Visual welding system for automotive wire harness

Abstract

The invention discloses a visual welding system for automotive wire harnesses, which relates to the field of automotive wire harness welding and comprises the following steps: S1, wire stripping: wire stripping is carried out by using a wire stripper; s2, color detection; s3, pre-welding is conducted; s4, heating: heating a lead and a bonding pad of the component by using the preheated solder tip; s5, wire feeding is conducted, specifically, the tin wire is held left and right and sent to the heated welded position; s6, leaving: after the solder wire is melted and slowly flows to two thirds of the area of the whole bonding pad, rapidly removing the solder wire and the solder tip; and S7, solidification: after the solder tip is withdrawn, components and the circuit board are prevented from being shaken, and the welding spots are naturally solidified. For components which are easy to burn, solidification is accelerated through welding and blowing; the method has the advantages that the classifier is created through machine vision, training is carried out to obtain the training model, the training model is applied to color detection of the automobile wire harness, whether color arrangement has errors or not is detected, the accuracy rate is high, and use is convenient.

Description

Technical Field

[0001] The present invention relates to the field of wire harness visual welding, and particularly to a vehicle wire harness visual welding system. Background Art

[0002] Visual inspection of vehicle wire harness welding is one of the important means to ensure the welding quality of automotive wire harnesses. By using a machine vision system to automatically detect the welding points, the production efficiency can be effectively improved, the product quality can be guaranteed, and the labor cost can be reduced.

[0003] Based on this, the present invention designs a vehicle wire harness visual welding system, aiming to detect the accuracy of the color arrangement and welding quality of automotive wire harnesses through machine vision technology. Through artificial intelligence, computer vision, and sensor technology, it simulates the visual ability of humans, analyzes and identifies the images of wire harness colors and welding quality, and performs detection and welding according to preset rules to improve the welding accuracy and achieve an automated processing process. Summary of the Invention

[0004] The object of the present invention is to solve the problems in the prior art, and a vehicle wire harness visual welding system is proposed.

[0005] A vehicle wire harness visual welding system includes the following steps:

[0006] S1: Stripping the wire, using wire strippers to strip the wire, and the stripping length is operated according to the process requirements;

[0007] S2: Color detection, using machine vision to create a classifier and train to obtain a trained model for applying to the color detection of automotive wire harnesses, so as to detect whether there are errors in the color arrangement;

[0008] S3: Pre-welding;

[0009] S4: Heating, using a pre-heated soldering iron tip to heat the component leads and pads;

[0010] S5: Feeding the solder wire, holding the solder wire with both hands and sending it to the heated welding place;

[0011] S6: Withdrawal, after the solder wire melts and slowly flows to two-thirds of the area of the entire pad, quickly withdraw the solder wire and the soldering iron tip;

[0012] S7: Solidification, after the soldering iron tip is withdrawn, avoid shaking the components and the circuit board, and let the solder joints solidify naturally. For components that are easily damaged by heat, blow air during welding to accelerate solidification;

[0013] S8: Inspection, inspect the welding quality to meet the state of moderate solder volume at the solder joints, bright solder joints, and firm welding.

[0014] In the above-mentioned vehicle wiring harness vision welding system, in step S2, the color detection includes image acquisition and preprocessing, color space conversion and recognition, color arrangement detection, and result output and verification steps. According to the requirements and data set of color arrangement recognition, a convolutional neural network structure is designed, and the model is trained using the method of deep learning. The trained convolutional neural network model is applied to the scenario of vehicle wiring harness color arrangement recognition. By collecting the wiring harness image in real time and inputting it into the model, the recognition result of the color arrangement is obtained, and corresponding operations are performed according to the recognition result.

[0015] In the above-mentioned vehicle wiring harness vision welding system, the image acquisition includes using a high-resolution camera or image acquisition device to obtain a clear image of the automotive wiring harness. These images contain various colors of the wiring harness and background information, and the image preprocessing includes median filtering and image enhancement preprocessing operations on the collected images to improve the image quality and reduce noise and interference.

[0016] In the above-mentioned vehicle wiring harness vision welding system, the color space conversion and recognition include:

[0017] Color space conversion: Convert the image from the RGB color space to the HSV color space;

[0018] Color recognition: In the HSV space, color recognition is performed according to different hue values. For each color, a hue range is set in the HSV space, and then the pixel hue values in the image are compared with this range to determine the color of the pixel.

[0019] In the above-mentioned vehicle wiring harness vision welding system, the color arrangement detection includes:

[0020] Color area segmentation: Use threshold segmentation, region filling, and connected segmentation techniques to segment the color areas in the image. These areas correspond to different color parts in the wiring harness.

[0021] Color area positioning: According to the position information of the segmented color areas, determine the position of each color in the wiring harness. This can be achieved by measuring the center position, boundary position, etc. of the color areas;

[0022] Color arrangement order determination: According to the position information of the color areas, determine the arrangement order of the colors, sort or encode the color areas, so as to clearly represent the color arrangement in the wiring harness.

[0023] In the above-mentioned vehicle wiring harness vision welding system, the convolutional neural network contains multiple convolutional layers, pooling layers, fully connected layers, and output layers. In the convolutional layer, different sizes and numbers of convolutional kernels are used to extract color features. In the pooling layer, max pooling or average pooling is used to reduce the dimensionality and computational amount of the data. In the fully connected layer, a weight matrix and a bias vector are used to perform a non-linear combination of the features;

[0024] In the output layer, the Softmax function is used to output the probability distribution of the color arrangement. The convolutional neural network is trained using a dataset. During the training process, a loss function and an optimization algorithm are adopted to minimize the prediction error, and the model performance is optimized by iteratively updating the weights and biases.

[0025] In the above-mentioned vehicle wiring harness vision welding system, the loss function and the optimization algorithm adopt the cross-entropy loss function. The cross-entropy loss function is used to measure the difference between two probability distributions. The formula for the cross-entropy loss function is:

[0026]

[0027] where p is the probability distribution of the true label and q is the predicted probability distribution.

[0028] Compared with the existing technologies, the advantages of the present invention are as follows: A classifier is created using machine vision and trained to obtain a trained model, which is applied to the color detection of automotive wiring harnesses to detect whether there are errors in the color arrangement, with high accuracy and convenient use. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 FIG. is a schematic flow chart of a vehicle wiring harness vision welding system proposed by the present invention.

[0030] Figure 2 FIG. is a flow chart of color detection in a vehicle wiring harness vision welding system proposed by the present invention.

[0031] Figure 3 FIG. is a model diagram of the convolutional neural network structure in a vehicle wiring harness vision welding system proposed by the present invention.

[0032] Figure 4 FIG. is a model diagram of the convolutional neural network structure in a vehicle wiring harness vision welding system proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] Referring to Figures 1-4 , a vehicle wiring harness vision welding system includes the following steps:

[0034] S1: Stripping the wire, using a wire stripper to strip the wire, and the stripping length is operated according to the process requirements;

[0035] S2: Color detection. Use machine vision to create a classifier and train it to obtain a trained model for applying to the color detection of automotive wiring harnesses, so as to detect whether there are errors in the color arrangement.

[0036] S3: Pre-welding;

[0037] S4: Heating. Use a pre-heated soldering iron tip to heat the component leads and pads.

[0038] S5: Wire feeding. Hold the solder wire with both hands and send it to the heated soldering area.

[0039] S6: Withdrawal. After the solder wire melts and slowly flows to two-thirds of the area of the entire pad, quickly withdraw the solder wire and the soldering iron tip.

[0040] S7: Solidification. After withdrawing the soldering iron tip, avoid shaking the components and the circuit board to let the solder joints solidify naturally. For components that are easily damaged by heat, blow air during soldering to accelerate solidification.

[0041] S8: Inspection. Check the soldering quality to meet the state where the solder volume of the solder joint is moderate, the solder joint is bright, and the soldering is firm.

[0042] Among them, the detection of the color arrangement of automotive wiring harnesses is mainly based on image processing and analysis technologies, including steps such as image acquisition and preprocessing, color space conversion and recognition, color arrangement detection, and result output and verification. These steps together constitute a complete detection system, which can achieve accurate detection of the color arrangement of automotive wiring harnesses. The specific steps are as follows:

[0043] I. Image acquisition and preprocessing

[0044] Image acquisition: Use a high-resolution camera or image acquisition device to obtain a clear image of the automotive wiring harness. These images usually contain various colors of the wiring harness and possible background information.

[0045] Image preprocessing: Perform preprocessing operations such as median filtering and image enhancement on the acquired images to improve the image quality, reduce noise and interference. These preprocessing steps help to improve the clarity and contrast of the images, making subsequent color recognition more accurate.

[0046] II. Color space conversion and recognition

[0047] Color space conversion: Convert the image from the RGB color space to the HSV (hue, saturation, value) color space. In the HSV space, the hue component can more intuitively reflect the color changes, which helps with color recognition.

[0048] Color recognition: In the HSV color space, color recognition is performed based on different hue values. For each color, a hue range can be set in the HSV color space, and then the hue values of the pixels in the image are compared with this range to determine the color of the pixels.

[0049] III. Color arrangement detection

[0050] Color region segmentation: Use image processing techniques (such as threshold segmentation, region filling, connected component segmentation, etc.) to segment the color regions in the image. These regions correspond to different color parts in the wire harness.

[0051] Color region localization: Determine the position of each color in the wire harness according to the position information of the segmented color regions. This can be achieved by measuring the central position, boundary position, etc. of the color regions.

[0052] Color arrangement order determination: Determine the arrangement order of the colors according to the position information of the color regions. This usually involves sorting or coding the color regions so that the arrangement of the colors in the wire harness can be clearly represented.

[0053] Steps of deep learning in the recognition of color arrangements in automotive wire harnesses:

[0054] 1. Dataset preparation: Collect a dataset of automotive wire harness images containing different color arrangements, and preprocess the images to ensure the consistency of the input data.

[0055] 2. Model design: Design a convolutional neural network structure according to the requirements of color arrangement recognition and the characteristics of the dataset. Among them, the network includes multiple convolutional layers, pooling layers, fully connected layers, and output layers.

[0056] In the convolutional layer, convolutional kernels of different sizes and numbers are used to extract color features.

[0057] In the pooling layer, max pooling or average pooling is used to reduce the dimensionality and computational complexity of the data.

[0058] In the fully connected layer, weight matrices and bias vectors are used to perform non-linear combinations of the features.

[0059] In the output layer, the softmax function is used to output the probability distribution of the color arrangements.

[0060] The softmax function is used for multi-classification problems, which converts a vector composed of K real numbers into a probability distribution with a range of (0, 1) and a sum of 1.

[0061] Assume that the input vector is z = [z1, z2, z3,..., zn], and the output vector after being processed by the softmax function is y = [y1, y2, y3,..., yn], where:

[0062]

[0063] In this way, each element of the output vector is between 0 and 1, and the sum of all elements is 1, which conforms to the definition of probability distribution.

[0064] In a neural network, Softmax is often connected after the fully connected layer to convert the linear output into probabilities for each class, which are used by the cross-entropy loss function to calculate the loss for optimizing the model parameters.

[0065] 3. Model training: The convolutional neural network is trained using the prepared dataset. During the training process, a loss function and an optimization algorithm are used to minimize the prediction error. In the present invention, the cross-entropy loss function is adopted.

[0066] The cross-entropy loss function is used to measure the difference between two probability distributions. In a classification problem, assuming that the probability distribution of the true label is p (usually a one-hot encoded vector, where only the position of the correct class is 1 and the other positions are 0), and the predicted probability distribution is q, then the cross-entropy loss function:

[0067]

[0068] Cross-entropy is used to train classification models. The smaller its value, the closer the predicted probability distribution is to the true probability distribution. During the model training process, the goal is to minimize the cross-entropy loss so that the model's prediction results are as close as possible to the true labels.

[0069] In this application, for the classification task of neural networks, the Softmax function and the cross-entropy loss function are combined. The Softmax function converts the output of the neural network into a probability distribution, and the cross-entropy loss function calculates the loss based on this predicted probability distribution and the true labels. The above combination method can effectively train the classification model, enabling the model to better learn the differences between different classes, improve the classification accuracy, and finally optimize the model performance by iteratively updating the weights and biases.

[0070] 4. Model evaluation and optimization: The trained model is evaluated using an independent test dataset. Metrics such as classification accuracy, precision, recall, and F1-score are calculated to evaluate the performance of the model. The model is optimized according to the evaluation results, such as adjusting the network structure, hyperparameters, or using regularization techniques, etc.

[0071] When performing wire harness color detection, the trained convolutional neural network model is applied to the actual scenario of identifying the color arrangement of vehicle wire harnesses. By collecting wire harness images in real time and inputting them into the model, the recognition result of the color arrangement is obtained. Corresponding operations are performed according to the recognition result. The diversity of the data set is expanded through the above data augmentation technology; and an efficient convolutional neural network structure and algorithm are used to reduce the amount of calculation and improve the accuracy of color detection.

[0072] As is known by common technical knowledge, the present invention can be implemented by other embodiments that do not depart from its spiritual essence or essential features. Therefore, the above-disclosed embodiments are illustrative in all aspects and not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are encompassed by the present invention.

Claims

1. A vehicle wiring harness visual welding system, characterized in that: The following steps are involved: S1: Wire stripping: use wire strippers to strip the wires, and the stripping length is adjusted according to the process requirements; S2: Color detection, using machine vision to create a classifier and train the trained model to be applied to the color detection of automotive wiring harnesses to detect whether there is an error in the color arrangement; S3: pre-welding; S4: Heating: Use the preheated soldering iron tip to heat the component leads and pads; S5: Wire feeding, holding the tin wire on both sides and feeding it to the heated welding area; S6: Leave. After the solder wire melts and slowly flows to two-thirds of the entire pad, quickly evacuate the solder wire and soldering iron tip. S7: Solidification. After the soldering iron tip is removed, avoid shaking the components and circuit boards to allow the solder joints to solidify naturally. For components that are easily damaged by scalding, blow air during welding to accelerate solidification. S8: Check the welding quality to ensure that the solder joints have moderate tin content, bright solder joints, and strong welding.

2. The automotive wiring harness visual welding system according to claim 1, characterized in that: In step S2, the color detection includes image acquisition and preprocessing, color space conversion and recognition, color arrangement detection, and result output and verification steps. According to the requirements and data sets of color arrangement recognition, a convolutional neural network structure is designed, and the model is trained using a deep learning method. The trained convolutional neural network model is applied to the scenario of color arrangement recognition of automotive wiring harnesses. By real-time acquisition of wiring harness images and inputting them into the model, the color arrangement recognition results are obtained, and corresponding operations are performed according to the recognition results.

3. The vehicle wiring harness visual welding system according to claim 2, characterized in that: The image acquisition includes using a high-resolution camera or image acquisition device to obtain clear images of the automotive wiring harness, which contain multiple colors of the wiring harness and background information, and the image preprocessing includes performing median filtering and image enhancement preprocessing operations on the acquired images to improve image quality and reduce noise and interference.

4. The automotive wiring harness visual welding system according to claim 2, characterized in that: The color space conversion and recognition includes: Color space conversion: convert the image from RGB color space to HSV color space; Color recognition: In the HSV space, color recognition is performed based on the difference in hue values. For each color, a hue range is set in the HSV space, and then the pixel hue value in the image is compared with this range to determine the color of the pixel.

5. The automotive wiring harness visual welding system according to claim 2, characterized in that: The color arrangement detection comprises: Color region segmentation: Use threshold segmentation, region filling, and connected segmentation techniques to segment the color regions in the image. These regions correspond to different color parts in the line bundle. Color area positioning: According to the position information of the segmented color area, determine the position of each color in the wire harness. This can be achieved by measuring the center position, boundary position, etc. of the color area; Color arrangement order determination: According to the position information of the color area, the color arrangement order is determined, and the color areas are sorted or encoded to clearly indicate the arrangement of the colors in the wiring harness.

6. The automotive wiring harness visual welding system according to claim 2, characterized in that: The convolutional neural network includes multiple convolutional layers, pooling layers, fully connected layers and output layers. In the convolutional layers, convolution kernels of different sizes and numbers are used to extract color features. In the pooling layers, maximum pooling or mean pooling is used to reduce the dimension and amount of calculation of data. In the fully connected layers, weight matrices and bias vectors are used to perform nonlinear combinations of features. In the output layer, the Softmax function is used to output the probability distribution of color arrangement. The convolutional neural network is trained using the dataset. During the training process, the loss function and optimization algorithm are used to minimize the prediction error, and the model performance is optimized by iteratively updating the weights and biases.

7. The automotive wiring harness visual welding system according to claim 6, characterized in that: The loss function and optimization algorithm use the cross entropy loss function, which is used to measure the difference between two probability distributions. The cross entropy loss function formula is: Among them, p is the probability distribution of the true label, and q is the predicted probability distribution.

Images