0 Introduction
Welding is a material joining technology that uses a physicochemical process to connect separated materials through intermolecular forces. With the continuous development of welding technology, its application in production has become increasingly widespread, and it has now become an important processing method. From the production of everyday consumer goods such as household appliances and plumbing equipment to cutting-edge technological products such as airplanes, submarines, rockets, and spacecraft, all rely on efficient and modern welding technology. Further improving welding quality, working conditions, and labor productivity has become a strong desire for all welding workers, and adopting automatic control technology is the correct way to achieve these goals. Research on automatic weld seam tracking systems is an important aspect of the welding field. For precise automatic welding, automatic weld seam tracking is essential.
International research on real-time welding tracking control mainly focuses on the selection of welding sensors and the improvement of control methods. In 1985, D. Lakov of Bulgaria proposed using a fuzzy model to describe the uncertainty of the arc welding process. With the help of a configured non-contact laser sensor, fuzzy control inference was used to estimate, predict, and control the motion of the teach-in robot, achieving automatic weld tracking. In 1989, S. Mursaami et al. of Japan studied fuzzy control for weld tracking in arc welding robots using arc sensing. This control system judged the horizontal and vertical displacement of the weld point based on the relationship between the amplitude and position of the welding torch and the distance between the welding wire and the workpiece. Under strong arc light, high temperature, and fumes, a fuzzy filter and fuzzy controller based on language rules were used to design a weld tracking control system, achieving good results. Motoman Corporation of Ohio, USA, launched an arc sensing tracking system capable of weld tracking at welding speeds up to 60 in_/min (152 cm/min).
Research on weld seam tracking and control technology in my country started relatively late. In the late 1980s, a research group led by Academician Pan Jiluan of Tsinghua University conducted extensive research on rotating arc sensors and achieved valuable results. In 1993, based on the research of his doctoral student Fei Yuejia, Liao Baojian, a PhD candidate at Tsinghua University, successfully developed a hollow shaft motor-driven rotating scanning sensor and obtained a national patent. Subsequently, Jiangxi University conducted in-depth research on miniaturization and vibration reduction, made further improvements, and manufactured a prototype. Wu Shide's doctoral dissertation at Tsinghua University systematically studied arc sensor information processing technology, conducted frequency domain characteristic analysis of scanning arc sensor signals through spatial transformation, and proposed a signal processing method for arc sensing using characteristic filtering vectors.
1. System Overall Indicator Analysis
As an industrial welding robot, the required specifications are first analyzed and defined when designing the entire system.
1.1 System Static Indicators
For welding robot systems, the static performance index refers to the robot's ability to track commands from the control system without deviation when in normal welding operation, reaching a steady state. For step and velocity signals, considering the characteristics of the mechanical structure, the steady-state error must be controlled within 0.2 %, and for acceleration signals, the steady-state error must not be less than 0.5 %.
1.2 System Dynamic Indicators
For welding robots, the dynamic performance indicators of the system mainly consider the dynamic response characteristics of the entire system to step input signals and sinusoidal input signals. For step signals, the overshoot should not exceed 10%, the transition time should not exceed 0.1 ms, and the number of oscillations should not exceed 2. For sinusoidal responses, the main considerations are its frequency characteristics and phase margin, with an operating frequency of approximately 5 to 12 Hz.
1.3 Motion accuracy indicators
The positional accuracy of the welding torch movement is less than 0.2 mm, the weld error is less than 1 mm, the speed accuracy is less than 1 mm/s, and the system response time is less than 0.3 s.
1.4 Analysis of Intelligence Indicators
The designed system is required to have a certain degree of intelligence, including system self-testing function, fault attempt and self-repair function, fault protection function, and automatic tracking and tracing function of weld seam.
1.5 Scalability Indicator Analysis
The system should have certain external interfaces to support both online and offline programming. With the development of internet and IoT technologies, specific interface modules are required. The system must also be upgradeable to accommodate reassembly and upgrades for different applications.
1.6 Application Indicators
The entire system uses 220V AC mains power. Power consumption should be minimized while ensuring system operation. The core control system measures 25x25x25cm. Due to its relatively high power consumption, the power management section is packaged separately, also measuring 25x25x25cm, and equipped with a cooling fan to avoid impacting the low-voltage system. The total cost of the system is below 20,000 yuan.
1.7 Environmental Requirements
The noise level during use must not exceed 40dB. The entire system should not cause electromagnetic interference to the mains power network or the surrounding environment. Electromagnetic compatibility must be controlled within national standards. There should be no strong electromagnetic radiation. Simple shielding should be applied to the high-intensity light portion of the welding torch during operation. The system is not waterproof, has an IP5 dustproof rating, and operates in environments below 90 degrees Celsius and above -15 degrees Celsius.
1.8 Assembly Index Analysis
The assembly accuracy of the entire system must be controlled within 0.1mm , as long as it meets the accuracy requirements of the cross carriage.
2 Overall Design Scheme
Figure 1 Overall Design Scheme and Module Diagram
Figure 1 shows the overall system design and module diagram. The entire system can be divided into 9 main modules, and the system's working logic is as follows:
Upon power-up of the main system, the working environment detection system is activated to monitor the ambient humidity and temperature of the current main control platform. If these conditions are not met, automatic adjustments are attempted. If adjustments are ineffective, the system alarms and automatically shuts down. Once the environment meets the requirements, the DSP main control system starts and performs a self-test. If any modules are found to be in an unready state, the system attempts automatic software repair. If the repair fails, the system automatically restarts. If the problem persists after restarting, the system triggers an alarm. After system initialization, the zero-point positioning sensor locates the initial zero position. After zeroing, a coordinate transformation is performed, converting the coordinates of the zero-point sensor to the coordinates of the welding torch. At this point, the system enters normal operating mode. The DSP uses PWM to control the stepper motor, moving the cross carriage and controlling the arc welding power supply and wire feeding mechanism for welding. After welding begins, the rotating arc sensor detects the arc voltage information, which is converted into corresponding displacement information by the DSP and fed back the displacement, forming a closed-loop operating mode for the entire control system. Simultaneously, various sensors equipped with the system are monitored, including mechanical limit protection, power overvoltage and overcurrent protection, and system overheat and overhumidity protection.
3. Design of the Overall Control Core System
The main control system uses a DSP as the control chip for the entire system, handling its adjustment and control. After comparing various options, the TI TMS320F2812 DSP chip was selected. This series of chips is a 32-bit fixed-point high-speed DSP chip from TI, employing an 8-stage instruction pipeline, a single-cycle 32x32-bit MAC function, and a maximum execution speed of 150 million instructions per second (150 MIPS), ensuring fast and real-time control and signal processing. Furthermore, the TMS320F2812 integrates abundant external resources, including 16-channel 12-bit ADCs, 6-channel PWM outputs, 3 32-bit general-purpose timers, 128KB of 16-bit Flash memory, 18KB of RAM, a Peripheral Interrupt Extension Module (PIE) supporting 45 peripheral interrupts, and interfaces such as McBSP , SPI , SCI, and an extended eCAN bus. The TMS320F2812 also supports external memory expansion up to 1MB.
The TMS320F2812 supports C/C++ programming languages, with its C language optimizer achieving a C compilation efficiency of up to 99%. It also features a virtual floating-point math function library, significantly shortening the development cycle for mathematical operations and control programs. The TMS320F2812 is well-suited for applications such as motor control, power supply design, and intelligent sensor design.
In this system design, a high-speed computing speed is required to complete the calculation of sensor information values and motor control. The information processing involves a large number of floating-point operations, which is precisely the strength of the DSP control system. After comparison and selection, it was decided to use the TMS320F2812 chip from TI.
4. Classification and Working Principle of Arc Sensors
Currently, the main types of arc sensors used in practical applications are as follows:
(1) Non-scanning dual-wire (multi-wire) parallel type
This type utilizes the static characteristics of electric arcs. When the welding torch is misaligned, the difference in height between the two arcs will be reflected in the current (voltage) difference, thus achieving weld seam tracking. However, because it requires two independent circuit power supplies with identical parameters to perform beveling welding in parallel, it is difficult to implement and therefore its practical use is limited.
(2) Oscillating Arc Sensor
Oscillating arc sensors use the arc generated by mechanical oscillation as the sensing medium. Due to the limitations of the mechanism, the scanning frequency is generally very low (below 5Hz), resulting in low sensitivity. They can only be used in low-speed welding. At the same time, the flow and filling of liquid metal in the molten pool also hinders the identification of weld bevels.
(3) Rotary scanning arc sensor
The basic principle of a rotating arc sensor is the same as other arc sensors. Its unique feature is the use of a DC motor to drive an eccentric mechanism, causing the welding wire and arc to rotate, thus achieving high-speed arc scanning. The scanning frequency is typically between 5 and 50 Hz. This design overcomes the problems caused by the low scanning frequency of oscillating arc sensors, greatly improving sensitivity, allowing operation in high-speed applications, and also improving weld quality.
Traditional rotating arc sensors utilize an external DC motor to drive an eccentric mechanism via gear transmission, thereby achieving the rotation of the welding wire and the arc. The problem with this type of structure is its large mechanical size, significant mechanical vibration, and the need to consider the added mass and torque, which in turn affects the selection of a series of related devices. Considering the vertical movement of the welding torch, the torch should be small and lightweight to reduce costs; therefore, a more rational device should be adopted.
The arc welding power supply provides the welding voltage, and the wire feeder completes the wire feeding (generally maintaining a constant speed, but the wire feeding speed can be adjusted by changing the voltage). A hollow shaft DC motor rotates at high speed, causing the welding wire and arc to rotate via an eccentric device. During this process, a current sensor detects the magnitude of the welding current flowing through the welding wire and obtains a voltage signal corresponding to the arc length. This voltage signal is output to the corresponding circuit for sampling and processing. Simultaneously, a photoelectric encoder measures the starting position and instantaneous position relative to the starting point of each arc scan. This is processed to obtain the rotational speed of the hollow shaft motor, enabling closed-loop control of the motor's rotational speed. The encoder output information and the voltage information from the Hall sensor are processed by the main control circuit DSP to control the actuator to adjust the position of the welding torch in the x, y, and z directions. The flowchart for this part of the system is shown in Figure 2. An opto-isolator is used to isolate high-voltage and low-voltage signals between the DSP output signal and the actuator, protecting the control circuit. A computer is used to display the weld condition in real time during the welding process. Through mathematical calculations, the internal condition of the weld can be obtained, such as the weld cross-section, width, depth, and crack plane orientation at a given moment.
Figure 5-14 Flowchart of the arc welding control system
According to relevant literature, the operating frequency of the rotating arc sensor is between 5-50Hz, and the scanning frequency of this system is 25Hz. The following parameters were also determined: the continuous moving speed of the welding torch (under normal circumstances) is 25mm/s, meaning the rotating arc sensor advances 1mm per revolution. Related experimental studies show that the weld seam error of the rotating arc sensor is 0.1mm , and the tracking error is 0.1mm .
5. Analysis of Arc Length Model and Plane Fitting Algorithm
Before performing algorithm analysis, a model of the arc length during the welding process must be established. Taking the tracking of a V-groove weld as an example, the offset distance of the welding torch axis from the weld groove symmetry line in the horizontal direction is called the deviation, denoted as e. Let the distance from the welding torch nozzle end face to the bottom of the weld groove be Hc, the angle between the weld groove and the horizontal plane be β, the arc rotation radius be r, the rotation period be 2T, and the angular velocity be ω.
If t=0 when the welding torch is rotating to the far right, then the arc length H(t) can be obtained from equation (1).
(1)
It refers to the state at a certain instant during the motion process.
According to relevant welding theory , under certain conditions , the transfer function from the arc length change H(s) to the welding current change I(s) during dynamic arc changes can be expressed as:
(2)
in,
Ka
This represents the potential gradient of the electric arc;
Kr
It is a constant related to the melting rate of the welding wire;
Kn
It is a constant related to the elongation resistance and the equivalent resistance of the polar region;
KM
Let P(s) be a constant related to power supply characteristics and welding materials; P(s) represents the dynamic characteristics of the power supply. When the power supply has excellent dynamic quality , P(s) can be considered as a proportional element, and the transfer function can be simplified to a first-order model. It can be seen that when the rotation frequency is constant, the arc length variation is directly proportional to the current variation, and therefore also directly proportional to the voltage variation.
Arc length
H(t)
It can be obtained from equation (1) , and
Therefore, they can be discretized into
Since the rotating arc sensor used in this paper samples 64 times in one scanning cycle, 64 discrete points are selected. According to spatial analytic geometry theory, the equation of a spatial plane can be expressed as:
(3)
For ease of discussion , the above equation is transformed into :
(4)
Introducing vector and matrix notation , let
(5)
Spatial plane fitting of the arc length is to seek
Make it satisfy the following conditions :
6 Conclusions
The entire system design fully considers the characteristics that industrial welding robots should possess. Firstly, it prioritizes safety and stability, employing protective measures including system startup protection, software protection, mechanical limit protection, and alarm protection. The system design also fully considers response speed, which affects the overall responsiveness of the application system and is a crucial indicator of a welding robot's application value. Regarding weld seam tracking and scanning, tracking error and weld seam error meet the required specifications, with a short response time and broad applicability to various weld seam types. In terms of application expansion, the system's scalability is fully considered, with standard serial and Ethernet interfaces reserved to provide necessary hardware support for system debugging and upgrades, and to facilitate the development and upgrading of subsequent series of similar welding robot products.
