Automatic spring production process and system

Abstract

The invention relates to the field of intelligent manufacturing and mechanical automation, and discloses an automatic spring production process which comprises the following steps: S1, performing quality inspection on a steel wire by using a machine vision system; s2, a steel wire is cut through a numerical control cutting machine according to the spring specification; s3, the steel wire is precisely conveyed to forming equipment through the conveying belt to be automatically formed; s4, a numerical control forming machine is used for automatic forming; s5, the formed spring is subjected to heat treatment, cooling is conducted, and then coding spraying is conducted; and S6, based on the historical production data and the real-time process data, predicting and optimizing spraying parameters through a machine learning model. By introducing a machine learning model trained based on historical production data and real-time process data into the spraying parameter optimization step, predictive adjustment of key process parameters such as spraying pressure, spraying speed and ink flow is realized, and a spraying control effect dynamically adapting to production change is obtained.

Description

Technical Field

[0001] The present invention relates to the fields of intelligent manufacturing and mechanical automation, and in particular to an automated spring production process and system. Background Art

[0002] As a basic mechanical component, springs are widely used in the automotive, electronics, electrical appliance and other fields. Although their production technology has a long history, most small and medium-sized enterprises currently still rely on semi-automated production methods, especially in the back-end identification and quality links, and the level of intelligence is relatively low.

[0003] On existing spring production lines, inkjet printing is typically performed after manually setting spray parameters. During operation, changes in ambient temperature and humidity, printhead wear, or slight fluctuations in ink quality can cause fluctuations in print quality. However, operators often struggle to detect these issues immediately, relying on post-processing inspections to identify problems. This approach results in significant delays and can easily lead to batch scrapping.

[0004] On the other hand, spray coating quality assessments often rely on manual visual inspection or simple image comparison methods. These algorithms are sensitive to lighting variations, limiting recognition accuracy and causing frequent misjudgments and omissions. This is especially true when spraying on small, reflective metal surfaces like springs, where traditional recognition methods struggle.

[0005] Furthermore, most current production lines lack an effective closed-loop quality control system. Even when coding errors are identified, there's no automated mechanism for handling them. Defective products are often manually removed, cleaned, and reworked, resulting in low efficiency and frequent introduction of new deviations through human error.

[0006] Furthermore, the final packaging and shipping process for springs still relies heavily on manual labor. The scanning, recording, and labeling processes lack systematic integration, making traceability information easily missed and data consistency difficult to ensure. With the growth of product batches and the expansion of product categories, this operating method has become increasingly difficult to meet the dual requirements of quality control and flexible production.

[0007] There have been some attempts in the industry. For example, some companies have introduced image recognition equipment or simple PLC automatic sorting logic. However, these lack deep integration with machine learning models, making it impossible to achieve true data-driven adjustment of spraying strategies, and the overall system has limited intelligence. Summary of the Invention

[0008] In response to the shortcomings of the existing technology, the present invention provides an automated spring production process, which solves the problems of unstable spraying quality, lack of automatic identification and parameter optimization, low rework efficiency and information islands in the outbound process in the existing spring production.

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions: A spring automated production process, comprising the following steps:

[0010] S1: Use machine vision system to check the quality of steel wire;

[0011] S2: Cut the steel wire according to the spring specifications using a CNC cutting machine;

[0012] S3: The conveyor belt accurately delivers the steel wire to the forming equipment for automatic forming;

[0013] S4: Use CNC molding machine for automated molding;

[0014] S5: heat-treating the formed spring and then spraying the coding after cooling;

[0015] S6: Predict and optimize spray parameters using machine learning models based on historical production data and real-time process data;

[0016] S7: The spring after spraying is inspected for quality by a machine vision system, and the DM code spraying effect is analyzed using a deep learning algorithm;

[0017] S8: Adjust spraying parameters according to spraying quality feedback;

[0018] S9: Sorting and reworking of unqualified springs, cleaning the sorted springs and spraying them again according to the optimized spraying parameters, and re-inspecting the quality;

[0019] S10: Automatically pack qualified springs and complete outbound management.

[0020] An automated spring production system, comprising:

[0021] Steel wire quality inspection module, used to inspect the quality of steel wire using a machine vision system;

[0022] Wire cutting module, used to receive the spring after the wire quality inspection module, and cut the wire according to the spring specifications through the CNC cutting machine;

[0023] A conveying module is used to transport spring raw materials and qualified springs and unqualified springs between modules through a conveyor belt;

[0024] The spring forming module is used to receive the steel wire delivered by the steel wire feeding module and complete the spring forming through the CNC forming machine;

[0025] Heat treatment and cooling module, used for heating and cooling the spring after forming;

[0026] Spraying parameter optimization module, which is used to optimize spraying parameters through machine learning based on historical data and real-time process data, and implement DM code spraying;

[0027] The spraying quality inspection module is used to inspect the quality of the DM code on the spring after spraying using a machine vision system;

[0028] Spraying quality feedback and adjustment module, used to adjust the parameters in the spraying process according to the quality inspection results;

[0029] The error spring processing and rework module is used to sort and clean the unqualified springs and then spray them again through the spray parameter optimization module;

[0030] The automatic packaging and outbound management module is used to automatically package qualified springs and manage outbound delivery.

[0031] Preferably, the spraying parameter optimization module includes:

[0032] Data acquisition unit, used to collect relevant data during the production process in real time;

[0033] A data processing unit, used for preprocessing and cleaning the collected data;

[0034] Machine learning model unit, used to train machine learning models based on historical data and real-time data to predict and optimize spraying parameters;

[0035] The optimization result output unit is used to output the optimized spraying parameters and transmit them to the spraying equipment for adjustment.

[0036] Preferably, the spraying quality detection module includes:

[0037] Image acquisition unit, used to capture the details of the DM code on the spring after spraying using a high-resolution camera

[0038] An image processing unit, used to perform processing on the collected images, including denoising, interference removal, and image enhancement;

[0039] A deep learning analysis unit for analyzing images based on convolutional neural networks;

[0040] The quality assessment unit is used to evaluate the spraying quality based on the deep learning analysis results and mark qualified and unqualified products.

[0041] Preferably, the spraying quality feedback and adjustment module includes:

[0042] A quality detection feedback unit, used to receive feedback information from the spraying quality detection module;

[0043] Parameter adjustment unit, used to adjust the spraying pressure, spraying speed and ink flow during the spraying process according to the quality inspection feedback results;

[0044] The control unit is used to adjust the operation of the spraying equipment in real time according to the optimized spraying parameters.

[0045] Preferably, the error spring processing and reworking module includes:

[0046] a non-conforming spring identification unit, connected to the quality assessment unit, for marking non-conforming springs;

[0047] Automatic sorting unit, used to automatically sort out unqualified springs from qualified springs and send them for cleaning;

[0048] Cleaning unit, used to clean unqualified springs;

[0049] Secondary spraying unit, used for secondary spraying of the cleaned spring according to optimized spraying parameters;

[0050] The rework quality inspection unit is used to inspect the quality of springs after secondary spraying.

[0051] Preferably, the quality assessment unit compares the recognition score based on the deep learning analysis unit with a preset threshold value. When the score is greater than the threshold value, the spring spraying is judged to be qualified, otherwise it is judged to be unqualified.

[0052] Preferably, the qualified springs are connected to the automatic packaging and outbound management module through the transmission module, and the unqualified springs are connected to the error spring processing and re-work module through the transmission module.

[0053] Preferably, the deep learning analysis unit extracts and recognizes features of the spraying image through a convolutional neural network model, and the expression of the convolution operation in the convolutional neural network model is:

[0054]

[0055] Among them: F x,y,c′ : Output the value of the c′th channel of the feature map at position (x, y); I x+i-1,y+j-1,c : The pixel value of the input image at the cth channel and position (x+i-1, y+j-1); K i,j,c,c′ : The convolution kernel weight between the cth input channel and the c′th output channel, at the convolution kernel position (i, j); B c′ : Bias term of output channel c′; C in : number of input channels; k: length of convolution kernel; σ(·): activation function, used to introduce ReLU.

[0056] Preferably, the quality assessment unit normalizes the classification scores output by the convolutional neural network model and determines whether the spraying quality is qualified using the following formula:

[0057] If Q≥θ, it is judged as "qualified", otherwise it is "unqualified";

[0058] Where: Q: represents the predicted probability of spraying quality being "qualified"; S ok : The score representing the “qualified” category in the deep learning model output; S fail : represents the score of the “unqualified” category; θ: the preset quality judgment threshold, used to control the confidence level of qualified judgment; exp(·): exponential function, used for Softmax normalization.

[0059] Material toggling mechanism, its both sides respectively have a cylinder pressure, and the cylinder pressure bar connects swing arm, and the swing arm end face has hook portion, and a bar passes position between the end of two swing arms and the hook portion.

[0060] The top of the hydraulic cylinder is connected with the hydraulic cylinder to the upper end of the hydraulic cylinder, and the lower end of the hydraulic cylinder is connected with the hydraulic cylinder to the upper end of the hydraulic cylinder.

[0061] The present invention provides an intelligent control process for automated spring production, which has the following beneficial effects:

[0062] 1. The present invention introduces a machine learning model based on historical production data and real-time process data training in the spray parameter optimization step to achieve predictive adjustment of key process parameters such as spray pressure, spray speed, and ink flow, thereby obtaining a spray control effect that dynamically adapts to production changes, effectively solving the technical problems of delayed response and inaccurate adjustment of manually set parameters.

[0063] 2. The present invention uses a convolutional neural network model to extract features and classify the post-spraying image in the spraying quality detection step, thereby realizing automated intelligent judgment of the DM code spraying quality, achieving higher adaptability and recognition accuracy than traditional rule recognition, and effectively solving the problem of low recognition rate of non-standard spraying defects.

[0064] 3. The present invention realizes closed-loop recovery and repair of unqualified spray products by setting up a resumption process of unqualified spring identification, automatic sorting, cleaning and secondary spraying, thereby improving the yield rate and reducing material waste, and effectively overcoming the problem of resource waste caused by direct rejection of defective products in traditional production lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a schematic diagram of the steps of the present invention;

[0066] Figure 2 Schematic diagram of the system architecture of the present invention;

[0067] Figure 3 This is a schematic diagram of the overall front structure of the deburring device of the present invention;

[0068] Figure 4 This is a schematic diagram of the overall rear structure of the deburring device of the present invention;

[0069] Figure 5 for Figure 4 Schematic diagram of the structure at A;

[0070] Figure 6 This is a schematic structural diagram of the positioning plate of the deburring device of the present invention;

[0071] Figure 7 This is a schematic diagram of the top structure of the bottom plate of the deburring device of the present invention;

[0072] Figure 8 This is a schematic diagram of the cross-sectional structure of the hydraulic oil cylinder of the deburring device of the present invention;

[0073] Figure 9 It is a schematic diagram of the cross-sectional structure of the auxiliary oil cylinder of the deburring device of the present invention. DETAILED DESCRIPTION

[0074] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0075] Example 1:

[0076] Please see the attached Figure 1 , an embodiment of the present invention provides a spring automated production process, comprising the following steps:

[0077] S1: Use machine vision system to check the quality of steel wire;

[0078] S2: Cut the steel wire according to the spring specifications using a CNC cutting machine;

[0079] S3: The conveyor belt accurately delivers the steel wire to the forming equipment for automatic forming;

[0080] S4: Use CNC molding machine for automated molding;

[0081] S5: heat-treating the formed spring and then spraying the coding after cooling;

[0082] S6: Predict and optimize spray parameters using machine learning models based on historical production data and real-time process data;

[0083] S7: The spring after spraying is inspected for quality by a machine vision system, and the DM code spraying effect is analyzed using a deep learning algorithm;

[0084] S8: Adjust spraying parameters according to spraying quality feedback;

[0085] S9: Sorting and reworking of unqualified springs, cleaning the sorted springs and spraying them again according to the optimized spraying parameters, and re-inspecting the quality;

[0086] S10: Automatically pack qualified springs and complete outbound management.

[0087] Specifically, in this embodiment, step S1 performs quality inspection on the original steel wire through the configured machine vision system.

[0088] Generally speaking, surface defects of steel wire (such as scratches, rust, and bends) will directly affect the quality of subsequent spring forming.

[0089] Specifically, the machine vision system includes an industrial camera, an image acquisition card, and an image recognition software module, which uses edge detection and color analysis algorithms to identify the wire state. In some embodiments, tolerance standards can be set, and if the detection results exceed the tolerance range, the system will issue a rejection instruction.

[0090] In this embodiment, step S2 uses a CNC cutting machine to cut qualified wire to a fixed length. The cutting length is automatically determined by a preset database based on the spring type. In practice, the control system adjusts the cutting length based on the order model to ensure process adaptability.

[0091] In this embodiment, step S3 uses a conveyor belt to precisely deliver the cut wire to the forming equipment. Specifically, the conveyor belt system is equipped with a position sensor and synchronous encoder to achieve fixed-point alignment control. In one possible implementation, a servo-driven transmission method is used, combined with adaptive conveyor speed adjustment, to ensure that the feeding action is synchronized with the downstream equipment.

[0092] In this embodiment, step S4 utilizes a CNC forming machine to automatically form the steel wire. Generally, this forming process is controlled based on spring parameters, including wire diameter, number of turns, and pitch. Alternatively, the CNC system reads a parameter table to adjust the motion trajectory of the forming die in real time. In some embodiments, a dynamic compensation module is provided to automatically correct for errors in the robotic arm to improve forming accuracy.

[0093] In this embodiment, step S5 performs heat treatment and cooling treatment on the spring after the forming is completed, and then performs DM code spraying. Specifically, the heat treatment adopts an electric heating furnace, and the structure stabilization treatment is performed in a constant temperature environment. The cooling method can be air cooling or water cooling, which is set according to the properties of the spring material. The spraying equipment has a built-in high-precision nozzle, which completes the coded spraying operation by controlling the pressure and speed. In one possible implementation, the model adopts a multi-layer perceptron structure (MLP) or a gradient boosting tree, with historical production data and real-time collected process parameters as input features. The optimization goals include minimizing the defect rate, improving the recognition contrast, and maintaining the uniformity of the spraying. The output of the model is mapped to the spraying pressure P s , spraying speed v s 、Ink flow Q s Parameters such as speed and load are sent to the control module for real-time adjustment.

[0094] For example, in an optimization model, the prediction function can be expressed as:

[0095]

[0096] in: Spraying parameter optimization output results; X i: the i-th input feature, including ambient temperature, nozzle pressure, material batch, etc.; f(·): machine learning prediction function.

[0097] In this embodiment, step S7 uses a machine vision system to detect the spring after spraying, and combines a deep learning algorithm to analyze the DM code spraying quality.

[0098] The image acquisition unit uses a high-resolution CMOS camera, and the image processing unit performs noise filtering, image enhancement and region segmentation on the image.

[0099] In this embodiment, step S8 automatically adjusts the spraying parameters according to the image analysis results.

[0100] In this embodiment, step S9 sorts and reprocesses unqualified springs. After being marked by the identification unit, unqualified springs are automatically sorted and sent to the cleaning unit by the automatic sorting system. After cleaning, they enter the second spraying process based on the newly optimized parameters and repeat the image inspection. Optionally, a maximum number of re-sprays can be set for re-processed springs to avoid wasting resources. If two consecutive spraying attempts fail, they are automatically rejected and classified as scrap.

[0101] In this embodiment, step S10 automatically packages and manages the release of qualified springs. The packaging module and the transport module work together to automatically box and label the springs based on the quality assessment results. In practice, a QR code can be used to bind the product number and production parameters, enabling traceable release management.

[0102] The process also includes a deburring step. After the steel wire is cut, the deburring operation is performed by a deburring device. The deburring device includes a base plate 1, and a support plate 2 fixedly connected to the top of the base plate 1 is symmetrically provided on the top of the base plate 1. One side of one support plate 2 is fixedly connected to a horizontally arranged hydraulic cylinder 3, and one side of the other support plate 2 is provided with a fixing component 4. The output shaft of the hydraulic cylinder 3 passes through the support plate 2 and is connected to the support plate 2 in a sliding manner. One end of the output shaft of the hydraulic cylinder 3 is fixedly connected to a rotating motor 5, and the output end of the rotating motor 5 is fixed. It is connected to a grinding disc 6, and a slide groove 7 is provided on one side of the support plate 2 and is opened at the top of the base plate 1. Positioning grooves 8 opened at the top of the base plate 1 are symmetrically provided on both sides of the slide groove 7. A slider 9 is slidably connected to the inner side of the slide groove 7, and a support rod 10 is fixedly connected to one side of the slider 9. A bolt 11 is fixedly connected to the top of the slider 9, and a nut 12 is spirally connected to the outside of the bolt 11. A positioning plate 13 is provided between the nut 12 and the slider 9, and a circular hole 14 is provided on the inner side of the positioning plate 13, which passes through the positioning plate 13 from top to bottom, and a positioning block 15 is fixedly connected to the bottom end of the positioning plate 13.

[0103] A symmetrically arranged side plate 16 is fixedly connected between the two support plates 2 and fixedly connected to the top of the bottom plate 1. One of the side plates 16 is rotatably connected to a top plate 17 at the top. A leakage hole 18 is provided on the inner side of the bottom plate 1, which passes through the bottom plate 1 from top to bottom. A collection box 19 is fixedly connected to the bottom end of the bottom plate 1. A drawer 20 is slidably connected to the inner side of the collection box 19. The collection box 19 and the drawer 20 can collect iron filings produced by grinding. The bolt 11 is located on the inner side of the circular hole 14, and the positioning block 15 is located on the inner side of the positioning groove 8. There are multiple leakage holes 18, and the leakage holes 18 are distributed in an array on the inner side of the bottom plate 1. The arrangement of multiple leakage holes 18 allows iron filings to enter the inner side of the drawer 20 more smoothly. The fixing component 4 includes a hydraulic oil cylinder 401 that is horizontally arranged and fixedly connected to one side of the support plate 2. A first piston 402 is provided on the inner side of the hydraulic oil cylinder 401. One side of the first piston 402 A driving rod 403 is fixedly connected to the hydraulic oil cylinder 401 and is slidably connected to the hydraulic oil cylinder 401. The end of the driving rod 403 away from the first piston 402 is fixedly connected to the disc 404. The outside of the hydraulic oil cylinder 401 is fixedly connected to a secondary oil cylinder 405 that passes through the hydraulic oil cylinder 401. The inside of the secondary oil cylinder 405 is provided with a second piston 406 that fits tightly with the inside of the secondary oil cylinder 405. One side of the second piston 406 is fixedly connected to a clamping rod 407 that passes through the secondary oil cylinder 405 and is slidably connected to the secondary oil cylinder 405. The end of the clamping rod 407 away from the piston is fixedly connected to a fixing plate 408. The arrangement of the fixing component 4 can make the spring more evenly stressed when fixed and not easy to fall off. The outer side of the disc 404 is fixedly connected to the top of the support rod 10. There are multiple secondary oil cylinders 405, and the secondary oil cylinders 405 are distributed in an array on the outside of the hydraulic oil cylinder 401.

[0104] Working process: First, fix the spring, put the spring on the outside of the hydraulic oil cylinder 401, and push the disc 404 at one end of the driving rod 403. At this time, the first piston 402 moves inside the hydraulic oil cylinder 401, and the slider 9 on the side of the support rod 10 slides inside the slide groove 7. The positioning block 15 at the bottom end of the positioning plate 13 moves inside the positioning groove 8. The hydraulic oil inside the hydraulic oil cylinder 401 is pushed into the inside of the auxiliary oil cylinder 405 by the first piston 402. The second piston 406 inside the auxiliary oil cylinder 405 moves to drive the clamping rod 407 and the fixing plate 408 to move until the outside of the fixing plate 408 is tightly fitted with the inside of the spring. The setting of the fixing component 4 can make the force on the spring more uniform when fixed and not easy to fall off. Then, The nut 12 is tightened on the outside of the bolt 11 on the inside of the circular hole 14 until the bottom end of the positioning block 15 is tightly fitted with the bottom end of the inner side of the positioning groove 8. At this time, the disc 404 can be fixed to prevent the hydraulic oil from flowing back to push the first piston 402 to move. Next, the spring is polished, the top plate 17 on the top of the side plate 16 is closed, the hydraulic cylinder 3 on one side of the support plate 2 is started, and the rotating motor 5 is started. The output end of the rotating motor 5 drives the grinding disc 6 to rotate, and the output shaft of the hydraulic cylinder 3 extends to drive the grinding disc 6 to contact the spring. The iron filings polished out enter the drawer 20 inside the collection box 19 through the leakage hole 18 on the bottom plate 1. When the drawer 20 is full, the drawer 20 is directly pulled out, and the iron filings in the drawer 20 are poured out for recycling.

[0105] Example 2:

[0106] Please see the attached Figure 2 , an embodiment of the present invention provides a spring automated production process, comprising:

[0107] Steel wire quality inspection module, used to inspect the quality of steel wire using a machine vision system;

[0108] Specifically, the image acquisition device usually uses an industrial-grade linear array or area array camera, which is selected and matched according to the outer diameter of the steel wire and the travel speed. In some embodiments, the area array camera has a resolution of not less than 5 million pixels, and the shutter time can be adjusted to the microsecond level. In order to avoid the reflection of the steel wire affecting the image quality, it is generally equipped with a coaxial light source or annular backlight, and the wavelength is often a monochromatic LED in the visible light range. As a possible implementation method, the red light band (625-660nm) is more effective in identifying small scratches, and is particularly suitable for surface inspection of carbon steel wire.

[0109] Image acquisition and synchronization triggering are performed by an encoder or laser beam sensor. As the wire moves, the camera captures image frames at regular intervals. Each frame corresponds to a specific location on the wire, completely reproducing the entire wire surface in a "scan-like" manner. The acquisition frame rate depends on the wire speed and is typically set between 50 and 150 fps.

[0110] The image processing terminal executes the defect detection algorithm. The processing typically includes steps such as image grayscale conversion, edge enhancement, morphological filtering, and contour analysis. For scratch detection, for example, adaptive threshold segmentation and line structure feature extraction are used. Surface rust identification combines color information with regional texture analysis. The detection control unit receives the recognition results and issues control commands. If a section of wire is deemed unqualified, it is automatically separated from the main wire using a linked robotic arm or meter. Qualified sections are then transferred to the subsequent CNC cutting module. The entire recognition-judgment-rejection process is a closed-loop process.

[0111] As an expansion option, the module supports multi-angle acquisition, adding two or more cameras to provide 360° composite imaging of the entire wire surface from different directions. This configuration is particularly suitable for high-strength steel wire, whose surface defects are more difficult to fully visualize in a single-angle image.

[0112] In terms of beneficial effects, deploying a visual inspection module at the wire feeding stage significantly improves the accuracy of incoming material quality control. Compared with traditional manual spot checks, this system offers faster inspection speed, better recognition repeatability, and independence from subjective judgment. Especially in high-volume continuous production, it effectively prevents downstream problems such as forming failures or heat treatment cracks caused by wire defects. Furthermore, defect images can be uploaded to a database for subsequent traceability analysis and model training, forming an iteratively optimized closed-loop quality system.

[0113] Wire cutting module, used to receive the spring after the wire quality inspection module, and cut the wire according to the spring specifications through the CNC cutting machine;

[0114] Specifically, in one embodiment of the present invention, the wire cutting module is located downstream of the wire quality inspection module. It receives quality-inspected raw wire and performs high-precision cutting according to the target spring specifications. The module primarily comprises a feed synchronization unit, a CNC cutting machine, a specification instruction control unit, and a safety interlock mechanism.

[0115] Typically, this module physically interfaces directly with the quality inspection module. A conveyor belt or roller guide mechanism guides the wire into the cutting area. To ensure accurate cutting position, a photoelectric sensor array is installed at the entrance of the cutting area to detect the position of the wire's leading edge and achieve dynamic synchronization.

[0116] The CNC cutting machine is the core actuator. Its typical structure includes a servo drive system, a cutter actuator, and a closed-loop position feedback unit. Optionally, the cutter can be a tungsten carbide blade, suitable for cutting high-strength carbon spring steel wire. In some embodiments, laser or plasma cutting can also be used for cutting special materials or when a burr-free process is required.

[0117] A conveying module is used to transport spring raw materials and qualified springs and unqualified springs between modules through a conveyor belt;

[0118] Specifically, the transmission module is used to realize the transportation and diversion of steel wire raw materials and spring products between various functional modules. Its core is composed of a multi-section variable-speed conveyor belt, a position sensing element, a classification execution unit and a logic controller. The front end of the module is connected to the wire cutting module, and the rear end is connected to the forming equipment, heat treatment unit, spraying system and inspection and resumption channel in sequence. The conveyor belt body is made of wear-resistant composite material, and a number of limit bumps are provided on the surface to prevent the spring from rolling or offset. In some areas, a roller + belt combination conveying structure can be used to adapt to spring parts of different shapes and diameters.

[0119] Several photoelectric switches or laser displacement sensors are embedded in the conveying path to monitor the spring position and beat synchronization in real time. The control system can determine whether to start the subsequent workstation or temporary buffer based on these position signals. In some embodiments, an electric slide rail or pneumatic swing arm is provided at the end of the conveying branch for the diversion of qualified and unqualified products. The unqualified parts are marked by the quality inspection module and then transferred to the resumption branch line by the actuator, while the qualified products enter the packaging preparation area. The conveying rhythm can be adjusted in real time by the main control PLC according to the operating status of the previous and next processes to avoid accumulation or idle time on the production line. At the same time, in order to improve the information tracking capability, the module can also be embedded with an RFID reader or visual recognition device.

[0120] The spring forming module is used to receive the steel wire delivered by the steel wire feeding module and complete the spring forming through the CNC forming machine;

[0121] Specifically, the spring forming module is arranged downstream of the wire cutting module, and is used to receive the steel wire that has been cut to a fixed length and complete the automatic forming operation of the spring. The core is composed of a CNC forming machine body, a wire guide mechanism, a wire feeding drive assembly, a winding control system and a forming parameter input interface. The forming machine body adopts a dual-axis CNC structure, one axis controls the wire feeding length, and the other axis drives the curling blade to complete the spiral winding. Springs of different specifications are adapted by replacing the forming tool and adjusting the guide radius. After the steel wire enters the wire guide assembly through the feed end of the forming machine, it is first clamped and then accurately advanced by the servo roller drive system. The advancement length and speed are adjusted in real time by the controller, and the forming radius is achieved by winding the wire with a side winding arm. The winding arm is fixed on the motion curve and is synchronized with the CNC trajectory control module. Common trajectory formats include step control and continuous interpolation. Both methods are based on the G code instruction library.

[0122] In some embodiments, in order to ensure pitch consistency and turn accuracy, the system introduces a real-time tension compensation mechanism, which dynamically adjusts the wire feeding speed by feeding back the force changes on the wire. Some processes have high requirements for the spring end structure. For example, if a fastening ring or a flattened end needs to be set, an additional curve segment needs to be added to the control program. After forming, the spring automatically separates from the forming shaft and falls into the receiving device. Some structures are equipped with a robotic arm to assist in transporting it to the next process. After forming, the size of the spring is spot-checked by an online laser measuring device. The detection parameters include outer diameter, turn pitch, total number of turns, flatness, etc. The detection data can be used to form a deviation model for subsequent automatic correction.

[0123] Heat treatment and cooling module, used for heating and cooling the spring after forming;

[0124] Specifically, the heat treatment and cooling module is arranged after the spring forming module, and is used to perform structural stabilization treatment on the formed spring to ensure that it has the required mechanical properties and dimensional retention capabilities. The module usually includes a continuous electric heating furnace, a cooling channel, a temperature control unit and a conveying component. The heat treatment part adopts far-infrared or induction heating to heat the spring to a set temperature range (such as 400°C~500°C) and maintain it for a certain time to eliminate residual stress. In some embodiments, the temperature control accuracy can reach ±3°C. The cooling part achieves rapid cooling through forced air cooling or spray cooling. The cooling rate is adjusted by the controller according to the material and spring specifications. The conveying system adopts a stainless steel chain plate structure to drive the spring to move continuously in the heat treatment and cooling zone. The beat and temperature zone length are coordinated to form a stable heating curve.

[0125] Spraying parameter optimization module, which is used to optimize spraying parameters through machine learning based on historical data and real-time process data, and implement DM code spraying;

[0126] Specifically, in one embodiment of the present invention, a spray parameter optimization module is used to intelligently adjust key process parameters in the spring spraying process to improve the clarity and adhesion stability of the DM code spraying. The module consists of a data acquisition unit, a parameter processing unit, a machine learning model unit and a spray instruction output interface. The data acquisition unit obtains process variables such as spraying pressure, nozzle temperature, ambient humidity, ink flow, and running speed in real time, and at the same time constructs a sample set in combination with historical spraying images and quality scores. The machine learning model adopts a multi-layer perceptron structure and outputs spray control parameters after repeated training.

[0127] The spraying quality inspection module is used to inspect the quality of the DM code on the spring after spraying using a machine vision system;

[0128] Specifically, in one embodiment of the present invention, a spraying quality detection module is arranged after the spraying process, and is used to perform online detection and quality assessment of the DM code printed on the surface of the spring. The module includes a high-resolution camera, a light source system, an image processing unit and a quality analysis algorithm component. The camera is used to capture the DM code image after spraying. An area array camera with more than 5 million pixels is usually used, and a coaxial light source or backlight is used to enhance image contrast and edge clarity. The image processing unit performs pre-processing operations such as grayscale normalization, edge enhancement, and area positioning on the collected image. Subsequently, a deep learning model (such as a convolutional neural network) is used to extract and identify image features. The system outputs the DM code recognition confidence and compares it with the set threshold to determine whether the spraying quality is qualified.

[0129] Spraying quality feedback and adjustment module, used to adjust the parameters in the spraying process according to the quality inspection results;

[0130] Specifically, the spraying quality feedback and adjustment module is used to receive the evaluation results output by the spraying quality detection module, and to make real-time corrections to the key parameters in the spraying process according to the detection conditions. The module consists of a quality feedback interface, a parameter adjustment unit, a control instruction generator, and an equipment execution interface. The quality feedback interface receives information such as the recognition confidence of the DM code, the defect type, and the image score.

[0131] The parameter adjustment unit compares the set threshold to determine whether the spraying status deviates from the target range. If an unqualified result occurs, the control logic immediately triggers the parameter optimization instruction. The instruction generator combines the spraying history data with the current environmental information to quickly adjust the spraying pressure, spraying speed, nozzle temperature or ink output flow and other parameters. In some embodiments, fuzzy control or PID algorithm is also introduced to ensure that the parameter adjustment process is smooth and not over-adjusted.

[0132] The error spring processing and rework module is used to sort and clean the unqualified springs and then spray them again through the spray parameter optimization module;

[0133] Specifically, the error spring processing and resumption module is used to automatically sort, clean and re-spray springs that fail the spray quality inspection results to ensure the overall product qualification rate. The module is mainly composed of an unqualified identification unit, an automatic sorting device, a cleaning unit, a resumption spray interface and a detection return path. The unqualified identification unit receives the marking signal from the spray quality inspection module, and numbers and records the springs with low identification confidence or defects such as broken codes, offsets, and ambiguities. The sorting device automatically guides the unqualified springs from the transmission path to the resumption channel through a pneumatic fork, slide rail or robotic arm.

[0134] The cleaning unit processes the residual ink on the surface of the spring, and usually adopts a multi-stage alkaline liquid spraying and water washing combination process to ensure that the original spray layer is completely removed to provide a clean surface for secondary spraying. After cleaning, the spring re-enters the spray parameter optimization module and completes the DM code spraying and quality inspection process again. In some embodiments, the resumption process limits the maximum number of re-spraying times, for example, it is set to 2 times. The spring that exceeds the limit is automatically scrapped and classified to avoid waste of resources. This module realizes the closed-loop reuse of unqualified products, greatly improving material utilization and production line yield.

[0135] The automatic packaging and outbound management module is used to automatically package qualified springs and manage outbound delivery.

[0136] Specifically, the automatic packaging and outbound management module is used to automatically collect, classify and package qualified springs that have passed quality inspection and complete the outbound process management. This module mainly includes a qualified product collection unit, an automatic boxing system, a label printing and pasting device, an outbound scanning and recording system, and a data upload interface. After the qualified springs are sent into the collection unit by the transmission module, they are automatically classified according to the preset batch or product model. The automatic boxing system will pack the specified number of springs into the packaging box or turnover box. The boxing process can be coordinated with the vibration alignment device to ensure neat arrangement. The label printing device automatically generates labels containing product number, specification, batch number, spray parameter summary and other information according to the production record and completes the pasting. The boxing is then confirmed by the scanning device and an outbound record is generated. The record is connected to the factory MES system to realize product traceability, inventory update and delivery information synchronization.

[0137] Example 3:

[0138] This embodiment provides a complete operational example of the intelligent control process for automated spring production. This example has been deployed and put into operation on a spring production line at a medium-sized automotive parts factory. The entire production line, starting with raw steel wire feeding, continues through multiple processes, including quality inspection, cutting, forming, heat treatment, spraying, testing, feedback, re-work, and packaging. System modules are interconnected via Industrial Ethernet and centrally managed and scheduled by a central control platform.

[0139] In this embodiment, the raw material steel wire is 0.8mm diameter oil-quenched and tempered spring steel, which is input into the system in the form of a coil. The steel wire first enters the steel wire quality inspection module, which uses two Basler acA2440 industrial cameras arranged in the upper and lower directions, and cooperates with a 650nm coaxial red light source system to continuously capture images of the steel wire surface. The detection logic is completed by the image processing module, which identifies common defects such as scratches, dents, and rust. Each image covers 60mm of steel wire length. The system analyzes the defect area ratio in real time and scores it according to the scoring threshold D. S>0.015 is used as the judgment standard. Steel wires exceeding the threshold are sent to the bypass for manual confirmation or directly rejected.

[0140] Qualified steel wire enters the wire cutting module. This embodiment utilizes a dual-servo roller feed system coupled with a high-speed shearing machine. The feed length is controlled by an incremental encoder, resulting in a cutting length of 180 mm ± 0.2 mm and a forming cycle of 1.6 seconds per piece. After cutting, the wire automatically falls into the alignment slot of the next station.

[0141] The steel wire is fed through a conveyor module into the spring forming module, using a Taiwanese brand dual-axis CNC forming machine with an automatic die change mechanism, enabling switching between different models within three minutes. The forming machine uses G-code to drive the wire feeding and winding operations. The forming program sets a total of nine turns, a free length of 23mm, and tight loops at the ends. After forming, the spring is grasped by a robotic arm and placed on a conveyor chain for the next heat treatment step.

[0142] The heat treatment and cooling module utilizes a continuous far-infrared heating furnace, with the heating section set at 460°C. The spring, with a length of 1.2m, passes through the furnace at a constant speed of 1.8m / min, maintaining heating for approximately 40 seconds. An air cooling channel is installed in the rear section, using directional air nozzles on both sides to rapidly cool the temperature to below 100°C within 3 seconds. Temperature changes within the module are monitored in real time by four sets of thermocouples to ensure uniform heating.

[0143] The heat-treated spring enters the spray parameter optimization module. The system automatically calls the corresponding spray parameters in the process database as initial values, including spray pressure of 0.22MPa, speed of 400mm / s, and ink flow rate of 0.6ml / min. The system's three-layer MLP model, built on the TensorFlow framework, receives real-time input features such as ambient temperature and humidity, nozzle temperature, and the spray ratings of the first five pieces. The model outputs the optimized spray speed and pressure, which are then transmitted to the inkjet controller for adjustment. S >0.015

[0144] After the DM code is sprayed, the paint quality inspection module immediately begins. A high-resolution GigE camera captures images and analyzes regional features. The images are then fed into a CNN model for recognition and scoring, outputting a pass probability value, Q. If Q exceeds 0.85, the paint is marked as unqualified. The system saves the images and inspection results for subsequent retrospective analysis.

[0145] The system then enters the spray quality feedback and adjustment module, which receives the latest spray result data and triggers the control logic to dynamically adjust the printhead parameters. For example, if the probability of recognition blur increases, the ink flow rate will be appropriately increased or the printhead speed will be reduced by 0.1 seconds per section. This feedback is updated every 20 springs.

[0146] Springs deemed unqualified are directed through the conveyor module to the faulty spring handling and reprocessing module. A sorting mechanism, comprised of pneumatic swing arms, directs springs into a cleaning channel based on control signals. The cleaning section is sprayed with an alkaline deinking agent for 60 seconds, then rinsed with high-pressure water and air-dried in the air-drying module. After treatment, the springs re-enter the spraying process and undergo re-inspection. If they again fail, they are sent to the rejection channel, while those that pass enter the packaging process.

[0147] The automatic packaging and delivery management module is equipped with a multi-station vibration arrangement system. Every 10 springs are automatically loaded into a customized blister. The laser printer prints the DM code summary label in real time and sticks it on the outer packaging. After packaging is completed, the delivery registration is carried out through the code scanning device and uploaded to the MES system, binding the order number, batch, parameters and inspection image data to support subsequent traceability queries.

[0148] The complete operation cycle of this embodiment is controlled within 4.5 seconds per spring, with a daily production capacity of more than 6,500 pieces. The system has been running stably for more than half a year with a failure rate of less than 0.3%. It has excellent scalability and industrial adaptability, verifying the technical feasibility and practical application value of the intelligent control process and system for automated spring production proposed in this invention.

[0149] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A spring automated production process, characterized in that: The following steps are involved: S1: Use machine vision system to check the quality of steel wire; S2: Cut the steel wire according to the spring specifications using a CNC cutting machine; S3: The conveyor belt accurately delivers the steel wire to the forming equipment for automatic forming; S4: Use CNC molding machine for automated molding; S5: heat-treating the formed spring and then spraying the coding after cooling; S6: Predict and optimize spray parameters using machine learning models based on historical production data and real-time process data; S7: The spring after spraying is inspected for quality by a machine vision system, and the DM code spraying effect is analyzed using a deep learning algorithm; S8: Adjust spraying parameters according to spraying quality feedback; S9: Sorting and reworking of unqualified springs, cleaning the sorted springs and spraying them again according to the optimized spraying parameters, and re-inspecting the quality; S10: Automatically pack qualified springs and complete outbound management.

2. A spring automated production system, according to the spring automated production process of claim 1, characterized in that: include: Steel wire quality inspection module, used to inspect the quality of steel wire using a machine vision system; Wire cutting module, used to receive the spring after the wire quality inspection module, and cut the wire according to the spring specifications through the CNC cutting machine; A conveying module is used to transport spring raw materials and qualified springs and unqualified springs between modules through a conveyor belt; The spring forming module is used to receive the steel wire delivered by the steel wire feeding module and complete the spring forming through the CNC forming machine; Heat treatment and cooling module, used for heating and cooling the spring after forming; Spraying parameter optimization module, which is used to optimize spraying parameters through machine learning based on historical data and real-time process data, and implement DM code spraying; The spraying quality inspection module is used to inspect the quality of the DM code on the spring after spraying using a machine vision system; Spraying quality feedback and adjustment module, used to adjust the parameters in the spraying process according to the quality inspection results; The error spring processing and rework module is used to sort and clean the unqualified springs and then spray them again through the spray parameter optimization module; The automatic packaging and outbound management module is used to automatically package qualified springs and manage outbound delivery.

3. The spring automated production system according to claim 2, characterized in that: The spraying parameter optimization module includes: Data acquisition unit, used to collect relevant data during the production process in real time; A data processing unit, used for preprocessing and cleaning the collected data; Machine learning model unit, used to train machine learning models based on historical data and real-time data to predict and optimize spraying parameters; The optimization result output unit is used to output the optimized spraying parameters and transmit them to the spraying equipment for adjustment.

4. The spring automated production system according to claim 2, characterized in that: The spraying quality detection module includes: Image acquisition unit, used to capture the details of the DM code on the spring after spraying using a high-resolution camera An image processing unit, used to perform processing on the collected images, including denoising, interference removal, and image enhancement; A deep learning analysis unit for analyzing images based on convolutional neural networks; The quality assessment unit is used to evaluate the spraying quality based on the deep learning analysis results and mark qualified and unqualified products.

5. The spring automated production system according to claim 2, characterized in that: The spraying quality feedback and adjustment module includes: A quality detection feedback unit, used to receive feedback information from the spraying quality detection module; Parameter adjustment unit, used to adjust the spraying pressure, spraying speed and ink flow during the spraying process according to the quality inspection feedback results; The control unit is used to adjust the operation of the spraying equipment in real time according to the optimized spraying parameters.

6. The spring automated production system according to claim 2, characterized in that: The error spring processing and resumption module includes: a non-conforming spring identification unit, connected to the quality assessment unit, for marking non-conforming springs; Automatic sorting unit, used to automatically sort out unqualified springs from qualified springs and send them for cleaning; Cleaning unit, used to clean unqualified springs; Secondary spraying unit, used for secondary spraying of the cleaned spring according to optimized spraying parameters; The rework quality inspection unit is used to inspect the quality of springs after secondary spraying.

7. The spring automated production system according to claim 4, characterized in that: The quality assessment unit compares the recognition score based on the deep learning analysis unit with a preset threshold value. When the score is greater than the threshold value, the spring spraying is determined to be qualified, otherwise it is determined to be unqualified.

8. The spring automated production system according to claim 7, characterized in that: The qualified springs are connected to the automatic packaging and outbound management module through the transmission module, and the unqualified springs are connected to the error spring processing and re-work module through the transmission module.

9. The spring automated production system according to claim 4, characterized in that: The deep learning analysis unit extracts and recognizes features of the spraying image through a convolutional neural network model. The expression of the convolution operation in the convolutional neural network model is: Among them: F x,y,c′ : Output the value of the c′th channel of the feature map at position (x, y); I x+i-1,y+j-1,c : The pixel value of the input image at the cth channel and position (x+i-1, y+j-1); K i,j,c,c′ : The convolution kernel weight between the cth input channel and the c′th output channel, at the convolution kernel position (i, j); B c′ : Bias term of output channel c′; C in : number of input channels; k: length of convolution kernel; σ(·): activation function, used to introduce ReLU.

10. The spring automated production system according to claim 4, characterized in that: The quality assessment unit normalizes the classification scores output by the convolutional neural network model and determines whether the spraying quality is qualified using the following formula: If Q≥θ, it is judged as "qualified", otherwise it is "unqualified"; Where: Q: represents the predicted probability of the spraying quality being "qualified"; S ok : The score representing the "qualified" category in the deep learning model output; S fail : represents the score of the "unqualified" category; θ: the preset quality judgment threshold, used to control the confidence level of the qualified judgment; exp(·): exponential function, used for Softmax normalization.

11. The spring automated production system according to claim 10, characterized in that: The invention also includes a spring deburring step. After the steel wire is cut, the deburring operation is performed by a deburring device. The deburring device includes a base plate (1). The top of the base plate (1) is symmetrically provided with a support plate (2) fixedly connected to the top of the base plate (1). One side of one of the support plates (2) is fixedly connected to a horizontally arranged hydraulic cylinder (3). The other side of the support plate (2) is provided with a fixing component (4). The output shaft of the hydraulic cylinder (3) passes through the support plate (2) and is connected to the support plate (2) in a sliding manner. One end of the output shaft of the hydraulic cylinder (3) is fixedly connected to a rotating motor (5). The output end of the rotating motor (5) is fixedly connected to a grinding disc (6). A slide groove (7) is provided on one side of the support plate (2) and is opened at the top of the bottom plate (1). Positioning grooves (8) are symmetrically provided on both sides of the slide groove (7) and are opened at the top of the bottom plate (1). A slider (9) is slidably connected to the inner side of the slide groove (7). A support rod (10) is fixedly connected to one side of the slider (9). A bolt (11) is fixedly connected to the top of the slider (9). A nut (12) is spirally connected to the outer side of the bolt (11). A positioning plate (13) is provided between the nut (12) and the slider (9). A circular hole (14) is provided on the inner side of the positioning plate (13) and passes through the positioning plate (13) from top to bottom. A positioning block (15) is fixedly connected to the bottom end of the positioning plate (13). A symmetrically arranged side plate (16) is fixedly connected between the two support plates (2) and is fixedly connected to the top of the bottom plate (1). The top of one of the side plates (16) is rotatably connected to a top plate (17). A material leakage hole (18) is provided on the inner side of the bottom plate (1) and passes through the bottom plate (1) from top to bottom. A collection box (19) is fixedly connected to the bottom end of the bottom plate (1). A drawer (20) is slidably connected to the inner side of the collection box (19). The fixed component (4) includes a hydraulic oil cylinder (401) that is horizontally arranged and fixedly connected to one side of the support plate (2). A first piston (402) is provided on the inner side of the hydraulic oil cylinder (401). One side of the first piston (402) is fixedly connected to a hydraulic oil cylinder (401). The oil cylinder (401) is provided with a driving rod (403) which is slidably connected to the hydraulic oil cylinder (401); the driving rod (403) is fixedly connected to a disc (404) at one end away from the first piston (402); the outer side of the hydraulic oil cylinder (401) is fixedly connected to a secondary oil cylinder (405) which is in communication with the hydraulic oil cylinder (401); the inner side of the secondary oil cylinder (405) is provided with a second piston (406) which is tightly fitted with the inner side of the secondary oil cylinder (405); one side of the second piston (406) is fixedly connected to a clamping rod (407) which penetrates the secondary oil cylinder (405) and is slidably connected to the secondary oil cylinder (405); the end of the clamping rod (407) which is away from the piston is fixedly connected to a fixing plate (408).

Images