Welding device for plastic oil tank production and intelligent control method thereof

By designing a welding device including slide rods, threaded pipes and fan chassis, the problems of pre-positioning and rear movement are solved, and rapid cooling is achieved using extraction pumps and water spray plates, improving welding efficiency and quality.

CN119078199AInactive Publication Date: 2024-12-06YANGZHOU CHANGYUN PLASTIC TECH CO LTD

Patent Information

Application Number
CN202411422280.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-11
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing welding devices for plastic oil tank production need to be positioned before welding, and cannot move quickly and cool down after welding, resulting in inefficiency.

Method used

A welding device including a base, a slide rod, a threaded pipe, an electric telescopic rod and a fan chassis is designed. The positioning and movement of the oil tank is achieved through the cooperation of the slide rod and a threaded pipe, and the extraction pump and a water spray plate are used for rapid cooling treatment.

Benefits of technology

It realizes convenient movement and rapid cooling of the fuel tank, improves welding efficiency, and reduces welding defect rate and production time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119078199A_ABST
    Figure CN119078199A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of plastic oil tank welding, in particular to a welding device for plastic oil tank production and an intelligent control method thereof.The welding device comprises a base, a supporting column is fixedly connected to the bottom of the base, a controller is fixedly connected to the outer side of the base, and a moving groove is formed in the base; a sliding rod is slidably connected to the interior of the moving groove, a mounting groove is fixedly connected to the top of the sliding rod, a threaded pipe is fixedly connected to the outer side of the mounting groove, a threaded rod is in threaded connection to the interior of the threaded pipe, a positioning cover is fixedly connected to the outer side of the threaded rod, and an L-shaped plate is fixedly connected to the outer side of the mounting groove; the oil tank welding device is simple in structure and convenient to use, an oil tank can be moved after being installed, the welding position of the oil tank can be rapidly cooled during welding, and therefore the welding efficiency is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a welding device for producing a plastic oil tank and an intelligent control method thereof, belonging to the technical field of plastic oil tank welding. Background Art

[0002] In addition to storage, the oil tank also plays the role of heat dissipation, separation of bubbles in the oil, and precipitation of impurities in the hydraulic system. There are many auxiliary parts installed in the oil tank, such as coolers, heaters, air filters, and liquid level gauges. Plastic oil tanks are generally installed on high-end light trucks. Plastic oil tanks have strong anti-corrosion capabilities. Plastic oil tanks will not produce some impurities due to corrosion, so that impurities will not enter the engine through the oil supply system and cause engine damage and reduce the service life. In the processing of plastic oil tanks, they need to be welded, so welding equipment will be used.

[0003] There are still some problems in the use of an existing welding device for producing plastic fuel tanks. Before the fuel tank is welded, it needs to be positioned, and the positioned fuel tank cannot be moved. After welding, the other side of the fuel tank needs to be welded. At this time, the fuel tank needs to be disassembled and reinstalled, which wastes a lot of time. After welding, the welding part of the fuel tank will generate high temperature, and the normal heat dissipation takes a long time, which makes it impossible to quickly dissipate heat. Summary of the invention

[0004] The purpose of the present invention is to provide a welding device for the production of plastic oil tanks and an intelligent control method thereof. The present invention has a simple structure and is easy to use. It can move the oil tank after installation and quickly cool the welding point of the oil tank during welding, thereby improving the welding efficiency and solving the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A welding device for producing plastic oil tanks and an intelligent control method thereof, comprising a base, a support column fixedly connected to the bottom of the base, a controller fixedly connected to the outer side of the base, a movable groove opened inside the base, a sliding rod slidably connected inside the movable groove, a mounting groove fixedly connected to the top of the sliding rod, a threaded tube fixedly connected to the outer side of the mounting groove, a screw connected to the inner thread of the threaded tube, a positioning cover fixedly connected to the outer side of the screw, an L-shaped plate fixedly connected to the outer side of the mounting groove, an electric telescopic rod fixedly connected to the outer side of the L-shaped plate, a push rod fixedly connected to the outer side of the electric telescopic rod, and a tank body mounted on the top of the mounting groove.

[0007] Furthermore, a gantry is fixedly connected to the top of the base, a connecting plate is fixedly connected to the top of the gantry, a motor is fixedly connected to the outer side of the connecting plate, a screw is fixedly connected to the output end of the motor, a ring is threadedly connected to the outer side of the screw, a moving rod is fixedly connected to the bottom of the ring, a slide groove is provided inside the gantry, an electric control box is fixedly connected to the bottom of the moving rod, a welding column is fixedly connected to the bottom of the electric control box, a water tank is fixedly connected to the outer side of the water tank, a display panel is fixedly connected to the outer side of the water tank, a water inlet cover is fixedly connected to the top of the water tank, an extraction pump is fixedly connected to the outer side of the extraction pump, a hose is fixedly connected to the outer side of the welding column, a connecting rod is fixedly connected to the outer side of the connecting rod, a water spray plate is fixedly connected to the outer side of the electric control box, an extrusion rod is fixedly connected to the outer side of the gantry, a fan box is fixedly connected to the top of the fan box, a power box is fixedly connected to the top of the power box, and an elastic switch is fixedly connected to the top of the power box.

[0008] Furthermore, the support columns are symmetrically distributed at the bottom of the base, and the mounting grooves are slidably connected to the top of the base via sliding rods.

[0009] Furthermore, the threaded tubes are symmetrically distributed on the outside of the installation groove, and the positioning cover is rotatably connected to the inside of the installation groove through a screw rod.

[0010] Furthermore, the electric telescopic rod is electrically connected to the controller, and one end of the push rod away from the electric telescopic rod is fixedly connected to the outside of the mounting groove.

[0011] Furthermore, the motor is electrically connected to the controller, and the screw rod is rotatably connected between the two connecting plates.

[0012] Furthermore, the moving rod is slidably connected inside the slide slot, and the electric control box is electrically connected to the controller.

[0013] Furthermore, the extraction pump is electrically connected to the controller, one end of the hose away from the extraction pump is fixedly connected to the inside of the water spray plate, and the squeezing rod and the elastic switch are on the same plane.

[0014] Furthermore, a welding device for producing a plastic fuel tank and an intelligent control method thereof include the following steps:

[0015] S1. First, place the oil tank body inside the installation groove, and position the oil tank body by using the screw and the positioning cover;

[0016] S2, the position of the oil tank body (13) inside the installation groove can be adjusted by using the electric telescopic rod and the push rod in coordination;

[0017] S3, starting through the controller and the electric control box, so that the welding column performs welding processing on the oil tank body at the bottom;

[0018] S4, the motor drives the screw rod to rotate, so that the moving rod drives the welding column to move on the top of the oil tank body;

[0019] S5. Use the extraction pump to extract the water from the water tank, and finally spray it out from the inside of the water spray plate to cool down the welded tank body and Xining:

[0020] S6, squeezing the elastic switch through the squeezing rod to start the fan box, thereby accelerating the drying of the water after cooling;

[0021] S7. Inspection and acceptance: After processing, the workpiece is manually inspected and accepted.

[0022] The beneficial effects of the present invention are:

[0023] (I) The present invention sets a base, and uses a sliding rod to make the installation groove slidably connected to the top of the base. First, the oil tank body is placed inside the installation groove, and the screw rod is rotated to rotate and move inside the threaded tube. The movement of the screw rod drives the positioning cover to move. At this time, the positioning cover makes the oil tank body fixedly installed inside the installation groove. When the position of the oil tank body needs to be moved, the electric telescopic rod on the outer side of the L-shaped plate is started by the controller, so that the electric telescopic rod drives the push rod to move. The movement of the push rod causes the installation groove to be pushed, so that the installation groove moves on the top of the base under the drive of the sliding rod. At this time, the other side of the oil tank body can be welded.

[0024] (ii) The present invention is provided with a base, and a gantry is fixedly connected to the top of the base. When the installed oil tank body needs to be welded, the electric control box is started by the controller, so that the welding column at the bottom of the electric control box welds the oil tank body. At this time, the motor is started to drive the screw rod to rotate between the connecting plates, and the rotation of the screw rod drives the ring to rotate. Since a moving rod is fixedly connected to the bottom of the ring, and the moving rod is slidably connected to the inside of the slide groove, when the screw rod rotates, the ring can drive the electric control box and the welding column at the bottom of the moving rod to move, thereby welding the oil tank body. The connecting edge is welded. During welding, the extraction pump on the outside of the water tank is started through the controller, so that the extraction pump draws out the water inside the water tank, and then enters the inside of the water spray plate through the hose. Since the water spray plate is located on the outside of the welding column, the water can be sprayed onto the welding point to cool the position after welding. When the electrical control box drives the welding column to move to the other side of the gantry, the extrusion rod will squeeze the elastic switch on the outside of the power box, so that the power box drives the fan box to start. At this time, the wind force generated by the start-up of the fan box will dry the cooled water, thereby achieving the effect of improving welding efficiency.

[0025] (III) Reinforcement learning (RL) enables the control system to autonomously learn the optimal strategy based on the interaction with the loop, while model predictive control (MPC) uses the prediction model to optimize the future system behavior. The combination of the two enables the control system to respond more intelligently to the external environment and internal state changes, thereby achieving adaptive control of unknown environments.

[0026] Combining reinforcement learning with model predictive control (RL-MPC) algorithm, through continuous trial and error learning, the control strategy can be gradually optimized in actual operation, making the decision of the control system more intelligent and efficient, thereby improving the intelligence level of the control system.

[0027] (IV) After the introduction of the reinforcement learning combined with model predictive control (RL-MPC) algorithm, the control system can predict the optimal control strategy for a period of time in the future based on the real-time collected sensor data and system status. This can avoid factors that may lead to a decline in welding quality, such as temperature changes, air pressure fluctuations, etc., to a certain extent, thereby effectively reducing the welding defect rate. By continuously optimizing the control strategy and combining the reinforcement learning with model predictive control (RL-MPC) algorithm, the welding process can be made more stable, the volatility of welding quality can be reduced, and the stability and consistency of welding quality can be improved.

[0028] (V) The combination of reinforcement learning and model predictive control (RL-MPC) algorithm can dynamically adjust welding parameters and control strategies according to the real-time environment and system status to maximize production efficiency. For example, under the premise of ensuring welding quality, the welding process can be optimized by adjusting parameters such as welding speed and temperature. Since the combination of reinforcement learning and model predictive control (RL-MPC) algorithm can predict future system behavior and optimize it, it can avoid unnecessary shutdowns and adjustments in the production process to a certain extent, thereby improving the utilization rate and production efficiency of the production line. The combination of reinforcement learning and model predictive control (RL-MPC) algorithm can perform real-time optimization according to real-time environmental changes and system dynamics, making the control system more adaptable and robust. Whether it is due to changes in the external environment or fluctuations in system parameters, the RL-MPC algorithm can adjust the control strategy to a certain extent to ensure the stability and performance of the system. After the introduction of the combination of reinforcement learning and model predictive control (RL-MPC) algorithm, the control system has a more intelligent decision-making ability and can better cope with uncertainty and variability, thereby improving the reliability and stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the specific embodiments of the present invention, but do not constitute a limitation of the present invention.

[0030] Figure 1 It is a structural schematic diagram of a welding device for producing plastic oil tanks according to the present invention;

[0031] Figure 2 It is a structural schematic diagram of the back side of a welding device for producing a plastic oil tank according to the present invention;

[0032] Figure 3 It is a structural schematic diagram of the outer side of a gantry frame of a welding device for producing plastic oil tanks according to the present invention;

[0033] Figure 4 It is a structural schematic diagram of a water storage tank and a water spray plate of a welding device for producing a plastic oil tank of the present invention;

[0034] Figure 5 It is a structural schematic diagram of the outer side of the installation groove of a welding device for producing a plastic oil tank according to the present invention;

[0035] Figure 6 It is a structural schematic diagram of the outer side of a fan box and a power box of a welding device for producing a plastic oil tank according to the present invention;

[0036] Figure 7 It is a structural schematic diagram of the outer side of an electric control box of a welding device for producing plastic oil tanks according to the present invention;

[0037] Figure 8 This is the S4 flow chart after combining reinforcement learning and model predictive control (RL-MPC) algorithm

[0038] Numbers in the figure: 1. Base; 2. Support column; 3. Controller; 4. Moving groove; 5. Slide rod; 6. Mounting groove; 7. Threaded pipe; 8. Screw; 9. Positioning cover; 10. L-shaped plate; 11. Electric telescopic rod; 12. Push rod; 13. Oil tank body; 14. Gantry; 15. Connecting plate; 16. Motor; 17. Screw rod; 18. Ring; 19. Moving rod; 20. Slide groove; 21. Electric control box; 22. Welding column; 23. Water storage tank; 24. Display board; 25. Water inlet cover; 26. Extraction pump; 27. Hose; 28. Connecting rod; 29. ​​Water spray plate; 30. Extrusion rod; 31. Fan box; 32. Power box; 33. Elastic switch. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] Example 1 Please refer to Figure 1-Figure 7 , the present invention provides a technical solution:

[0041] A welding device for producing plastic oil tanks includes a base 1, a support column 2 is fixedly connected to the bottom of the base 1, a controller 3 is fixedly connected to the outside of the base 1, a movable groove 4 is opened inside the base 1, a slide rod 5 is slidably connected inside the movable groove 4, a mounting groove 6 is fixedly connected to the top of the slide rod 5, a threaded tube 7 is fixedly connected to the outside of the mounting groove 6, a screw 8 is threadedly connected inside the threaded tube 7, a positioning cover 9 is fixedly connected to the outside of the screw 8, an L-shaped plate 10 is fixedly connected to the outside of the mounting groove 6, an electric telescopic rod 11 is fixedly connected to the outside of the L-shaped plate 10, a push rod 12 is fixedly connected to the outside of the electric telescopic rod 11, and a tank body 13 is installed on the top of the mounting groove 6.

[0042] Specifically, Figure 1 As shown, the support columns 2 are symmetrically distributed at the bottom of the base 1, the mounting groove 6 is slidably connected to the top of the base 1 through the sliding rod 5, the threaded tube 7 is symmetrically distributed on the outside of the mounting groove 6, the positioning cover 9 is rotatably connected to the inside of the mounting groove 6 through the screw 8, the electric telescopic rod 11 is electrically connected to the controller 3, and the end of the push rod 12 away from the electric telescopic rod 11 is fixedly connected to the outside of the mounting groove 6.

[0043] In this embodiment: by setting up the base 1, the mounting groove 6 is slidably connected to the top of the base 1 through the slide bar 5, the fuel tank body 13 is first placed inside the mounting groove 6, and the screw rod 8 is rotated and moved inside the threaded tube 7 by rotating the screw rod 8, and the positioning cover 9 is driven to move by the movement of the screw rod 8. At this time, the positioning cover 9 will make the fuel tank body 13 fixedly installed inside the mounting groove 6. When the position of the fuel tank body 13 needs to be moved, the electric telescopic rod 11 on the outside of the L-shaped plate 10 is started by the controller 3, so that the electric telescopic rod 11 drives the push rod 12 to move, and the movement of the push rod 12 causes the mounting groove 6 to be pushed, so that the mounting groove 6 is moved on the top of the base 1 under the drive of the slide bar 5, and the other side of the fuel tank body 13 can be welded.

[0044] Example 2 Please refer to Figure 1 , Figure 2 and Figure 7 The difference between this embodiment and the embodiment 1 is that: the top of the base 1 is fixedly connected with a gantry 14, the top of the gantry 14 is fixedly connected with a connecting plate 15, the outer side of the connecting plate 15 is fixedly connected with a motor 16, the output end of the motor 16 is fixedly connected with a screw rod 17, the outer side of the screw rod 17 is threadedly connected with a ferrule 18, the bottom of the ferrule 18 is fixedly connected with a moving rod 19, a slide groove 20 is provided inside the gantry 14, the bottom of the moving rod 19 is fixedly connected with an electric control box 21, the bottom of the electric control box 21 is fixedly connected with a welding column 22, and the outer side of the gantry 14 is fixedly connected with a water tank 23, a display panel 24 is fixedly connected to the outside of the water tank 23, a water inlet cover 25 is fixedly connected to the top of the water tank 23, a pumping pump 26 is fixedly connected to the outside of the water tank 23, a hose 27 is fixedly connected to the outside of the pumping pump 26, a connecting rod 28 is fixedly connected to the outside of the welding column 22, a water spray plate 29 is fixedly connected to the outside of the connecting rod 28, an extrusion rod 30 is fixedly connected to the outside of the electric control box 21, a fan box 31 is fixedly connected to the inside of the gantry 14, a power box 32 is fixedly connected to the top of the fan box 31, and an elastic switch 33 is fixedly connected to the top of the power box 32.

[0045] Specifically, Figure 1-7 As shown, the motor 16 is electrically connected to the controller 3, the screw rod 17 is rotatably connected between the two connecting plates 15, the moving rod 19 is slidably connected to the inside of the slide groove 20, the electric control box 21 is electrically connected to the controller 3, the extraction pump 26 is electrically connected to the controller 3, the end of the hose 27 away from the extraction pump 26 is fixedly connected to the inside of the water spray plate 29, and the extrusion rod 30 and the elastic switch 33 are on the same plane.

[0046] In this embodiment: by setting up a base 1, a gantry 14 is fixedly connected to the top of the base 1. When it is necessary to weld the installed oil tank body 13, the controller 3 is used to start the electric control box 21, so that the welding column 22 at the bottom of the electric control box 21 welds the oil tank body 13. At this time, the motor 16 is started to drive the screw rod 17 to rotate between the connecting plates 15, and the rotation of the screw rod 17 drives the ring 18 to rotate. Since a moving rod 19 is fixedly connected to the bottom of the ring 18, and the moving rod 19 is slidably connected to the inside of the slide groove 20, when the screw rod 17 rotates, the ring 18 can drive the electric control box 21 and the welding column 22 at the bottom of the moving rod 19 to move, thereby welding the oil tank body 13. The connecting edge of the box body 13 is welded. When welding, the extraction pump 26 on the outside of the water tank 23 is started through the controller 3, so that the extraction pump 26 extracts the water inside the water tank 23, and then enters the inside of the water spray plate 29 through the hose 27. Since the water spray plate 29 is located on the outside of the welding column 22, the water can be sprayed onto the welding point to cool the position after welding. When the electrical control box 21 drives the welding column 22 to move to the other side of the gantry 14, the squeezing rod 30 will squeeze the elastic switch 33 on the outside of the power box 32, so that the power box 32 drives the fan box 31 to start. At this time, the wind force generated by the start of the fan box 31 will dry the cooled water, thereby achieving the effect of improving welding efficiency.

[0047] An intelligent control method for plastic oil tank production, applied to a welding device for plastic oil tank production as claimed in any one of claims 1 to 8, comprises the following steps:

[0048] S1. First, place the oil tank body 13 inside the installation groove 6, and position the oil tank body 13 by using the screw 8 and the positioning cover 9 in coordination;

[0049] S2. The position of the oil tank body 13 inside the installation groove 6 can be adjusted by using the electric telescopic rod 11 and the push rod 12 in coordination;

[0050] S3, starting the controller 3 and the electric control box 21, so that the welding column 22 performs welding processing on the oil tank body 13 at the bottom;

[0051] S4, the motor 16 drives the screw rod 17 to rotate, so that the moving rod 19 drives the welding column 22 to move on the top of the oil tank body 13;

[0052] S5, extract the water inside the water storage tank 23 through the extraction pump 26, and finally spray it out from the inside of the water spray plate 29, and cool down the welded oil tank body 13 and Xining:

[0053] S6, squeezing the elastic switch 33 by the squeezing rod 30, so that the fan box 31 is started, thereby accelerating the drying of the water after cooling;

[0054] S7. Inspection and acceptance: After processing, the workpiece is manually inspected and accepted.

[0055] S4. In S4, the RL-MPC algorithm combining reinforcement learning and model predictive control is introduced to monitor and optimize various parameters in the plastic tank welding process in real time; the specific process is:

[0056] Step 1: Establish state space and action space:

[0057] 1. Establish state space:

[0058] In order to accurately capture the state changes during the welding process, multiple high-precision sensors are installed in the welding system to monitor the following key parameters in real time:

[0059] Welding column position (x w ,y w ) and the tank body position (x t ,y t ):

[0060] (x w ,y w ) represents the horizontal and vertical coordinates of the welding column, (x t ,y t ) represents the horizontal and vertical coordinates of the fuel tank body, which are used to describe the relative positions of the welding column and the fuel tank body during welding, and are represented by a Cartesian coordinate system; accurately represents the positional relationship between the key components in the welding system, thereby further monitoring and controlling the welding process; the sensor is installed on the welding robot arm to provide real-time feedback on the spatial position of the welding column to ensure the accuracy and consistency of the welding path; the sensor is installed at a key position of the fuel tank to monitor the position change of the fuel tank in real time to avoid welding defects caused by the movement of the fuel tank;

[0061] The position of the welding column and the position of the fuel tank body are offset, resulting in an unstable welding process or welding defects; therefore, the range of the welding column position and the fuel tank body position is determined according to the design of the welding system and the working environment, limiting the movement range of the welding column and the fuel tank body to ensure that the welding process is carried out within a safe and effective range;

[0062] Ambient temperature T env :

[0063] Considering the influence of ambient temperature on the welding process, it is expressed in degrees Celsius (℃); by obtaining real-time data from environmental monitoring equipment and setting it according to historical data and environmental characteristics, the influence of ambient temperature on welding quality and process stability is accurately understood; welding in a high temperature environment causes the welding material to solidify prematurely, while in a low temperature environment, the fluidity of the welding material decreases; temperature sensors are distributed around the welding area to collect ambient temperature data in real time and provide accurate temperature information;

[0064] Ambient humidity H env :

[0065] Consider the influence of ambient humidity on the welding process, expressed as relative humidity (%); obtain real-time data based on environmental monitoring equipment and set according to historical data and environmental characteristics. In a high humidity environment, the welding arc may be affected by humid air and become unstable, thus affecting the welding quality; therefore, monitor humidity changes and adjust welding parameters to ensure the stability and consistency of the welding process; humidity sensors are installed in the welding workshop to monitor the ambient humidity in real time,

[0066] Environmental pressure P env :

[0067] Consider the impact of environmental pressure on the welding process, expressed in Pascal (Pa); obtain real-time data from environmental monitoring equipment and set according to historical data and environmental characteristics. In a high-pressure environment, the welding arc may be affected by additional gas pressure and become more stable, thereby improving welding quality; adjust welding parameters according to real-time pressure data, and install pressure sensors in the welding environment to provide real-time feedback on environmental pressure changes.

[0068] 2. Establish the action space:

[0069] The motor drives the screw to rotate at an angle θ motor :

[0070] Indicates the angle at which the motor drives the screw to rotate, usually expressed in degrees (°), which controls the movement of the welding column; the angle determines the position of the welding column relative to the tank body, affecting the accuracy and speed of welding. The range of the angle at which the motor drives the screw to rotate depends on the design of the welding system and the rotation range of the motor. Parameters are usually adjusted and set during the experimental stage;

[0071] Welding speed V weld :

[0072] Indicates the speed at which the welding column or welding head moves during the welding process, expressed in millimeters per second (mm / s); welding speed directly affects the efficiency and quality of welding; when welding large workpieces, higher welding speeds can improve production efficiency, but may affect welding quality; by setting according to welding materials and process requirements, adjusting and optimizing through experiments and simulations; then evaluating the welding quality, and finally determining the optimal welding speed;

[0073] Welding pressure F weld :

[0074] Indicates the welding pressure applied during welding, expressed in Newton (N); the size of welding pressure directly affects the quality and strength of the welded joint; when welding metal workpieces, appropriate welding pressure can ensure the fullness and firmness of the weld; the range of welding pressure is set according to the properties of the welding materials and process requirements to ensure that the pressure applied during welding meets the requirements; for different types of metal materials, the most suitable welding pressure is determined according to their hardness and plasticity; this ensures that the pressure applied during welding can meet the requirements of welding quality and process stability;

[0075] Soldering temperature T weld :

[0076] Indicates the welding temperature during the welding process, expressed in degrees Celsius (℃); welding temperature is an important parameter that controls the melting and flow of materials during the welding process; for different types of plastic materials, the most suitable welding temperature is determined according to their melting point and thermal sensitivity. It is necessary to ensure that the welding temperature is higher than the melting point of the material to ensure that the welding material can be fully melted and form a strong connection; the welding temperature can be set according to the melting point of the welding material and the process requirements to ensure that the temperature during the welding process is controlled within an appropriate range;

[0077] Step 2: Build the model

[0078] In the welding process, a recurrent neural network (RNN) is used as a state transition model to capture the nonlinear and time-varying characteristics of the system. RNN has the advantage of processing time series data and can describe the dynamic evolution of the state during the welding process, thus providing accurate predictions for optimal control.

[0079] Model training: The data from the sensors in step 1 are transmitted to the data processing center in real time via a wireless network or wired connection to form a historical data set; this data set is used to train the RNN model to capture and understand the state transition rules in the welding process; the RNN model is trained using the collected data;

[0080] The current status is:

[0081] s t =[x w,t ,yw,t , x t,t ,y t,t , T env,t , H env,t , P env,t ]

[0082] The historical data is: (s t-n ,s t-n+1 , ..., s t ) Input RNN and train the model to learn the law of state transition; historical data includes not only the state at the current moment, but also the state at several previous moments. Through these data, the law of state change over time can be captured;

[0083] Among them, s t : actual state value, indicating the state at the current time step; s i : represents the i-th group state vector; x w,t represents the horizontal coordinate position of the welding column at time t; w,t represents the ordinate position of the welding column at time t; x t,t represents the horizontal coordinate position of the fuel tank body at time t; y t,t : Indicates the vertical coordinate position of the fuel tank body; T env,t : represents the ambient temperature at time t; H env,t represents the ambient humidity at time t; P env,t represents the environmental pressure at time t;

[0084] The state transition formula is:

[0085] h t =tanh(W h ·h t-1 +U h ·s t +b h )

[0086] s` t+1 =V·h t +c

[0087] in:

[0088] h t is the hidden state, which represents the state of RNN at time t; the hidden state is the core part of RNN, which captures the dependencies in the time series through continuous updating;

[0089] Tanh is an activation function used to introduce nonlinearity. Tanh can limit the output range to [-1, 1], which is more suitable for processing certain types of time series data.

[0090] W his the weight matrix, which represents the parameters of the model; the weight matrix is ​​adjusted through training, and finally the model can accurately describe the state transition process;

[0091] U h is the weight matrix, which represents the parameters of the model; the weight matrix is ​​adjusted through training, and finally the model can accurately describe the state transition process;

[0092] V is the weight matrix, which represents the parameters of the model. The weight matrix is ​​adjusted through training, and ultimately the model can accurately describe the state transition process.

[0093] b h The bias term adjusts the model output; the bias term provides an additional degree of freedom, allowing the model to fit the data more flexibly;

[0094] c is the bias term, which adjusts the model output; the bias term provides an additional degree of freedom, allowing the model to fit the data more flexibly;

[0095] s t : actual state value, indicating the state at the current time step;

[0096] During the training process, the Adam optimization algorithm is used to minimize the loss function to adjust the weights and bias terms; the loss function is defined as:

[0097]

[0098] L: loss function, which represents the difference between the model prediction value and the actual value;

[0099] T: time step, indicating the total number of steps in the welding process;

[0100] ∑: summation symbol, indicating the cumulative summation of the losses of all time steps;

[0101] t: time step index, indicating the current time step;

[0102] s t+1 : The actual state value, indicating the next state after the current time step;

[0103] s` t+1 : The state value predicted by the model, indicating the next state predicted by the model after the current time step;

[0104] The update rules of the Adam algorithm are as follows:

[0105] m t =β1·m t-1 +(1-β1)·g t

[0106]

[0107] Among them, g t is the gradient of the current time step; m t : Moving average of gradient; v t : Moving average of the square of the gradient; β1: exponential decay rate of the first-order moment estimate; β2: exponential decay rate of the second-order moment estimate; The exponential decay rate of the first-order moment estimate to the power of t; The exponential decay rate of the second-order moment estimate to the power of t; θ t is the parameter of the current time step; is the corrected estimate of the first and second moments of the gradient; m t : First-order moment estimate, representing the exponential moving average of the gradient; v t : Second-order moment estimate, representing the exponential moving average of the square of the gradient; a is the coefficient;

[0108] By minimizing the loss function, the RNN model can more accurately predict the state changes in the welding process. Through the Adam optimization algorithm, the weights and bias terms of the model are continuously adjusted during the training process to minimize the loss function, so that the RNN model can more accurately predict the state changes in the welding process, and converge faster and remain stable during the training process.

[0109] After the model training is completed, the prediction accuracy and generalization ability of the model need to be verified; the specific steps include:

[0110] Select a portion of data from the actual welding process as a validation set to ensure that these data are not involved in model training; this can test the performance of the model on unseen data and evaluate its generalization ability;

[0111] Use the trained RNN model to predict the validation set and compare the predicted results with the actual data; by comparing the predicted values ​​and the true values, the accuracy of the model can be evaluated;

[0112] Calculate the prediction error to evaluate the performance of the model; when the error is small, it means that the model has good prediction and generalization capabilities; commonly used evaluation indicators include mean square error (MSE), mean absolute error (MAE), etc.

[0113] If the prediction accuracy of the model is not high enough, it is necessary to further adjust the model parameters, retrain and verify until the model performance reaches the expected standard; optimize the model by adjusting the hyperparameter learning rate, hidden layer size, changing the network structure and increasing the number of layers;

[0114] Reward Function

[0115] In order to optimize the welding process, a reward function that comprehensively considers welding quality, energy consumption and production efficiency is designed to achieve multi-objective optimization. During the welding process, the corresponding reward value R is calculated after each welding action is completed. By continuously adjusting the welding parameter motor rotation angle θ motor , welding speed V weld , welding pressure F weld , welding temperature T weld , maximize the reward function, and thus select the optimal welding strategy; according to the current reward value and historical data, adjust the welding speed V weld , welding pressure F weld Soldering temperature T weld and the motor rotation angle θ motor , to optimize welding quality and efficiency; by continuously adjusting parameters, find the welding strategy that maximizes the reward value;

[0116] Reward function design

[0117] Assuming welding quality is Q, energy consumption is E, and production efficiency is P, the reward function R is defined as:

[0118] R=w q Q+w e (1-E)+w p ·P

[0119] in:

[0120] w q w e w p is a weight parameter that reflects the importance of each indicator; by adjusting these weight parameters, the relationship between different optimization objectives can be balanced;

[0121] w q :Indicates the importance of welding quality. For welding tasks with high quality requirements,

[0122] w e : Indicates the importance of energy consumption. For tasks with low energy consumption requirements,

[0123] w p : Indicates the importance of production efficiency. For tasks requiring high efficiency,

[0124] Q stands for welding quality index, which can be evaluated based on the strength and tightness of the weld; the weld can be evaluated by non-destructive testing methods, ultrasonic testing and X-ray testing;

[0125] E is the energy consumption index, which is calculated based on the energy consumed during the actual welding process; this index can be obtained by monitoring the energy consumption of the welding equipment;

[0126] P is the production efficiency index, which is based on the consideration of welding time and work efficiency; it can be calculated by recording the start and end time of welding and the welding speed;

[0127] These weight parameters can be optimized through experiments and simulations to find the most suitable parameter combination for a specific welding task; by adjusting the weight parameters through multiple experiments and simulations, the optimal weight combination can be found so that the welding process achieves the best balance between different objectives;

[0128] Step 3: Combine reinforcement learning and model predictive control (RL-MPC) algorithm application:

[0129] Reinforcement Learning (RL): Use genetic algorithms to better handle multi-objective optimization problems and find better control strategies;

[0130] Define parameter range: During welding, determine the welding speed V weld and welding pressure F weld The scope is crucial;

[0131] Initialize the population: Initially, generate a certain number of individuals, which represent various combinations of welding parameters; randomly generate N different individuals, each of which contains a set of random V weld and F weld ;

[0132] In the combined reinforcement learning and model predictive control (RL-MPC) algorithm, the design of the fitness function is crucial because it determines the pros and cons of each individual. The fitness function is designed based on the reward function, which can be expressed as:

[0133] R=w q Q+w e (1-E)+w p ·P

[0134] In the selection operation, the roulette wheel selection method is used to select the next generation of individuals; specifically, the probability of each individual being selected is proportional to its fitness value; this can be expressed by the following formula:

[0135]

[0136] in

[0137] P(i) is the probability that individual i is selected,

[0138] f(i) is the fitness value of individual i, f(j) is the fitness value of individual j,

[0139] N is the number of individuals in the population;

[0140] In genetic algorithms, crossover and mutation operations are key steps in generating new individuals; the crossover operation simulates gene recombination, while the mutation operation introduces diversity in the population; these operations help maintain the diversity of the population and help avoid local optimal solutions;

[0141] Once new individuals are generated, they will replace some of the old individuals to form a new generation of population;

[0142] Through repeated selection, crossover and mutation operations, individuals in the population gradually approach the optimal solution; in the iterative process, an iteration termination condition is set, and the iteration is stopped when a certain number of iterations or a certain fitness threshold is reached.

[0143] Model Predictive Control (MPC): In model predictive control, nonlinear optimization technology and constrained optimization methods are introduced to solve the control problems of nonlinear systems and improve the accuracy and stability of control;

[0144] Consider a plastic oil tank welding system, in which the moving rod is connected to the motor through a screw. The rotation of the motor drives the screw to rotate, so that the moving rod drives the welding column to move on the top of the oil tank body. The dynamic equation of the system can be modeled as:

[0145] m·x``(t)=F motor (t)-F fricyion (x`(t))

[0146] Where m is the mass of the moving rod; x(t) is the position of the moving rod, x`(t) and x``(t) are the velocity and acceleration of the moving rod, respectively; F motor (t) is the torque of the motor, F fricyion is the friction force;

[0147] Optimization goal:

[0148] The optimization goal is to minimize the error between the weld column position and the desired position:

[0149]

[0150] in,

[0151] x desired (t) is the desired weld column position;

[0152] Nonlinear optimization algorithms:

[0153] The gradient descent method is used to minimize the objective function, and the welding column position is controlled by adjusting the speed and direction of the motor; the update rule of the gradient descent method is:

[0154]

[0155] in,

[0156] α is the learning rate, which is used to control the update step size;

[0157] Apply constrained optimization methods:

[0158] Considering the constraints of motor speed and direction, the maximum torque τ of the motor max and maximum speed ω max , and the maximum speed of the moving rod ω max and displacement range; these constraints can be expressed as:

[0159] -τ max ≤F motor (t)≤τ max

[0160] -ω max ≤x`(t)≤ω max

[0161] Constrained optimization algorithms can be processed using the Lagrange multiplier method, constructing the Lagrangian function, and solving the optimal solution using the gradient descent method;

[0162] The Lagrange multiplier method is a method for dealing with constrained optimization problems. By introducing Lagrange multipliers, the constraints are integrated into the objective function, and a Lagrange function containing the original objective function and the constraints is constructed. Specifically, the following optimization problem is set:

[0163] Minimize the objective function: J(x)

[0164] Satisfy a series of constraints: g i (x)≤0, i=1, 2, …, m

[0165] Where x is the optimization variable vector, J(x) is the objective function, and g i (x) is the constraint function;

[0166] According to the Lagrange multiplier method, the Lagrange function is constructed as follows:

[0167]

[0168] Among them, λ i is the Lagrange multiplier;

[0169] By taking the partial derivative of the Lagrangian function and setting it to zero, we can obtain a set of equations about the optimization variable x and the Lagrangian multiplier λ in the following form:

[0170]

[0171] g i(t)≤0, i=1, 2, …, m

[0172] in, is the gradient of the Lagrangian function with respect to the variable x;

[0173] Finally, by solving the above equations, we can obtain the optimal solution x* and the corresponding optimal value J(x*);

[0174] Step 4: Real-time control

[0175] In the plastic tank welding process, real-time control is crucial to ensure welding quality and production efficiency. In order to achieve high-precision and high-stability control goals, adaptive control methods and model parameter online update technology are introduced to cope with environmental changes and system dynamic uncertainties, thereby improving the robustness and adaptability of the control system.

[0176] Adaptive control method

[0177] The adaptive control method responds to changes in system parameters and environmental uncertainties by adjusting control parameters online. The specific steps are as follows:

[0178] System parameter identification: By collecting sensor data in real time, the dynamic parameters of the welding system, such as motor torque and friction coefficient, are estimated;

[0179] Controller adjustment: Based on the real-time estimated system parameters, the controller parameters are adjusted online to maintain the system stability and control accuracy;

[0180] System dynamics model:

[0181] m·x``(t)=F motor (t)-F fricyion (x`(t))

[0182] Real-time identification of friction coefficient k f and motor gain k m :

[0183] F fricyion (x`(t))=k f ·x f (t)

[0184] F motor (t) = k m u(t)

[0185] Controller online adjustment formula:

[0186]

[0187] u(t): control input voltage or current of the motor;

[0188] Model parameter online update technology

[0189] The model parameter online update technology is used to dynamically update the system model during the control process to reflect the actual status of the current system; this can be achieved through the following steps:

[0190] Data collection: Install sensors to monitor various parameters in the plastic tank welding process in real time, including temperature, pressure, flow, displacement, and speed;

[0191] Model update: Based on the real-time collected data, the parameters of the system model are updated using the online learning algorithm recursive least squares method;

[0192] Assume the dynamic model is

[0193] x``(t)=a·x(t)+b·x`(t)+c·u(t)

[0194] Where, a, b, c: parameters of the system model;

[0195] Update parameters a, b, c in real time:

[0196] a`(t+1)=a`(t)+γ·(x(t)-a`(t)·x(t-1))

[0197] b`(t+1)=b`(t)+γ·[x(t)-b`(t)·x(t-1))

[0198] c`(t+1)=c`(t)+γ·(x(t)-c`(t)·x(t-1))

[0199] in,

[0200] a`(t), b`(t), c`(t): parameter estimates at time t

[0201] x(t): the position of the moving rod

[0202] x`(t): speed of the moving rod

[0203] u(t): control input

[0204] γ is the learning rate, which controls the step size of parameter updates.

[0205] Although 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 the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A welding device for producing a plastic oil tank, comprising a base (1), Features: The bottom of the base (1) is fixedly connected to a support column (2), the outer side of the base (1) is fixedly connected to a controller (3), the interior of the base (1) is provided with a movable groove (4), the interior of the movable groove (4) is slidably connected to a slide rod (5), the top of the slide rod (5) is fixedly connected to a mounting groove (6), the outer side of the mounting groove (6) is fixedly connected to a threaded tube (7), the interior of the threaded tube (7) is threadedly connected to a screw rod (8), the outer side of the screw rod (8) is fixedly connected to a positioning cover (9), the outer side of the mounting groove (6) is fixedly connected to an L-shaped plate (10), the outer side of the L-shaped plate (10) is fixedly connected to an electric telescopic rod (11), the outer side of the electric telescopic rod (11) is fixedly connected to a push rod (12), and the top of the mounting groove (6) is installed with an oil tank body (13).

2. A welding device for producing plastic fuel tanks according to claim 1, Features: The top of the base (1) is fixedly connected to a gantry (14), the top of the gantry (14) is fixedly connected to a connecting plate (15), the outer side of the connecting plate (15) is fixedly connected to a motor (16), the output end of the motor (16) is fixedly connected to a screw rod (17), the outer side of the screw rod (17) is threadedly connected to a ferrule (18), the bottom of the ferrule (18) is fixedly connected to a moving rod (19), a sliding groove (20) is provided inside the gantry (14), the bottom of the moving rod (19) is fixedly connected to an electric control box (21), the bottom of the electric control box (21) is fixedly connected to a welding column (22), the outer side of the gantry (14) is fixedly connected to a water storage tank (23), the water storage tank The outer side of the water storage tank (23) is fixedly connected with a display panel (24), the top of the water storage tank (23) is fixedly connected with a water inlet cover (25), the outer side of the water storage tank (23) is fixedly connected with an extraction pump (26), the outer side of the extraction pump (26) is fixedly connected with a hose (27), the outer side of the welding column (22) is fixedly connected with a connecting rod (28), the outer side of the connecting rod (28) is fixedly connected with a water spray plate (29), the outer side of the electric control box (21) is fixedly connected with an extrusion rod (30), the inner side of the gantry (14) is fixedly connected with a fan box (31), the top of the fan box (31) is fixedly connected with a power supply box (32), and the top of the power supply box (32) is fixedly connected with an elastic switch (33).

3. A welding device for producing plastic fuel tanks according to claim 1, Features: The support columns (2) are symmetrically distributed at the bottom of the base (1), and the mounting grooves (6) are slidably connected to the top of the base (1) via sliding rods (5).

4. A welding device for producing plastic fuel tanks according to claim 1, Features: The threaded tubes (7) are symmetrically distributed on the outside of the installation groove (6), and the positioning cover (9) is rotatably connected to the inside of the installation groove (6) via a screw rod (8).

5. A welding device for producing plastic fuel tanks according to claim 1, Features: The electric telescopic rod (11) is electrically connected to the controller (3), and one end of the push rod (12) away from the electric telescopic rod (11) is fixedly connected to the outside of the mounting groove (6).

6. A welding device for producing plastic fuel tanks according to claim 2, Features: The motor (16) is electrically connected to the controller (3), and the screw rod (17) is rotatably connected between the two connecting plates (15).

7. A welding device for producing plastic fuel tanks according to claim 2, Features: The moving rod (19) is slidably connected inside the sliding groove (20), and the electric control box (21) is electrically connected to the controller (3).

8. A welding device for producing plastic fuel tanks according to claim 2, Features: The extraction pump (26) is electrically connected to the controller (3), one end of the hose (27) away from the extraction pump (26) is fixedly connected to the inside of the water spray plate (29), and the squeezing rod (30) and the elastic switch (33) are on the same plane.

9. The intelligent control method of a welding device for producing a plastic oil tank according to claim 1, Features: A welding device for producing a plastic oil tank as described in any one of claims 1 to 8, comprising the following steps: S1. First, the oil tank body (13) is placed inside the installation groove (6), and the oil tank body (13) is positioned by using the screw rod (8) and the positioning cover (9); S2, the position of the oil tank body (13) inside the installation groove (6) can be adjusted by using the electric telescopic rod (11) and the push rod (12) in coordination; S3, starting the controller (3) and the electric control box (21), thereby causing the welding column (22) to weld the oil tank body (13) at the bottom thereof; S4, the motor (16) drives the screw rod (17) to rotate, so that the moving rod (19) drives the welding column (22) to move on the top of the oil tank body (13); S5. The water inside the water storage tank (23) is extracted by the extraction pump (26), and finally sprayed out from the inside of the water spray plate (29), and the welded oil tank body (13) and the cooling treatment are performed: S6, squeezing the elastic switch (33) by the squeezing rod (30) to start the fan box (31), thereby accelerating the drying of the water after the temperature is reduced; S7. Inspection and acceptance: After processing, the workpiece is manually inspected and accepted.

10. The intelligent control method of a welding device for producing a plastic oil tank according to claim 9, It is characterized in that In S4, the RL-MPC algorithm combining reinforcement learning and model predictive control is introduced to monitor and optimize various parameters in the plastic tank welding process in real time; the specific process is as follows: Step 1: Establish state space and action space:

1. Establish state space: In order to accurately capture the state changes during the welding process, multiple high-precision sensors are installed in the welding system to monitor the following key parameters in real time: Welding column position (x w ,y w ) and the tank body position (x t ,y t ): (x w ,y w ) represents the horizontal and vertical coordinates of the welding column, (x t ,y t ) represents the horizontal and vertical coordinates of the fuel tank body, which is used to describe the relative position of the welding column and the fuel tank body during welding, and is represented by a Cartesian coordinate system; accurately represents the positional relationship between the key components in the welding system, thereby further monitoring and controlling the welding process; the sensor is installed on the welding robot arm to provide real-time feedback on the spatial position of the welding column to ensure the accuracy and consistency of the welding path; the sensor is installed at a key position of the fuel tank to monitor the position change of the fuel tank in real time to avoid welding defects caused by the movement of the fuel tank; Ambient temperature T env : Consider the impact of ambient temperature on the welding process, obtain real-time data through environmental monitoring equipment and set it according to historical data and environmental characteristics; Ambient humidity H env :Consider the impact of ambient humidity on the welding process; obtain real-time data from environmental monitoring equipment and set it based on historical data and environmental characteristics; Environmental pressure P env : Consider the impact of environmental pressure on the welding process, obtain real-time data from environmental monitoring equipment and set it according to historical data and environmental characteristics; 2. Establish the action space: The motor drives the screw to rotate at an angle θ motor : Indicates the angle at which the motor drives the screw to rotate, controlling the movement of the welding column; the angle determines the position of the welding column relative to the tank body, affecting the accuracy and speed of welding. The range of the angle at which the motor drives the screw to rotate depends on the design of the welding system and the rotation range of the motor. Parameters are usually adjusted and set during the experimental stage. Welding speed Vwel d : Indicates the speed at which the welding column or welding head moves during the welding process. The welding speed directly affects the efficiency and quality of welding; Welding pressure F weld : Indicates the welding pressure applied during the welding process. The size of the welding pressure directly affects the quality and strength of the welded joint; Soldering temperature T weld : Indicates the welding temperature during the welding process. The welding temperature is an important parameter to control the melting and flow of materials during the welding process; Step 2: Build the model In the welding process, a recurrent neural network (RNN) is used as the state transfer model to capture the nonlinear and time-varying characteristics of the system. RNN has the advantage of processing time series data and can describe the dynamic evolution of the state during the welding process, thereby providing accurate predictions for optimal control. Model training: The data from the sensors in step 1 are transmitted to the data processing center in real time via a wireless network or wired connection to form a historical data set; this data set is used to train the RNN model to capture and understand the state transition rules in the welding process; the RNN model is trained using the collected data; The current status is: s t =[x w,t ,y w,t ,x t,t ,y t,t ,T env,t ,H env,t ,P env,t ] The historical data is: (s t-n ,s t-n+1 , ..., s t ) Input RNN and train the model to learn the law of state transition; historical data includes not only the state at the current moment, but also the state at several previous moments. Through these data, the law of state change over time can be captured; Among them, t: actual state value, indicating the state at the current time step; i: indicates the i-th group state vector; x w,t represents the horizontal coordinate position of the welding column at time t; w,t Indicates the ordinate position of the welding column at time t; x t,t represents the horizontal coordinate position of the fuel tank body at time t; y t,t : Indicates the vertical coordinate position of the fuel tank body; T env,t : represents the ambient temperature at time t; H env,t represents the ambient humidity at time t; P env,t represents the environmental pressure at time t; The state transition formula is: h t =tanh(W h ·h t-1 +U h ·S t +b h ) s` t+1 =V·h t +c Where: h t is the hidden state, indicating the state of RNN at time t; W h is the weight matrix, representing the parameters of the model; U h is the weight matrix, which represents the parameters of the model; V is the weight matrix, which represents the parameters of the model; b h is the bias term, which adjusts the model output; c is the bias term, which adjusts the model output; during the training process, the Adam optimization algorithm is used to minimize the loss function to adjust the weight and bias term; the loss function is defined as: L: loss function, which represents the difference between the model prediction value and the actual value; T: time step, indicating the total number of steps in the welding process; t: time step index, indicating the current time step; s t+1 : The actual state value, indicating the next state after the current time step; s` t+1 : The state value predicted by the model, indicating the next state predicted by the model after the current time step; The update rules of the Adam algorithm are as follows: m t =b 1 ·m t-1 +(1-β 1 )·g t Among them, g t is the gradient of the current time step; m t : Moving average of gradient; v t : Moving average of the squared gradient; β 1 : exponential decay rate of the first-order moment estimate; β 2 : exponential decay rate of the second-order moment estimate; The exponential decay rate of the first-order moment estimate raised to the power of t; The exponential decay rate of the second-order moment estimate to the power of t; θ t is the parameter of the current time step; is the corrected estimate of the first and second moments of the gradient; t : The first-order moment estimate, which represents the exponential moving average of the gradient; t: The second-order moment estimate, which represents the exponential moving average of the square of the gradient; a is the coefficient; Reward Function In order to optimize the welding process, a reward function that comprehensively considers welding quality, energy consumption and production efficiency is designed to achieve multi-objective optimization. During the welding process, the corresponding reward value R is calculated after each welding action is completed. By continuously adjusting the welding parameters, the motor rotation angle θ m o tor , welding speed Vwe ld , welding pressure Fwe ld , welding temperature T weld , maximize the reward function, and thus select the optimal welding strategy; according to the current reward value and historical data, adjust the welding speed V weld , welding pressure F weld Soldering temperature T weld and the motor rotation angle θ motor , to optimize welding quality and efficiency; by continuously adjusting parameters, find the welding strategy that maximizes the reward value; Reward function design Assuming welding quality is Q, energy consumption is E, and production efficiency is P, the reward function R is defined as: R=w q ·Q+w e ·(1-E)+w p ·P in: w q w e w p is a weight parameter that reflects the importance of each indicator; by adjusting these weight parameters, the relationship between different optimization objectives can be balanced; w q :Indicates the importance of welding quality. For welding tasks with high quality requirements, w e : Indicates the importance of energy consumption. For tasks with low energy consumption requirements, w p : Indicates the importance of production efficiency. For tasks requiring high efficiency, Q stands for welding quality index, which can be evaluated based on the strength and tightness of the weld; the weld can be evaluated by non-destructive testing methods, ultrasonic testing and X-ray testing; E is the energy consumption index, which is calculated based on the energy consumed during the actual welding process; this index can be obtained by monitoring the energy consumption of the welding equipment; P is the production efficiency index, which is based on the consideration of welding time and work efficiency; it can be calculated by recording the start and end time of welding and the welding speed; Step 3: Combine reinforcement learning with model predictive control RL-MPC algorithm application: Reinforcement Learning (RL): Use genetic algorithms to better handle multi-objective optimization problems and find better control strategies; Define parameter range: During welding, determine the welding speed V weld and welding pressure F weld The scope is crucial; Initialize the population: Initially, generate a certain number of individuals, which represent various combinations of welding parameters; randomly generate N different individuals, each of which contains a set of random V weld and F weld ; In the RL-MPC algorithm that combines reinforcement learning and model predictive control, the design of the fitness function is crucial because it determines the quality of each individual. The fitness function is designed based on the reward function, which can be expressed as: R=w q ·Q+w e ·(1-E)+w p ·P In the selection operation, the roulette wheel selection method is used to select the next generation of individuals; specifically, the probability of each individual being selected is proportional to its fitness value; this can be expressed by the following formula: Among them, P(i) is the probability that individual i is selected, f(i) is the fitness value of individual i, f(j) is the fitness value of individual j, and N is the number of individuals in the population; In genetic algorithms, crossover and mutation operations are key steps in generating new individuals; the crossover operation simulates the recombination of genes, while the mutation operation introduces diversity in the population; these operations help maintain the diversity of the population and help avoid local optimal solutions; Once new individuals are generated, they will replace some of the old individuals to form a new generation of population; Through repeated selection, crossover and mutation operations, individuals in the population gradually approach the optimal solution; in the iterative process, an iteration termination condition is set, and the iteration is stopped when a certain number of iterations or a certain fitness threshold is reached; Model Predictive Control (MPC): In model predictive control, nonlinear optimization technology and constrained optimization methods are introduced to solve the control problems of nonlinear systems and improve the accuracy and stability of control; Consider a welding system for a plastic oil tank, in which the moving rod is connected to the motor through a screw. The rotation of the motor drives the screw to rotate, so that the moving rod drives the welding column to move on the top of the tank body. The dynamic equation of the system can be modeled as: m·x``(t)=F motor (t)-F fricyion (x(t)) Where m is the mass of the moving rod; x(t) is the position of the moving rod, x`(t) and x``(t) are the velocity and acceleration of the moving rod, respectively; F motor (t) is the torque of the motor, F fricyion is the friction force; Optimization goal: The optimization goal is to minimize the error between the weld column position and the desired position: Among them, x desired (t) is the desired weld column position; Nonlinear optimization algorithms: The gradient descent method is used to minimize the objective function, and the welding column position is controlled by adjusting the speed and direction of the motor; the update rule of the gradient descent method is: Among them, α is the learning rate, which is used to control the update step size; Apply constrained optimization methods: Considering the constraints of motor speed and direction, the maximum torque τ of the motor max and maximum speed ω max , and the maximum speed of the moving rod ω max and displacement range; these constraints can be expressed as: -t max ≤F motor (t)≤τ max -ω≤x`(t)≤ω max Constrained optimization algorithms can be processed using the Lagrange multiplier method, constructing the Lagrangian function, and solving the optimal solution using the gradient descent method; The Lagrange multiplier method is a method for dealing with constrained optimization problems. By introducing Lagrange multipliers, the constraints are integrated into the objective function, and a Lagrange function containing the original objective function and the constraints is constructed. Specifically, the following optimization problem is set: Minimize the objective function: J(x) Satisfy a series of constraints: g i (x)≤0, i=1, 2, …, m Where x is the optimization variable vector, J(x) is the objective function, and g i (x) is the constraint function; According to the Lagrange multiplier method, the Lagrange function is constructed as follows: Among them, λ i is the Lagrange multiplier; By taking the partial derivative of the Lagrangian function and setting it to zero, we can obtain a set of equations about the optimization variable x and the Lagrangian multiplier λ in the following form: g i (x)≤0,i=1,2,…,m in, is the gradient of the Lagrangian function with respect to the variable x; Finally, by solving the above equations, we can obtain the optimal solution x* and the corresponding optimal value J(x*); Step 4: Real-time control In the plastic tank welding process, real-time control is crucial to ensure welding quality and production efficiency. In order to achieve high-precision and high-stability control goals, adaptive control methods and model parameter online update technology are introduced to cope with environmental changes and system dynamic uncertainties, thereby improving the robustness and adaptability of the control system. Adaptive control method The adaptive control method responds to changes in system parameters and environmental uncertainties by adjusting control parameters online. The specific steps are as follows: System parameter identification: By collecting sensor data in real time, the dynamic parameters of the welding system, such as motor torque and friction coefficient, are estimated; Controller adjustment: Based on the real-time estimated system parameters, the controller parameters are adjusted online to maintain the system stability and control accuracy; System dynamics model: m·x``(t)=F motor (t)-F fricyion (x`(t)) Real-time identification of friction coefficient k f and motor gain k m : F fricyion (x`(t))=k f ·x f (t) F motor (t)=k m ·u(t) Controller online adjustment formula: Where, u(t): input voltage or current of the control input motor; Model parameter online update technology The model parameter online update technology is used to dynamically update the system model during the control process to reflect the actual status of the current system; this can be achieved through the following steps: Data collection: Install sensors to monitor various parameters in the plastic tank welding process in real time, including temperature, pressure, flow, displacement, and speed; Model update: Based on the real-time collected data, the parameters of the system model are updated using the online learning algorithm recursive least squares method; Assume the dynamic model is x``(t)=a·x(t)+b·x`(t)+c·u(t) Where, a, b, c: parameters of the system model; Update parameters a, b, c in real time: a`(t+1)=a`(t)+γ·(x(t)-a`(t)·x(t-1)) b`(t+1)=b`(t)+γ·(x(t)-b`(t)·x(t-1)) c`(t+1)=c`(t)+γ·(x(t)-c`(t)·x(t-1)) Among them, a`(t), b`(t), c`(t): parameter estimates at time t; γ is the learning rate, which controls the step size of parameter update.

Citation Information

Patent Citations

  • Welding method based on artificial intelligence and welding robot

    CN118237825A

  • Polytetrafluoroethylene tube plate welding device

    CN213472229U

  • An ultrasonic welding device for plastic welding

    CN218804074U

  • Corrugated pipe steel belt welding device

    CN220073755U

Cited By

  • Process parameter setting system and method for engineering equipment

    CN119987205A