Accurate assembly robot of curved surface offset press for high-end equipment

Abstract

The invention relates to the technical field of accurate assembly, and discloses a curved surface offset press accurate assembly robot for high-end equipment, which comprises a robot body fixedly mounted on a main machine of a curved surface offset press through a rigid support structure; the press-fit execution tail end system is connected to the operation end of the robot body and used for executing the press-fit connection technology, the press-fit execution tail end system comprises a base connected with the robot body, a reserved containing cavity is formed in the base, and a quick-change tool head is detachably installed on the base; and a quick disassembly assembly is arranged on the quick-change tool head. A high-precision six-dimensional force sensing module and a servo press fitting driving module are integrated, a real-time sensing-online comparison-decision correction closed-loop control system is constructed, a force / torque signal is dynamically monitored in the press fitting process, active compensation and process parameter adjustment are conducted on centering deviation or interference magnitude fluctuation in real time, and the precision of press fitting is improved. And the printing precision and the operation stability of the whole machine are ensured.

Description

Technical Field

[0001] This invention relates to the field of precision assembly technology, specifically to a precision assembly robot for a curved surface offset printing machine used in high-end equipment. Background Technology

[0002] In the field of high-end printing equipment manufacturing, curved surface offset printing machines are crucial because they can achieve exquisite printing on complex curved surfaces. Their final printing accuracy, registration stability, and long-term operational reliability largely depend on the assembly precision of core moving components such as the printing plate cylinder, rubber cylinder, and their supporting bearings. Currently, the precision assembly of this type of high-end equipment mainly faces technical bottlenecks such as over-reliance on manual experience and skills for assembly quality, a lack of real-time monitoring and adaptive adjustment capabilities in the assembly process, and the difficulty in effectively accumulating and iterating assembly process knowledge.

[0003] Regarding automated assembly technology, existing publicly available literature contains relevant research and improvement solutions. A search revealed patent document CN120095846A, which discloses a multi-dimensional precision positioning automated assembly robot. This robot uses positioning mechanisms mounted on multiple carriages to track and photograph the robotic arm in real time, and utilizes motor-driven rotation of industrial cameras in both the X and Y axes to achieve precise positioning of the robotic arm during component assembly. However, this solution primarily involves visual positioning and tracking during the assembly process, focusing on improving position tracking accuracy through an adjustable-angle multi-camera system. It does not address the real-time perception and dynamic compensation of force / torque signals during the pressing process of core components. In actual pressing operations, even with accurate initial positioning, the stress state during pressing can still change abruptly due to factors such as part size variations, slight alignment deviations, or interference fluctuations. The lack of closed-loop control over these dynamic process parameters makes it difficult to avoid part damage or discrepancies in assembly accuracy caused by uneven stress.

[0004] Furthermore, patent document CN119458429A discloses an assembly robot for manufacturing curved offset printing machines with precise positioning. Through the design of protective and fixing components, it utilizes pressure sensors and airbags to limit the movement of curved offset printing parts, achieving stable loading and unloading. The clamping force is determined by the pressure sensor values. However, this solution primarily focuses on the stability control of workpiece clamping and loading / unloading during the assembly process. Its pressure sensing application is limited to determining whether the clamping force is appropriate, rather than real-time monitoring and adaptive correction of force / torque signals during the pressing process. For complex conditions such as centering deviations and changes in interference fit resistance that occur during the pressing stage, this solution still cannot achieve dynamic perception and immediate correction. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a precision assembly robot for high-end curved surface offset printing machines, which solves the problems of inconsistent assembly accuracy and poor quality due to reliance on human experience and lack of dynamic adaptive capabilities during the pressing and assembly of core components in existing curved surface offset printing machines.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a precision assembly robot for a curved surface offset printing machine in high-end equipment, comprising: The robot body is fixedly mounted on the main unit of the curved offset printing machine by a rigid support structure; The press-fitting end effector system is connected to the operating end of the robot body and is used to perform the press-fitting connection process. The press-fitting end effector system includes a base connected to the robot body. The base has a reserved receiving cavity inside and a quick-change tool head that is detachably installed on the base. The quick-change tool head is equipped with a quick-disassembly component. The press-fit force control and sensing system, integrated into the quick-change tool head, is used to achieve precise control and real-time monitoring of the press-fit process, including: Servo press-fitting drive module, used to provide precision feed power along the press-fitting axis; A multi-dimensional force sensing module is set at the output end of the servo press-fitting drive module to detect multi-dimensional force and torque signals in real time during the press-fitting process; The intelligent tool quick-change system includes a tool library for storing various press-fitting tool heads, and a positioning and docking mechanism for guiding the press-fitting execution end system to grasp, dock, and lock tool heads. The adaptive pressing and fitting control system is connected to the robot body, the pressing and fitting end effector system, the pressing and fitting force control and sensing system, and the intelligent tool quick-change system. It is used to adaptively adjust the pressing and fitting process based on real-time sensing data.

[0007] Preferably, the rigid support structure is a gantry frame, which spans above the main unit of the curved offset printing machine and across its assembly station, and the robot body is mounted on the crossbeam of the gantry frame.

[0008] Preferably, the adaptive pressure-fitting control system includes: The press-fitting process database stores standard press-fitting force-displacement curves and their tolerance ranges related to different press-fitting objects; The real-time monitoring and comparison module is used to acquire the real-time force signal detected by the multi-dimensional force sensing module and the displacement signal of the servo press-fitting drive module, generate a real-time press-fitting curve, and compare it online with the standard curve retrieved from the press-fitting process database. The pressing and assembly correction decision and execution module is used to generate and execute at least one of the following correction instructions in real time during the pressing and assembly process, based on the comparison results: a. Control the robot body to perform active compliant adjustment of the end-effector pose to correct the alignment deviation of the press-fit axis; b. Adjust the feed speed or pressure limit of the servo press-fitting drive module.

[0009] Preferably, the quick-change tool head further includes an integrated vision positioning module, wherein the camera of the integrated vision positioning module is embedded in the outer side of the quick-change tool head housing and is used to identify the features of the parts to be pressed and assembled and the assembly reference.

[0010] Preferably, the quick-release assembly includes a quick-connect plate, one end of which extends into the receiving cavity. A spring is installed on one side of the inner wall of the receiving cavity, and one end of the spring is fixedly connected to a locking frame. The locking frame and one side of the receiving cavity are provided with a slide rail. The tool heads inside the tool magazine are all welded with corresponding quick-connect plates.

[0011] Preferably, in the intelligent tool quick-change system, each press-fitting tool head in the tool library is equipped with an identification tag, and the quick-change tool head is equipped with a tag reader / writer corresponding to the identification tag.

[0012] Preferably, the identification tag is an RFID tag, which stores at least the unique identifier of the tool head and the corresponding standard pressing process parameter index.

[0013] Preferably, it also includes a global three-dimensional measuring instrument fixed in the robot's workspace, used to perform non-contact scanning and accuracy verification of the spatial pose of key components before and after the pressing process.

[0014] Preferably, the robot body is configured to operate as follows: S1: Press-fit task triggering and process loading: Receive press-fit assembly instructions for specific components and load the corresponding standard press-fit process parameters from the press-fit process database; S2: Quick Change and Identification of Press-fit Tool Head: Control the robot body to drive the press-fit execution end system to move to the tool library, complete the grasping and installation of the target press-fit tool head through the positioning and docking mechanism, and read its identity information through the tag reader to confirm the match; S3: Component visual guidance and coarse positioning: Control the robot to grasp the component to be pressed or move it to the pressing station, and use the integrated visual positioning module to identify the reference of the target installation position to complete the coarse positioning of the component. S4: Force-guided adaptive press fitting: Start the servo press fitting drive module to execute the press fitting operation; at the same time, the real-time force-displacement data collected is continuously compared with the standard process parameters through the real-time monitoring and comparison module. When the press fitting correction decision and execution module determines that the real-time data deviates from the allowable range, the corresponding online correction action is immediately executed until the press fitting process is completed. S5: Press-fit result verification and data feedback: The press-fitted components are measured in three dimensions using a global three-dimensional measuring instrument to evaluate the position and attitude accuracy after press-fitting, and the result data is fed back to the control system; S6: Press-fitting process data iteration: The data from this successful adaptive press-fitting process, including the final force-displacement curve and the implemented correction parameters, are updated and stored in the press-fitting process database as empirical data for optimizing the process parameters of subsequent similar press-fitting tasks.

[0015] Preferably, in step S4, the force-guided adaptive press fitting specifically includes: If real-time monitoring detects an abnormal axial pressure curve but a normal torque signal, it is determined to be an interference fit dimensional deviation, triggering an adjustment of the speed or pressure parameters of the servo press-fitting drive module. If real-time monitoring detects an abnormal torque signal, indicating an alignment error, it triggers active compliant fine-tuning of the robot's end-effector pose to correct the press-fit axis.

[0016] Working Principle: By deeply integrating high-rigidity mechanical structures, multimodal perception, real-time intelligent decision-making, and data closed-loop, the traditional discrete assembly operation, reliant on human experience, is transformed into a continuous, adaptive, and self-optimizing precision digital process. Specifically, after receiving instructions, the robot body fixed above the host first accurately grasps and identifies a dedicated quick-change tool head from the intelligent tool library according to the task, and its integrated vision system performs coarse positioning of the parts. Then, it enters the core press-fitting stage: the servo press-fitting module performs axial pressing, while the six-dimensional force sensor integrated into the tool head collects force / torque signals in real time across all dimensions; the adaptive control system compares this real-time data stream with the pre-stored process standard curve at the millisecond level and makes an immediate decision based on a preset algorithm—if abnormal torque is detected, it is determined to be an alignment deviation, and the robot body is instructed to perform sub-millimeter-level active compliant posture fine-tuning to correct the axis; if abnormal axial pressure is detected but the torque is normal, it is determined to be interference fluctuation, and the press-fitting speed or pressure threshold is dynamically adjusted. This "perception-decision-execution" closed loop runs through the entire press-fitting process, ensuring that each assembly dynamically approaches the optimal state. After the press-fit is completed, the global three-dimensional measuring instrument scans the component, compares the actual position with the theoretical model, and generates a quantitative accuracy report. This result data is used for quality judgment and equipment compensation on the one hand, and is associated with the data of the entire press-fit process (force-displacement curve, correction record) on the other hand, and fed back to the process database to drive the continuous iterative optimization of assembly process parameters, thereby achieving the self-evolution of assembly accuracy and consistency.

[0017] This invention provides a precision assembly robot for curved surface offset printing machines in high-end equipment. It offers the following advantages: 1. This invention integrates a high-precision six-dimensional force sensing module and a servo press-fitting drive module, and constructs a closed-loop control system of real-time perception, online comparison, and decision-making correction. During the press-fitting process, it dynamically monitors force / torque signals and proactively compensates for minor alignment deviations or interference fluctuations, adjusting process parameters accordingly. This fundamentally solves the problem of precision dispersion caused by individual differences and fatigue in manual assembly, ensuring that the press-fitting quality of the core components of every curved offset printing machine leaving the factory reaches an extremely high and consistent standard, thereby guaranteeing the final printing accuracy and long-term operational stability of the entire machine.

[0018] 2. This invention combines a dedicated tool head with identification capabilities with integrated visual positioning, enabling robots to automatically and accurately change and calibrate different specifications of fixtures and matching press-fit tool heads according to production instructions. This process achieves calibration during production changeover, significantly reducing the time-consuming and repetitive manual adjustments required in the traditional model. It meets the urgent needs of flexible production of high-end equipment with small batches and multiple varieties, effectively improving equipment utilization and production capacity.

[0019] 3. In this invention, by introducing the multivariable coupled dynamic adaptive threshold model and the position-based impedance control spatial compensation strategy, the situational awareness and adaptive precise correction of the centering deviation in the press-fitting process are realized. This enables the robot to autonomously adjust its sensitivity and correction strategy according to the fitting characteristics of different parts, real-time pressing depth and stress state. Thus, without the need for manual preset of complex parameters, the adaptability to complex and variable working conditions is significantly improved. Ultimately, this ensures that a very high first press-fitting success rate and reliable centering accuracy can be achieved under a wider range of parts batches and assembly conditions, greatly enhancing the process robustness and universality of intelligent assembly of high-end equipment. Attached Figure Description

[0020] Figure 1 This is a perspective view of the precision assembly robot in this invention; Figure 2 This is a diagram illustrating the precision assembly robot used in this invention. Figure 3 This is a schematic diagram of the precision assembly robot in this invention; Figure 4 This is an exploded view of the pressure distribution execution terminal system in this invention; Figure 5 This is an enlarged view of point A in the present invention; Figure 6 This is a cross-sectional view of the quick-disassembly component in this invention; Figure 7 This is an architecture diagram of the pressure distribution control and sensing system in this invention; Figure 8 This is a diagram of the architecture of the adaptive pressure distribution control system in this invention.

[0021] The components include: 1. Robot body; 2. Press-fitting end effector system; 21. Base; 22. Quick-change tool head; 23. Receiving cavity; 24. Spring; 25. Clamping frame; 26. Slide rail; 27. Quick-connect plate; 28. Tag reader / writer; 3. Press-fitting force control and sensing system; 31. Servo press-fitting drive module; 32. Multi-dimensional force sensing module; 33. Integrated vision positioning module; 4. Intelligent tool quick-change system; 41. Tool library; 411. Identification tag; 42. Positioning and docking mechanism; 5. Adaptive press-fitting control system; 6. Gantry frame; 7. Global 3D measuring instrument. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see the appendix Figure 1 -Appendix Figure 8 This invention provides a precision assembly robot for a curved surface offset printing machine used in high-end equipment, comprising: The robot body 1 is fixedly mounted on the main unit of the curved offset printing machine by a rigid support structure; The pressing and fitting execution end system 2 is connected to the operating end of the robot body 1 and is used to perform the pressing and fitting connection process. The pressing and fitting execution end system 2 includes a base 21 connected to the robot body 1. The base 21 has a reserved receiving cavity 23 inside, and a quick-change tool head 22 detachably installed on the base 21. The quick-change tool head 22 is provided with a quick disassembly component. The press-fitting force control and sensing system 3, integrated on the quick-change tool head 22, is used to achieve precise control and real-time monitoring of the press-fitting process, including: Servo press-fitting drive module 31 is used to provide precision feed power along the press-fitting axis; A multi-dimensional force sensing module 32 is set at the output end of the servo press-fitting drive module 31 and is used to detect multi-dimensional force and torque signals in real time during the press-fitting process. The intelligent tool quick-change system 4 includes a tool library 41 for storing various press-fitting special tool heads, and a positioning and docking mechanism 42 for guiding the press-fitting execution end system 2 to grasp, dock and lock tool heads; The adaptive pressing and fitting control system 5 is connected to the robot body 1, the pressing and fitting execution end system 2, the pressing and fitting force control and sensing system 3, and the intelligent tool quick-change system 4. It is used to adaptively adjust the pressing and fitting process based on real-time sensing data.

[0024] The adaptive pressure-fitting control system 5 includes: The press-fitting process database stores standard press-fitting force-displacement curves and their tolerance ranges related to different press-fitting objects; The real-time monitoring and comparison module is used to acquire the real-time force signal detected by the multi-dimensional force sensing module 32 and the displacement signal of the servo press-fitting drive module 31, generate a real-time press-fitting curve, and compare it online with the standard curve retrieved from the press-fitting process database. The pressing and assembly correction decision and execution module is used to generate and execute at least one of the following correction instructions in real time during the pressing and assembly process, based on the comparison results: a. Control the robot body 1 to perform active compliant adjustment of the end-effector pose in order to correct the alignment deviation of the press-fit axis; b. Adjust the feed speed or pressure limit of the servo press-fitting drive module 31.

[0025] Specifically, the precise implementation of the press-fitting process begins with a reliable data benchmark, and the press-fitting process database is a dynamically growing experiential knowledge base. In the initial stage, through theoretical calculations, digital twin simulations, and statistical analysis of numerous historical successful assembly cases, at least one standard press-fitting force-displacement curve and its tolerance range are established for each specific component. The standard press-fitting force-displacement curve is pre-defined based on the theoretical interference of the part, material properties, and historical successful assembly data, through mechanical model simulation and statistical learning. This curve defines the ideal axial pressure and feed displacement relationship throughout the entire process from the start to the end of press-fitting, while also including the upper and lower limits of allowable deviations. When the robot receives a specific press-fitting task instruction, the system first accurately retrieves the corresponding standard curve and process parameters from the database according to the task code (identification tag 411 in this embodiment), as the target path for this operation.

[0026] After the pressing action is initiated, the system's real-time monitoring and comparison module immediately enters high-speed operation. The multi-dimensional force sensing module 32 continuously acquires millisecond-level multi-dimensional force and torque signals (Fx, Fy, Fz, Mx, My, Mz) at the pressing contact point, while the encoder of the servo pressing drive module 31 acquires displacement signals. These raw displacement signals are rapidly processed and fused to generate a real-time pressing curve reflecting the current assembly status.

[0027] The key is that this module doesn't perform data analysis after the pressing process is complete, but rather conducts millisecond-level online comparisons. It divides the entire pressing process into characteristic stages based on the pressing depth, such as the "no-load contact stage," "initial interference stage," "steady-state pressing stage," and "fitting stage." Each stage has a preset dynamic reference curve and adaptive tolerance zone. During comparison, the algorithm not only calculates the absolute deviation between the real-time force value (Fz) and the standard value, but more importantly, it analyzes the changing trends of the first derivative (i.e., the force gradient dF / dS) and second derivative of the real-time force-displacement curve. For example, in the "steady-state pressing stage," if the real-time force gradient continuously exceeds the gradient tolerance of the standard curve, even if the absolute force value has not yet exceeded the limit, the system will issue an early warning of potential gradual interference. It instantly compares each data point generated in real time with the theoretical value at the corresponding position of the standard curve retrieved from the database. The comparison includes not only whether the axial pressure (Fz) is within the expected range, but also focuses on analyzing whether the torque signals (Mx, My) approach zero in the ideal alignment state, and the changing trend of the lateral forces (Fx, Fy). Subtle anomalies that deviate from the target path are instantly captured and quantified.

[0028] Once the real-time monitoring and comparison module detects that the real-time data exceeds the preset tolerance range, the assembly deviation correction decision and execution module is immediately triggered. This module incorporates rule-based or model-based decision algorithms, which can quickly determine the root cause of the fault and select the optimal correction strategy based on the pattern and degree of deviation. Mode A: Centering Deviation Correction. If real-time data shows significant anomalies in the torque components (Mx, My), while the axial pressure remains within the normal range, the system determines that there is a micro-centering deviation in the press-fit axis. In this case, the system will not stop operation but will immediately generate an active compliant adjustment command. This command controls the joint servo drive of robot body 1, causing its end effector to make a small, compliant translational or rotational movement in the opposite direction of the torque feedback. This action corrects the robot's end effector pose in real time, actively finding and restoring it to the centering state, thereby eliminating abnormal torque. Furthermore, the system's determination of the torque signal (Mx, My) incorporates a short-term energy statistics and trend consistency check. Simple single-point exceedances may originate from noise; therefore, the algorithm calculates the root mean square value of the torque signal over the past N sampling periods and compares it with a dynamic threshold based on the current axial pressure (Fz) and stage identifier. This threshold model is expressed as:

[0029] The threshold value for determining the torque at the current moment; This refers to the real-time axial pressing force. Current push-in depth; Theoretical fit length; These are the nominal diameters of the mating hole and the shaft, respectively. This is a sensitivity coefficient matrix based on the assembly stage (such as the guiding stage, stabilizing stage, and refining stage); , , The system gain coefficient is obtained through extensive process experiments and machine learning calibration.

[0030] The coefficients a and b were calibrated through numerous process experiments. Once an alignment deviation is determined, the system does not perform simple directional reversal compensation. Instead, based on the direction of the torque vector in the XY plane and the real-time force gradient change, a fuzzy PID controller calculates the spatially optimal compensation path. This path may be a small circular arc trajectory rather than a straight line, in order to minimize the additional stress during the adjustment process.

[0031] Mode B: Interference Fit Adaptive Correction. If real-time data shows a normal torque signal, but the axial pressure (Fz) deviates from the standard curve (e.g., pressure rises too early or increases too quickly), the system determines that there is a manufacturing variance in the actual interference fit of the mating surfaces of the parts. To address this, the system employs a simplified model predictive control strategy. When the real-time Fz curve continuously deviates from the standard band, the system not only adjusts the current speed but also predicts the force values ​​that may occur in the next few steps (e.g., a 2mm displacement) based on the current degree of deviation and force gradient, and proactively adjusts the preset speed curve for subsequent segments. Simultaneously, the system associates the parts' identity information (e.g., batch number). If parts from the same batch frequently trigger specific types of adjustments, the controller will fine-tune its internal model parameters online, achieving batch-level adaptive learning.

[0032] At the level of the entire robot system, its workflow is an unfolding of the above core principles on a timeline: The robot first accurately acquires and identifies the special tool head through the intelligent tool quick change system 4; it then uses the vision system on the tool head for coarse positioning; subsequently, it enters the force-guided adaptive pressing and fitting stage, which is entirely dominated by the above closed-loop principle until the pressing and fitting is completed; finally, the results can be verified by external measuring equipment, and the successful process data can be fed back to the database to achieve continuous learning and optimization.

[0033] Furthermore, after each assembly, the system generates a multi-dimensional process feature vector containing: [the principal eigenvalues ​​of the final force-displacement curve, the type and frequency of correction triggered at each stage, the effectiveness coefficient of the correction action, and the key dimension deviations of the three-dimensional measurement results]. These feature vectors are stored in association with the part material number, batch number, and environmental temperature and humidity information. The system periodically analyzes these vectors using an unsupervised clustering algorithm to automatically discover potential quality pattern clusters. For example, it automatically identifies that "a batch of bearings from supplier A typically clusters its feature vectors within clusters requiring 'low-speed press-fitting'." Subsequently, the system automatically recommends or directly creates a new optimized process curve for such parts. This transforms the process database from a static reference library into an intelligent system capable of autonomously discovering correlations, summarizing knowledge, and generating optimization suggestions, achieving automation and intelligence in process optimization.

[0034] The rigid support structure is a gantry frame 6, which spans above the main unit of the curved offset printing machine and across its assembly station. The robot body 1 is mounted on the crossbeam of the gantry frame 6.

[0035] The quick-change tool head 22 also includes an integrated vision positioning module 33. The camera of the integrated vision positioning module 33 is embedded in the outer side of the housing of the quick-change tool head 22 and is used to identify the features of the parts to be pressed and assembled and the assembly reference.

[0036] The quick-release assembly includes a quick-connect plate 27, one end of which extends into the receiving cavity 23. A spring 24 is installed on one side of the inner wall of the receiving cavity 23. One end of the spring 24 is fixedly connected to a locking frame 25. The locking frame 25 and one side of the receiving cavity 23 are provided with a slide 26. The tool heads inside the tool magazine 41 are all welded with corresponding quick-connect plates 27.

[0037] In the intelligent tool quick-change system 4, each press-fitting special tool head in the tool library 41 is equipped with an identification tag 411, and the quick-change tool head 22 is equipped with a tag reader 28 corresponding to the identification tag 411.

[0038] The identification tag 411 is an RFID tag, and the information it stores includes at least the unique identifier of the tool head and the corresponding standard pressing process parameter index.

[0039] Specifically, the core of the precision assembly robot is the robot body 1, which is fixedly mounted above the curved offset printing machine main unit via a high-rigidity, high-stability gantry frame 6. The gantry frame 6 spans the core assembly station of the main unit, forming a stable support frame. The robot body 1 is mounted upside down or sideways on the crossbeam of the gantry frame 6, allowing its operating end to vertically cover all critical assembly areas below. This layout ensures that the robot's workspace does not interfere with the ground personnel activity area or the workpiece conveying path.

[0040] The intelligent tool quick-change system 4 is the foundation for the robot to perform multi-functional operations. It includes a tool library 41, which stores in an orderly manner specialized quick-change tool heads 22 designed for different press-fitting tasks (such as bearing press-fitting, bushing press-fitting, and gear installation). Each quick-change tool head 22 has a quick-connect plate 27 welded to its base and embedded with an RFID tag for use with a tag reader 28. The tag reader stores the tool head's unique ID and the corresponding standard press-fitting process parameter index.

[0041] The quick-change process begins when the adaptive press-fit control system 5 receives the task instruction. The robot body 1 drives the press-fit execution end effector 2 to move to the tool magazine 41. With the assistance of the positioning and docking mechanism 42, the end effector base 21 is precisely aligned with the target tool head. The base 21 has a receiving cavity 23 inside. After alignment, the quick-connect plate 27 on the tool head is inserted into this receiving cavity 23. A spring 24 is installed on one side of the inner wall of the receiving cavity 23, and its end is connected to a locking bracket 25. During insertion, the quick-connect plate 27 pushes the locking bracket 25 along the slide 26 to compress the spring 24 until it reaches the preset position. The spring 24 rebounds and drives the locking bracket 25 to engage with the corresponding groove of the quick-connect plate 27 to achieve mechanical locking. When the quick-connect plate 27 is inserted into the predetermined position along the slide 26, the locking bracket 25 automatically springs up under the action of the spring 24 and engages with the groove at the bottom of the quick-connect plate 27 to achieve locking. Applying a release force in a specific direction can overcome the spring force and disengage the locking bracket 25 to achieve quick disassembly. Simultaneously, the gas and electrical interfaces are automatically connected. The tag reader 28 on the base 21 immediately reads the information from the tool head identification tag 411, confirms the tool head's identity, and automatically retrieves the corresponding standard press-fit curve and parameters from the press-fit process database, completing the synchronization of tool quick change and process preparation.

[0042] It also includes a global 3D measuring instrument 7 fixed in the robot's workspace, used for non-contact scanning and accuracy verification of the spatial pose of key components before and after the pressing process.

[0043] Specifically, before the assembly work begins, the global 3D measuring instrument 7 must first complete system calibration. By scanning standard gauge blocks placed in the robot's workspace, a precise transformation relationship is established between the measuring instrument's own coordinate system, the robot's base coordinate system, and even the theoretical design coordinate system of the offset printing machine. Simultaneously, the 3D models of the key components being measured have been pre-imported into the measurement system's analysis software as digital benchmarks for comparison. Before some high-precision assembly tasks begin, the robot can guide the measuring instrument to pre-scan the installed benchmark components (such as the bearing mounting hole system on the offset printing machine's wall panel). By quickly acquiring the actual 3D point cloud data of these benchmark features and fitting and comparing it with the theoretical model, the system can calculate the minute installation deviations between the actual benchmarks and the theoretical positions. These deviation data can be fed back to the robot in real time, enabling it to compensate for and position the target installation location during subsequent press-fit assembly. After the robot completes a press-fit process, the measuring instrument performs a controlled or automatic 3D scan of the assembly. For some adjustable assemblies, the measured deviation data can be immediately fed back to the offset printing machine's central control system. The central control system can use this mechanical error as a software compensation factor in subsequent printing registration control, dynamically adjusting the electronic axis phase. This ensures printing accuracy even when mechanical precision is not absolutely ideal, thanks to the control algorithm. Furthermore, for certain adjustable assemblies, the measured deviation data can be immediately fed back to the offset printing press's central control system.

[0044] Robot body 1 is configured to operate as follows: S1: Press-fit Task Triggering and Process Loading: When the upper-level manufacturing execution system or operator issues a press-fit command for a specific component, the adaptive press-fit control system 5 is immediately triggered. Its core action is to accurately retrieve the assembly process package uniquely corresponding to that component from the press-fit process database. This database is generated through digital twin simulation, statistical analysis of a large amount of historical successful assembly data, and machine learning, resulting in an ideal path with tolerance range. The loading process not only reads data but also includes fine-tuning based on the component batch number. The range of active compliant fine-tuning is usually within ±0.1mm, automatically associating the possible dimensional statistical characteristics of the batch, and setting a personalized initial benchmark that combines theoretical optimality and historical experience for subsequent adaptive control. S2: Quick-change and Identification of Press-fit Tool Heads: This step enables on-demand configuration and identity verification of the robot's execution capabilities. The robot body 1 drives the end effector base 21 to move to the intelligent tool library 41. The positioning and docking mechanism 42 ensures that the base 21 and the target quick-change tool head 22 achieve sub-millimeter precision alignment. After alignment, the receiving cavity 23 of the base 21 engages with the quick-connect plate 27 of the tool head, and the internal spring 24 engages, completing the mechanical locking and the connection of the air and electrical circuits. At the same time, the tag reader 28 on the base 21 instantly reads the information of the identification tag 411 on the tool head. The tool head identification tag 411 read in the identification step is compared with the tool type required by the task instruction in step S1, realizing double confirmation of hardware matching and preventing misuse of tools. After confirmation, the system automatically binds the inherent performance parameters of the tool head with the process parameters to be executed at the software level, completing the entire process from physical connection to information fusion. S3: Component Visual Guidance and Coarse Positioning: This step establishes a spatial starting point for high-precision press-fitting. The robot utilizes the integrated visual positioning module 33 on the quick-change tool head 22 to photograph and identify specific reference features on the components to be press-fitted (e.g., bearings) and the target mounting positions (e.g., roller journals). These specific reference features include, but are not limited to, chamfered edges, scribing lines, or pre-attached markers. Through image processing algorithms, the system calculates the deviation (ΔX, ΔY, ΔZ, ΔRx, ΔRy, ΔRz) between the current actual position of the component and the ideal assembly position in three-dimensional space. Based on this deviation, the robot then plans and executes a motion path to move the component to a pre-press-fitting preparation position. This position is typically set where the component is about to contact the mounting hole but is not under force. Visual guidance reduces the initial assembly error from the millimeter level (possibly due to manual placement) to a fraction of a millimeter level (repeatable by the robot), eliminating most macroscopic positional uncertainties for the subsequent fine-tuning stage, which relies entirely on force perception. This avoids insufficient force control correction range or failure due to excessive initial deviation. S4: Force-Guided Adaptive Press Fitting: This step is the execution and assurance stage of the core process, achieving a leap from programmed execution to intelligent adaptation. The servo press fitting drive module 31 begins to apply axial feed at a predetermined speed. Simultaneously, the multi-dimensional force sensing module 32 begins to collect full-dimensional force / torque signals at the press fitting contact points at frequencies up to several kilohertz. The real-time monitoring and comparison module performs millisecond-level synchronous comparisons of these signals with the standard curve loaded in S1.

[0045] If the monitoring detects that the torque signal (Mx, My) continuously exceeds the threshold, it indicates a micro-tilt or eccentricity in the press-fit axis, generating a harmful lateral bending moment. The press-fit correction decision and execution module immediately calculates the torque direction and instructs the joint servo system of the robot body 1 to make its end effector perform a very small-amplitude translational or rotational compensation movement in accordance with the force direction while maintaining axial feed. This process is dynamic and continuous, as if the robot is using "force-sensing fingers" to sense and actively align itself until the torque signal returns to near zero, ensuring that the press-fit process is always under pure axial force, protecting the parts from damage.

[0046] If monitoring reveals that the axial force (Fz) curve deviates from the standard band while the torque is normal, it mainly reflects a slight fluctuation in the actual interference fit of the mating surfaces. The decision module will dynamically adjust the control parameters of the servo press-fitting drive module 31 accordingly: if the resistance is too high, the press-fitting speed will be appropriately reduced or the pressure will be paused to prevent jamming; if the resistance is too low, the speed may be slightly increased to ensure proper assembly. This is an online optimization of process parameters based on real-time feedback, ensuring that different batches of parts can obtain suitable press-fitting force. S5: Press Fitting Result Verification and Data Feedback: Success is not automatically granted upon completion of the press fitting process. The global 3D measuring instrument 7 (using a laser tracker in this embodiment), fixed next to the workstation, is triggered to perform a non-contact 3D scan of the assembly, acquiring high-density point cloud data of key features (such as the outer ring of the bearing after press fitting). By optimally fitting and aligning the point cloud data with the CAD theoretical model of the component, the software automatically calculates the key geometric tolerances after assembly, such as radial runout, end face runout, and coaxiality, and generates a quantitative report. This report is not only used for conformity assessment, but its contained vector deviation data (including but not limited to runout and phase angle) can be fed back to the offset printing machine's main control system in real time. The main control system can use this mechanical error as an electronic compensation amount to offset it in subsequent printing control, thereby further improving the overall machine's output performance at the mechanical precision level. S6: Iterative Data Processing for Press Fitting: After a successful adaptive press fitting operation, the system does not simply end the task. It treats all the data from the entire press fitting cycle—including the final force-displacement curve, all triggered correction commands and their magnitudes, batch information of the corresponding parts, and the final measurement results from S5—as a complete case package, storing it in a structured manner and performing correlation analysis. Through data mining, the system can summarize patterns: for example, the optimal press fitting speed for a certain batch of parts from a specific supplier generally needs to be reduced by 5%. Subsequently, when similar parts re-enter the assembly queue, the system will prioritize recommending or directly apply the optimized new process parameters. This allows the press fitting process database to evolve from a static reference library into a dynamically growing, continuously optimized intelligent process, with assembly accuracy and efficiency constantly improving as production experience accumulates.

[0047] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A precision assembly robot for a high-end curved surface offset printing machine, characterized in that, The application relates to a robot system for precision assembly of curved surface parts, comprising: a robot body (1) fixedly installed on a main machine of a curved surface printing machine through a rigid support structure; a press-fit execution end system (2) connected to an operation end of the robot body (1) and used for executing a press-fit connection process, wherein the press-fit execution end system (2) comprises a base (21) connected to the robot body (1), the inside of the base (21) is provided with a reserved accommodating cavity (23), and a quick-change tool head (22) is detachably installed on the base (21), and the quick-change tool head (22) is provided with a quick-release assembly; a press-fit force control and sensing system (3) integrated on the quick-change tool head (22) and used for realizing precision control and real-time monitoring of a press-fit process, comprising: a servo press-fit driving module (31) used for providing precision feeding power in the direction of a press-fit axis; a multi-dimensional force sensing module (32) arranged at the output end of the servo press-fit driving module (31) and used for real-time detection of multi-dimensional force and torque signals in the press-fit process; an intelligent tool quick-change system (4) comprising a tool library (41) used for storing a plurality of press-fit special tool heads and a positioning and docking mechanism (42) used for guiding the press-fit execution end system (2) to carry out tool head grabbing, docking and locking; an adaptive press-fit control system (5) in signal connection with the robot body (1), the press-fit execution end system (2), the press-fit force control and sensing system (3) and the intelligent tool quick-change system (4) and used for adaptively adjusting the press-fit process according to real-time sensing data.

2. A precision assembly robot for a high-end curved surface offset printing press as claimed in claim 1, wherein The rigid support structure is a portal frame (6), the portal frame (6) is arranged above the main machine of the curved surface printing machine and crosses the assembly station, and the robot body (1) is installed on the crossbeam of the portal frame (6).

3. A precision assembly robot for a high-end curved surface offset printing press as claimed in claim 1, wherein The adaptive press-fit control system (5) comprises: a press-fit process database used for storing standard press-fit force-displacement curves and tolerance ranges associated with different press-fit objects; a real-time monitoring and comparison module used for acquiring real-time force signals detected by the multi-dimensional force sensing module (32) and displacement signals of the servo press-fit driving module (31), generating a real-time press-fit curve and carrying out online comparison with a standard curve called from the press-fit process database; a press-fit deviation correction decision and execution module used for generating and executing at least one deviation correction instruction in the press-fit process according to the comparison result, wherein the deviation correction instruction comprises: a, controlling the robot body (1) to actively and flexibly adjust the end position to correct the press-fit axis centering deviation; and b, adjusting the feeding speed or the pressure upper limit of the servo press-fit driving module (31).

4. A precision assembly robot for a high-end curved surface offset printing press as claimed in claim 1, wherein, The quick-change tool head (22) further comprises an integrated visual positioning module (33), the camera of the integrated visual positioning module (33) is embedded on the outside of the shell of the quick-change tool head (22) and is used for identifying the features of a part to be press-fitted and an assembly reference.

5. A precision assembly robot for a high-end curved surface offset printing press as claimed in claim 1, wherein, The quick disassembly assembly comprises a quick connection plate (27), one end of the quick connection plate (27) extends into a containing cavity (23), a spring (24) is mounted on one side of the inner wall of the containing cavity (23), one end of the spring (24) is fixedly connected with a clamping frame (25), the clamping frame (25) and one side of the containing cavity (23) are jointly provided with a sliding channel (26), and the tool heads inside the tool library (41) are all welded with corresponding quick connection plates (27).

6. A precision assembly robot for a high-end curved surface offset printing press as claimed in claim 1, wherein, In the intelligent tool quick change system (4), each press-fitting special tool head in the tool library (41) is provided with an identity recognition label (411), and the quick change tool head (22) is provided with a label reader / writer (28) corresponding to the identity recognition label (411).

7. A precision assembly robot for a high-end curved surface offset printing press as claimed in claim 6, wherein The identity recognition label (411) is an RFID label, and the information stored in the RFID label at least includes the unique identification of the tool head and the corresponding standard press-fitting process parameter index.

8. A precision assembly robot for a high-end curved surface offset printing press according to claim 1, wherein A global three-dimensional measuring instrument (7) fixed in the robot workspace is further included, which is used for non-contact scanning and accuracy verification of the space pose of a key component before and after the press-fitting process.

9. A precision assembly robot for a high-end curved surface offset printing press according to claim 3, wherein, The robot body (1) is configured to perform the following operations: S1: press-fitting task triggering and process loading: receiving a press-fitting assembly instruction for a specific component, and loading corresponding standard press-fitting process parameters from the press-fitting process database; S2: press-fitting tool head quick change and identification: controlling the robot body (1) to drive the press-fitting execution end system (2) to move to the tool library (41), completing the grabbing and installation of the target press-fitting tool head through the positioning and docking mechanism (42), and reading the identity information of the target press-fitting tool head through the label reader / writer (28) to confirm the matching; S3: component visual guidance and coarse positioning: controlling the robot to grab a component to be press-fitted or to move to a press-fitting station, identifying the reference of the target installation position by using the integrated visual positioning module (33), and completing the coarse positioning of the component; S4: force sensing guided adaptive press-fitting: starting the servo press-fitting driving module (31) to execute the press-fitting operation; at the same time, continuously comparing the real-time force-displacement data collected by the real-time monitoring and comparison module with the standard process parameters, and when the press-fitting correction decision and execution module determines that the real-time data deviates from the allowable range, immediately performing corresponding online correction actions until the press-fitting process is completed; S5: press-fitting result verification and data feedback: performing three-dimensional measurement on the component after the press-fitting by using the global three-dimensional measuring instrument (7), evaluating the position and attitude accuracy after the press-fitting, and feeding back the result data to the control system; S6: press-fitting process data iteration: storing the data of the successful adaptive press-fitting process, including the final force-displacement curve and the implemented correction parameters, into the press-fitting process database as experience data for updating, and used for optimizing the process parameters of subsequent similar press-fitting tasks.

10. A precision assembly robot for a high-end curved surface offset printing press according to claim 9, wherein, In step S4, the force sensing guided adaptive press-fitting specifically comprises: If the real-time monitoring finds that the axial pressure curve is abnormal while the torque signal is normal, it is determined that the interference fit size deviation is triggered, and the speed or pressure parameter of the servo press-fitting driving module (31) is adjusted. If the real-time monitoring finds that the torque signal is abnormal, indicating that there is a centering error, active soft fine adjustment of the end pose of the robot body (1) is triggered to correct the press-fit axis.

Images