Figure 5The figure is a complete flow chart of a method for intelligently identifying abnormal assembly curves of fasteners according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical scheme and advantages of the implementation of this application clearer, the technical scheme in the embodiment of this application will be described in more detail below in conjunction with the drawings in the embodiment of this application. In the drawings, the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions. The described embodiments are part of the embodiments of this application, not all of them. The embodiments described below with reference to the drawings are exemplary and are intended to be used to explain this application, and should not be construed as limitations on this application. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0038] Figure 1 The figure is a flow chart of a method for intelligently identifying abnormal assembly curves of fasteners according to an embodiment of the present invention. The method for intelligently identifying abnormal assembly curves of fasteners comprises:

[0039] Step S101, image preprocessing, preprocessing the fastener assembly curve training image;

[0040] Step S102, constructing a neural network model, marking the fastener assembly curve training graph and performing network training;

[0041] Step S103, using the trained neural network model to collect and identify the fastener assembly curve graph, including:

[0042] Step S1031, collecting a fastener assembly curve diagram and preprocessing the image;

[0043] Step S1032, identify and judge the pre-processed image, and if it is judged that there is an abnormality in the assembly, issue a warning notification to relevant personnel.

[0044] Specifically, preprocessing is to reduce unnecessary interference information in the image and improve the accuracy and effectiveness of neural network training.

[0045] In one embodiment, preprocessing includes: cropping the image, unifying pixels, and grayscale processing.

[0046] Specifically, in order to reduce unnecessary interference information in the image, the original image is cropped by curve main body, unified pixel size and grayscale processing.

[0047] In one embodiment, building a neural network includes configuring an input layer, a convolution layer, a pooling layer, an activation layer, a fully connected layer, and a final classification layer, and setting hyperparameters.

[0048] Specifically, building neural networks, especially convolutional neural networks (CNNs) for image processing, involves a systematic process that includes defining the various layers of the network and configuring the corresponding hyperparameters.

[0049] Furthermore, the input layer is the starting point of the neural network, which is responsible for receiving raw data. For image data, the size of the input layer usually matches the image's dimensions of height, width, and number of color channels. For example, for a 224x224 RGB image, the size of the input layer will be set to 224x224x3. The convolution layer is the core of CNN and is used to extract features from images. In the convolution layer, a series of convolution kernels or filters are defined, which slide over the input image or the feature map of the previous layer, performing convolution operations to generate new feature maps. Each convolution kernel can be regarded as a feature detector that can capture specific patterns in the input data.

[0050] Among them, the hyperparameter configuration includes:

[0051] The number of convolution kernels determines the depth of the output feature map;

[0052] The size of the convolution kernel determines the size of the receptive field;

[0053] The step size determines the distance the convolution kernel slides;

[0054] Padding, used to control the size of the output feature map;

[0055] Pooling layers usually follow convolutional layers to reduce the height and width of feature maps, reducing the amount of computation and the number of parameters while retaining important features.

[0056] The hyperparameter configuration includes: pooling window size and step size.

[0057] The activation layer introduces nonlinearity into the neural network, enabling the network to learn complex patterns. In CNN, ReLU is the most commonly used activation function because it is simple to calculate and can effectively alleviate the gradient vanishing problem.

[0058] After the convolutional layer and the pooling layer, there are usually several fully connected layers to integrate the local features extracted previously into global features, which are ultimately used for classification or regression tasks.

[0059] The hyperparameter configurations include: the number of fully connected layers and the number of neurons in each layer.

[0060] For classification tasks, the last layer of the network is usually a classification layer, which outputs the predicted probability of each class. The classification layer is usually a fully connected layer followed by a softmax activation function to convert the output into a probability distribution.

[0061] In addition to the above-mentioned layer-specific hyperparameters, there are also some global hyperparameters that need to be set, including: learning rate, optimizer, loss function, number of iterations / training cycle, batch size, and regularization method.

[0062] The selection and adjustment of these hyperparameters have a crucial impact on the performance of the network, and usually requires experiments and verification to find the optimal configuration.

[0063] In one embodiment, the constructed neural network is a convolutional neural network, which automatically extracts and identifies different types of assembly curve features through deep learning and marks them by labeling.

[0064] In one embodiment, the fastener assembly curve training diagram includes a normal assembly curve training diagram, a jump gun assembly curve training diagram, and an over-yield assembly curve training diagram;

[0065] The fastener assembly curve training diagram includes part fitting points and process torque points;

[0066] Among them, the process torque point is the tightening end position, which is determined according to the torque value of the fastener.

[0067] Specifically, smaller bolts or nuts may only require a smaller torque value, such as 20 Nm, to achieve sufficient tightening effect, while larger components or connection points that bear greater loads may require a larger torque value, such as 100 Nm, to ensure a secure connection.

[0068] Specifically, the part fit point, also called the fit torque point, is the point at which the fastened joint begins to deform at an approximately constant rate. At this point, the deformation of the connected components and the entire fastened joint begins to become constant. In the assembly curve, this is usually expressed as a torque value, when this value is reached, the connection begins to fit tightly, and further increase in torque results in a constant axial clamping force in the connection. The part fit point is an important reference point for evaluating the quality of the fastened joint and determining the subsequent tightening strategy.

[0069] In one embodiment, in step S102, marking includes marking a normal assembly label, a jump gun assembly label, and a yield assembly label;

[0070] When the slope of the fastener assembly curve from the part fitting point to the process torque point remains basically unchanged, a normal assembly label is affixed;

[0071] When the curve of the fastener assembly curve drops sharply during the process from the part fitting point to the process torque point, the jump gun assembly label is affixed;

[0072] When the slope of the curve decreases significantly during the process from the part fitting point of the fastener assembly curve to the process torque point, the yield assembly label is affixed.

[0073] In one embodiment, in step S103:

[0074] When the fastener assembly curve is judged to be normal assembly, it will be displayed as normal assembly and no warning notification will be issued;

[0075] When the fastener assembly curve is judged to be a jump-gun assembly, the abnormal assembly is displayed and an early warning notification is issued;

[0076] When the fastener assembly curve is judged to be an over-yield assembly, abnormal assembly is displayed and an early warning notification is issued.

[0077] In one embodiment, when the neural network makes an incorrect judgment on the fastener assembly curve graph, the incorrectly judged image is put into the neural network, the neural network is retrained, and the training and iteration of the neural network are continued.

[0078] In one embodiment, Figure 5 As shown, after the early warning notification is triggered, an email is automatically sent to relevant personnel.

[0079] In one embodiment, the normal assembly curve training graphs include at least 1000 graphs, the gun-jump assembly curve training graphs include at least 1000 graphs, and the over-yield assembly curve training graphs include at least 1000 graphs.

[0080] In one embodiment, three types of fastener assembly curve training diagrams are prepared, and the fastener assembly curve training diagrams include normal assembly curve training diagrams, gun-jumping assembly curve training diagrams, and over-yield assembly curve training diagrams, so that the training diagrams include normal assembly curves, over-yield assembly curves, and gun-jumping assembly curves. Exemplarily, 1,000 training diagrams of each curve are selected.

[0081] In order to reduce unnecessary interference information in the image, the fastener assembly curve training image is cropped, the pixel size is unified, and the grayscale is processed.

[0082] In this embodiment, two types of defects that are difficult to identify in the existing assembly curve monitoring window are selected, namely, the over-yielding assembly curve and the gun-jumping assembly curve. The over-yielding assembly curve will cause the bolts to be over-tightened, increasing the risk of bolt breakage. The gun-jumping assembly curve is an abnormal assembly process, which may cause the bolt assembly angle to be insufficient, and ultimately lead to a low assembly clamping force of the bolt, and the bolt becomes loose and fails.

[0083] In one embodiment, 3000 assembly curves may be selected, of which the 1st to 1000th are normal assembly curves, and such curves are labeled as normal assembly curves; the 1001st to 2000th are over-yield assembly curves, and such curves are labeled as yield assembly curves; and the 2001st to 3000th are gun-jumping assembly curves, and such curves are labeled as gun-jumping assembly curves.

[0084] The monitoring window is a reasonable fluctuation range of the process torque point. The monitoring window point is a monitoring area set to ensure that the fastener assembly process meets the preset standards and requirements. In the assembly curve, the monitoring window is usually expressed as a torque and angle interval range, which is determined based on actual assembly data and statistical methods. The setting of the monitoring window helps to identify abnormal conditions and potential problems in the assembly process, such as bolt tilt, repeated tightening, thread damage, etc.

[0085] In the conventional window method recognition mode, when both the angle and the torque fall within the monitoring window, the curve is judged to be qualified. This method is also applicable in the method of this patent.

[0086] Figure 2 This is a normal assembly curve image. The horizontal axis of the curve is the angle, and the vertical axis is the torque. The curve in the figure is a normal assembly curve. When the assembly curve reaches point B, the process torque point, from point A, the part fitting point, the slope of the assembly curve remains basically unchanged, and the torque increases linearly with the increase of the angle. This type of curve is a normal assembly curve and is labeled as a normal assembly.

[0087] In the conventional window recognition mode, if the angle and torque of the end both fall within the monitoring window range, it is also judged as qualified. If the angle or torque of the end exceeds the monitoring window range, it is judged as unqualified. Since "jumping gun" may occur at any stage of the tightening process, it is impossible to make a correct judgment using the conventional window method. This application can identify any stage of the tightening process to correct the erroneous judgment of the conventional window recognition mode.

[0088] Figure 3 This is the jump gun assembly curve. The horizontal axis of the curve is the angle, and the vertical axis is the torque. The curve in the figure is the jump gun assembly curve. When the assembly curve reaches the process torque point B from point A where the parts fit, the curve begins to show a rapid drop in torque from point C, indicating that the tightening gun at this position has disengaged from the bolt head and is idling until point D, when the tightening gun is put on the bolt head again, and the torque rises rapidly again until the set final torque point B. The drop in torque from point C to point D is the process of the assembly curve jumping the gun.

[0089] Furthermore, in some high-torque bolt position tightening processes that have high safety requirements, the material yield point will be slightly exceeded in the final stage of tightening to obtain the best clamping force. However, in this working condition, once over-yielding occurs, it will cause plastic deformation damage to the material itself.

[0090] Figure 4 This is an over-yield assembly curve. The horizontal axis of the curve is the angle, and the vertical axis is the torque. The curve in the figure is an over-yield assembly curve. When the assembly curve moves from point A, the part fitting point, to point B, the process torque point, the slope of the assembly curve decreases significantly at point C, that is, the slope of the second curve CB is significantly lower than the slope of the first curve AC. This indicates that the bolt has material yield at point C. After point C, as the angle increases, the torque increases slowly. After reaching the black bold section, the bolt has over-yield fracture failure. This type of curve is an over-yield assembly curve.

[0091] Therefore, when the tightening end position B exceeds point C, the bolt enters the yield area, which will cause the risk of fracture. In the conventional window method recognition mode, only the end point B is judged by the window. Even if the torque over-yield occurs, the single data point of the end torque B can also fall within the window range. Therefore, the window method is judged as "qualified". It can be concluded that the recognition accuracy of the window method recognition mode is not high.

[0092] After preparing the image, you can start building the neural network model, configure the relevant input layer, convolution layer, pooling layer, activation layer, fully connected layer and the final classification layer; then set the relevant hyperparameters. After the model configuration and debugging, you can perform network training and continuously optimize the model's hyperparameters based on the training results to improve prediction accuracy.

[0093] Furthermore, the neural network constructed by the model is a convolutional neural network. By giving the convolutional neural network model different types of assembly curves for learning and labeling them with type labels, the neural network generated after training has the ability to judge and identify curve types.

[0094] The acquisition and recognition of fastener assembly curve images include the following steps: data acquisition of fastener assembly images; preprocessing of the acquired images; and image recognition after preprocessing.

[0095] The data of fastener assembly in the workshop is collected and converted into fastener assembly curve images. Then the main body of the original image is cropped, pixel size is unified, and grayscale processing is performed.

[0096] The pre-processed image is identified by the trained neural network to determine whether the curve is abnormal and give the result. When the curve is abnormal, such as a jump gun or over-yield assembly curve, the system will issue an alarm and automatically send an email to notify the person in charge.

[0097] When the neural network makes an error in judging the fastener assembly curve image, for example, the assembly curve is actually a jump gun curve, but the system judges it as a normal curve. At this time, the incorrectly judged curve can be placed in the neural network training set and the neural network training can be re-performed. After multiple neural network training and iterations, the prediction accuracy will be further improved.

[0098] The present invention has the following beneficial effects:

[0099] 1. Operational convenience: The present invention uses a convolutional neural network artificial intelligence algorithm to automatically collect and identify the tightening curve of the workshop and determine whether there is any abnormality in the tightening curve.

[0100] 2. High accuracy: This method is based on the overall shape of the torque image rather than the value of a single data point, so it has a better recognition effect than the conventional window method. Moreover, through continuous iteration and learning, the accuracy of the neural network can further improve product quality and reduce the risk of abnormal fastener assembly.

[0101] 3. High efficiency: The trained neural network can automatically identify the tightening curve of the workshop without the need for human judgment, thereby improving production efficiency, reducing errors and delays, and significantly improving corporate benefits.

[0102] In the description of the present application, it should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. For ease of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship. The technology, methods and equipment known to ordinary technicians in the relevant field may not be discussed in detail, but in appropriate cases, the technology, methods and equipment should be regarded as part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as merely exemplary, rather than as a limitation. Therefore, other examples of the exemplary embodiments may have different values. It should be noted that similar numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0103] It should be noted that, in the present application, the terms "comprises", "includes" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. It should also be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0104] The above embodiments are provided for persons familiar with the art to implement or use the present application. Personnel familiar with the art can make various modifications or changes to the above embodiments without departing from the application concept of the present application. Therefore, the protection scope of the present application is not limited to the above embodiments, but should be the maximum scope of the innovative features mentioned in the claims.

Claims

1. An intelligent identification method for abnormal assembly curve of fasteners, characterized in that: The intelligent identification method of abnormal fastener assembly curve comprises: Step S101, image preprocessing, preprocessing the fastener assembly curve training image; Step S102, constructing a neural network model, marking the fastener assembly curve training graph and performing network training; Step S103, using the trained neural network model to collect and identify the fastener assembly curve graph, including: Step S1031, collecting a fastener assembly curve diagram and preprocessing the image; Step S1032, identify and judge the pre-processed image, and if it is judged that there is an abnormality in the assembly, issue a warning notification to relevant personnel.

2. The intelligent identification method of abnormal fastener assembly curve according to claim 1 is characterized in that: Preprocessing includes: Crop images, unify pixels, and process grayscale.

3. The intelligent identification method of abnormal fastener assembly curve according to claim 1 is characterized in that: Building a neural network includes configuring the input layer, convolution layer, pooling layer, activation layer, fully connected layer, and final classification layer, and setting hyperparameters.

4. The intelligent identification method of abnormal fastener assembly curve according to claim 3 is characterized in that: The constructed neural network is a convolutional neural network, which automatically extracts and identifies the features of different types of assembly curves through deep learning and marks them by labeling.

5. The intelligent identification method of abnormal fastener assembly curve according to claim 4 is characterized in that: The fastener assembly curve training diagram includes the normal assembly curve training diagram, the jump gun assembly curve training diagram and the over-yield assembly curve training diagram; The fastener assembly curve training diagram includes part fitting points and process torque points; Among them, the process torque point is the tightening end position, which is determined according to the torque value of the fastener.

6. The intelligent identification method of abnormal fastener assembly curve according to claim 5 is characterized in that: In step S102, marking includes marking a normal assembly label, a jump gun assembly label, and a yield assembly label; When the slope of the fastener assembly curve from the part fitting point to the process torque point remains basically unchanged, a normal assembly label is affixed; When the curve of the fastener assembly curve drops sharply during the process from the part fitting point to the process torque point, the jump gun assembly label is affixed; When the slope of the curve decreases significantly during the process from the part fitting point of the fastener assembly curve to the process torque point, the yield assembly label is affixed.

7. The intelligent identification method of abnormal fastener assembly curve according to claim 6 is characterized in that: In step S103: when it is determined that the fastener assembly curve diagram is normal assembly, normal assembly is displayed and no warning notification is issued; When the fastener assembly curve is judged to be a jump-gun assembly, the abnormal assembly is displayed and an early warning notification is issued; When the fastener assembly curve is judged to be an over-yield assembly, abnormal assembly is displayed and an early warning notification is issued.

8. The intelligent identification method of abnormal fastener assembly curve according to claim 1 is characterized in that: When the neural network makes an incorrect judgment on the fastener assembly curve diagram, the incorrectly judged image is put into the neural network, and the neural network is retrained, and the training and iteration of the neural network are continued.

9. The intelligent identification method of abnormal fastener assembly curve according to claim 7, characterized in that: After the early warning notification is triggered, an email is automatically sent to relevant personnel.

10. The intelligent identification method of abnormal fastener assembly curve according to claim 5, characterized in that: The normal assembly curve training diagrams include at least 1,000 diagrams, the gun jumping assembly curve training diagrams include at least 1,000 diagrams, and the over-yield assembly curve training diagrams include at least 1,000 diagrams.

Images