Production system and method for high-energy-efficiency three-cylinder compressor parts

Abstract

The invention discloses a production system and method for high-energy-efficiency three-cylinder compressor parts. The production system comprises a disc stock bin, a power cutter tower type numerical control lathe, a truss mechanical arm, a turning-around transfer device, a clamping jaw, a workpiece locating module and a control system. The disc stock bin is used for containing and positioning a to-be-machined air cylinder for the three-cylinder compressor, composite positioning is achieved by inserting a mandrel into an inner hole, and materials are supplied to the truss mechanical arm. The power cutter tower type numerical control lathe integrates the machining functions of plane turning, positioning hole machining, threaded hole machining, lateral air suction hole machining, spring hole machining and enthalpy increasing hole machining. The truss mechanical arm cooperates with the disc stock bin, the lathe and the turning-around transfer device and is used for automatically transferring workpieces. The turning-around transfer device is arranged between procedures, and a servo motor drives a rotating platform to adjust the direction of a workpiece. The clamping jaw is installed at the tail end of the mechanical arm and used for grabbing a workpiece. The workpiece locating module is used for rapidly locating a workpiece; the control system communicates with all the modules, monitors the state of the cutter in real time and triggers parameter adjustment.

Description

Technical Field

[0001] The present invention relates to the technical field of compressor component manufacturing, and particularly to a production system and method for high-efficiency three-cylinder compressor components. Background Art

[0002] As a core component of refrigeration and power equipment, the processing accuracy (such as flatness, hole position tolerance) and production efficiency of key components such as cylinders of a three-cylinder compressor directly affect the energy efficiency and reliability of the compressor. With the continuous growth of the market demand for high-efficiency compressors, the following limitations have gradually emerged in the traditional production technology of three-cylinder compressor components: First of all, the coordination between loading positioning and feeding is insufficient. In the existing technology, the loading of cylinder-type workpieces mostly relies on manual placement or simple bin positioning, which can only achieve single-dimensional positioning in the radial or circumferential direction, and often requires manual secondary adjustment of the direction, resulting in positioning deviation; at the same time, there is no dynamic matching mechanism between the feeding rhythm of the bin and the grasping cycle of the robotic arm, which is prone to feeding accumulation (leading to mis-grasping) or waiting (idle running of the equipment), reducing the production line efficiency.

[0003] Secondly, the integration degree of multi-process machining is low. Traditional numerical control lathes mostly adopt a single-process machining mode. Processes such as turning the plane, drilling, and tapping require multiple clamping switches. The connection between processes takes time, and the time coordination of multi-directional machining cannot be achieved, resulting in a long single-piece machining cycle and being difficult to meet the rhythm requirements of large-scale production of high-efficiency compressors.

[0004] Furthermore, the efficiency and accuracy of direction adjustment between processes are insufficient. Cylinder-type workpieces often need to be machined on multiple surfaces (such as the front and back end faces, lateral hole positions). The existing technology relies on manual flipping or a simple rotating table to adjust the direction, with a long adjustment time and low accuracy, which is prone to subsequent machining deviation and affects product consistency.

[0005] In addition, the management of cutting fluid and the protection of clamping jaws are lacking. The residual cutting fluid in the machining process is easy to corrode the clamping jaws, and the cutting fluid drips onto the machine tool guide rail during transfer, shortening the equipment life; at the same time, the grasping force control of the clamping jaws is rough (too tight causes workpiece deformation, too loose causes dropping), further affecting the machining stability.

[0006] Finally, the monitoring of tool status and the response to abnormalities are lagging. Traditional technologies mostly monitor tool wear through manual inspection or a single sensor (such as a current sensor), and cannot identify initial wear in real time. Often, workpiece scrapping is caused by tool abnormalities (such as excessive wear); and when dealing with abnormalities, manual intervention is required to stop the machine and re-clamp, and it is difficult to continue machining after the interruption, resulting in a long production interruption time.

[0007] In summary, the existing production technologies for components of three-cylinder compressors have significant deficiencies in aspects such as positioning accuracy, process integration, direction adjustment, cutting fluid protection, and tool monitoring. There is an urgent need for a production system and method with high energy efficiency, high precision, and intelligent collaboration capabilities to improve machining quality and production line efficiency and meet the large-scale production requirements of high-energy efficiency compressors. Summary of the Invention

[0008] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a production system and method for components of high-energy efficiency three-cylinder compressors, which are used to solve the technical problems of low positioning accuracy, poor process integration, insufficient direction adjustment efficiency and accuracy, lack of cutting fluid management, and lag in tool status monitoring during the production process of existing components of three-cylinder compressors (especially key components such as cylinders).

[0009] To solve the above problems, the technical solutions adopted by the present invention are as follows: A production system for components of high-energy efficiency three-cylinder compressors, comprising: A disk magazine, which is used to place and position the cylinders for three-cylinder compressors to be processed and automatically supply materials to the truss robot arm. A power turret CNC lathe configured to integrate multi-process machining functions. A truss robot arm, which is cooperatively arranged with the disk magazine, the power turret CNC lathe, and the turning and transfer device, and is used for the automatic transfer of workpieces. A turning and transfer device, which is arranged between adjacent processing procedures and is configured to adjust the direction of the workpiece within a preset time. Jaws, which are installed at the end of the truss robot arm and are used to grasp the workpiece, and the structural design can reduce the damage of cutting fluid to the workpiece and the jaws themselves. A workpiece positioning module, which integrates a laser displacement sensor, an industrial camera, and a vision processing unit, and is installed on the truss robot arm and the turning and transfer device. It is configured to perform multi-source data error characteristic modeling, dynamic adjustment of Kalman filter parameters, adaptive correction of reflectivity compensation parameters, and verification of positioning confidence, and output the final positioning coordinates and angle compensation values to identify the position of the workpiece. A control system, which is communicatively connected to the disk magazine, the power turret CNC lathe, the truss robot arm, the turning and transfer device, the jaws, and the workpiece positioning module, and is configured to perform fusion analysis including multi-modal data time-frequency domain feature fusion and dynamic evolution of the random forest model to obtain the tool usage status and adjust the processing parameters.

[0010] A production method for components of high-energy efficiency three-cylinder compressors, comprising the following steps: Manually place the cylinders for three-cylinder compressors in the disk magazine in a preset direction, position the workpiece through the core shaft of the disk magazine, and automatically supply materials to the truss robot arm. After the workpiece location is quickly determined using the workpiece location module, the workpiece is grabbed by the truss robot arm using the gripper and transferred to the power turret CNC lathe; The power turret CNC lathe is used to process planes, positioning holes, threaded holes, lateral air intake holes, spring holes and enthalpy-increasing holes; Between adjacent processing steps, the workpiece direction is adjusted within a preset time through the U-turn transfer device to provide a suitable position for subsequent processing; Based on multimodal data, the control system uses a machine learning model to monitor the tool usage status in real time and dynamically adjust the processing parameters to complete the automated processing of cylinders for three-cylinder compressors.

[0011] Preferably, positioning the workpiece and automatically feeding the truss robot arm includes: By inserting the mandrel into the inner hole of the workpiece, the radial positioning and circumferential positioning of the workpiece are achieved at the same time, so that the workpiece is fixed on the surface of the disc bin in the preset processing direction; Through the rotary indexing device configured in the disc silo, the positioned workpiece is transported to the grabbing area of ​​the truss robot arm at a preset beat according to the grabbing cycle of the truss robot arm; During the feeding process, the contact sensor monitors the matching status of the workpiece and the mandrel. If it is detected that the workpiece is not completely inserted into the mandrel, the disc bin alarm is triggered and the feeding is suspended. Wherein, the contact sensor is integrated at a key position where the side surface of the core shaft contacts the inner hole of the workpiece.

[0012] Preferably, when transporting to the grasping area of ​​the truss robot arm, it includes: The actual grabbing cycle T1 of the truss robot arm is obtained through real-time interaction between the silo controller and the control system; Compare the difference ΔT between the actual grabbing cycle T1 and the theoretical grabbing cycle T0 pre-stored in the silo controller, and dynamically adjust the speed of the rotary indexing device using a pre-compensation mechanism based on the cycle fluctuation trend, so that the matching error between the actual feeding beat T2 and the actual grabbing cycle T1 is lower than the set value; The grasping area of ​​the truss robot arm is defined as the zero grasping point and is pre-programmed and bound to the initial reset position of the truss robot arm and the starting point coordinates of the motion trajectory; The positioned workpiece is pre-sent to the cache position of the grasping area by the rotary indexing device one beat cycle in advance. After the truss robot arm completes the grasping of the previous workpiece and returns to the initial position, the workpiece is pushed from the cache position to the zero-position grasping point by the rotary indexing device.

[0013] Preferably, monitoring the matching state between the workpiece and the mandrel includes: The control system pre-stores a standard pressure threshold value P0 that matches the cylinder for the three-cylinder compressor; When the disk bin starts, calibrate the zero point of the contact sensor through no-load testing; When the workpiece is sleeved onto the mandrel, the contact pressure value P is collected in real time through the contact sensor. If P≥P0, it is determined that the workpiece is fully sleeved. If P<P0, it is determined that the workpiece is not fully sleeved; Among them, when it is detected that P<P0, the bin controller triggers a three-level response mechanism.

[0014] Preferably, when using the workpiece positioning module to quickly determine the workpiece position, it includes: Use a laser displacement sensor to scan the surface of the workpiece to obtain the initial spatial coordinates and contour features of the workpiece, providing a pre-positioning area for visual fine positioning; Use an industrial camera to collect images of the workpiece in the pre-positioning area, and run a feature recognition algorithm through the vision processing unit to identify the key positioning features of the workpiece, and calculate the accurate coordinates and tilt angle of the workpiece; Input the initial spatial coordinates and accurate coordinates into the data fusion algorithm, perform multi-source data error characteristic modeling, dynamic adjustment of Kalman filter parameters, adaptive correction of reflectivity compensation parameters, and verification of positioning confidence, and output the final positioning coordinates and angle compensation values; Transmit the final positioning coordinates and angle compensation values to the control system through the workpiece positioning module to control the truss manipulator to dynamically adjust the grasping path.

[0015] Preferably, when transferring to a power turret lathe, it includes: When the truss manipulator receives the final positioning coordinates and angle compensation values, dynamically adjust the jaw posture through the joint servo motor; After the truss manipulator grasps the workpiece, transfer it to the power turret lathe along the preset lifting-translation-lowering path. During the transfer process, the jaws maintain a preset tilt angle; Before the truss manipulator reaches the lathe station, communicate with the control system in real time to obtain the position coordinates of the current loading port of the lathe and the opening state of the fixture; The truss manipulator fine-tunes the jaw position according to the acquired information and accurately places the workpiece in the lathe fixture.

[0016] Preferably, when performing multi-process integrated machining, it includes: Based on the structural characteristics of the cylinder for a three-cylinder compressor, the control system pre-programs to generate a master-slave parallel machining path; For the different process requirements of turning and drilling, the control system synchronously adjusts the parameters in real time; After machining, use the contact probe built into the lathe to perform on-line detection of the flatness of the turned plane and the position accuracy of each hole through a composite path combining global grid scanning and local feature fine measurement. If the tolerance is exceeded, trigger multiple compensation programs based on the tolerance.

[0017] Preferably, when adjusting the workpiece direction, it includes: After the previous process is completed, the workpiece is placed on the positioning platform of the turning transfer device by the truss robot. Through the pre-programmed coordinate matching between the truss robot and the positioning platform, the initial position of the workpiece on the turning transfer device is fixed, providing a reference for direction adjustment. Through the real-time communication between the controller of the turning transfer device and the lathes of the previous and subsequent processes, the target angle θ1 required for the subsequent process is obtained, and the current rotation angle θ0 identified by the workpiece positioning module is synchronously received, and the required adjustment amount Δθ = θ1 - θ0 is calculated. The workpiece is fixed by the pneumatic clamping component on the rotating platform equipped with the turning transfer device, and the servo motor drives the rotating platform according to Δθ to complete the direction adjustment of the workpiece within the preset time. If tilt adjustment is required, the tilt axis built in the pneumatic clamping component moves synchronously. After the adjustment is completed, an industrial camera equipped with the workpiece positioning module is used to collect images of the preset marks on the workpiece surface, and the actual angle θ2 of the marks is calculated through visual algorithms. If the actual angle θ2 is lower than the set value compared with the target angle θ1, it is determined that the adjustment is qualified. If it is out of tolerance, the device automatically triggers secondary adjustment until the accuracy meets the standard.

[0018] Preferably, when real-time monitoring the tool usage status and dynamically adjusting the machining parameters, it includes: The control system collects tool status data in real time through multi-type sensors integrated in the turret. The control system performs fusion analysis on vibration, current, and temperature data through a machine learning model, including multi-modal data time-frequency domain feature fusion and dynamic evolution of the random forest model, and outputs tool status labels. Based on the tool status labels output by the machine learning model, the control system combines the sensor data features to subdivide the abnormal types. For the subdivided abnormal types, the control system dynamically adjusts the machining parameters according to the principle of minimum intervention - quality priority - efficiency balance to maintain the machining quality.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: First, the coordination between positioning and feeding is improved, reducing the defective rate and production line waiting. The disk magazine adopts core axis radial - circumferential composite positioning (restricting horizontal offset and rotational angle deviation), combined with real-time monitoring by contact sensors (sampling frequency ≥ 10Hz), to ensure that the workpiece is accurately fixed in the preset direction, avoiding manual secondary adjustment, eliminating the direction deviation caused by traditional single-dimensional positioning, and greatly reducing the defective rate.

[0020] Dynamic feeding rhythm adjustment mechanism (the matching error between the actual feeding rhythm and the robotic arm grasping cycle ≤ ±1 second) and pre-feeding buffer position design are adopted to achieve zero-wait synchronization of "grasping immediately upon arrival" for the truss robotic arm, basically eliminating the problems of material accumulation or waiting in the production line feeding process, and significantly improving the equipment utilization rate.

[0021] II. Integrated processing of multiple processes, significantly shortening the single-piece production cycle The power turret numerical control lathe integrates 6 types of processes such as turning the plane, drilling, and tapping (the 6 independent tool positions of the turret can be expanded). Through the master-slave parallel processing path planning (the overlapping time of turning the plane and the processes of positioning holes and threaded holes ≥ 4 seconds), the traditional step-by-step processing of multiple processes is optimized into integrated processing, greatly shortening the single-piece processing cycle and significantly improving the production efficiency; at the same time, the secondary clamping of workpieces is reduced, effectively reducing the clamping error, and significantly improving the processing consistency.

[0022] III. Efficient and accurate direction adjustment between processes, ensuring the quality of subsequent processing The turning and transfer device drives the rotating platform through a servo motor and uses visual verification for closed-loop control, replacing the traditional manual flipping or simple rotating table, ensuring that the processing orientation in subsequent processes is accurately adapted, avoiding hole position deviation caused by direction adjustment errors, and thus improving the qualified rate of products.

[0023] IV. Enhanced protection of cutting fluid and reliability of grippers, extending the equipment life The gripper adopts a composite structure of corrosion-resistant alloy + elastic sealing layer, combined with the design of a diversion groove (groove depth 0.5 - 0.7 mm) and an oil-repellent coating (contact angle ≥ 110°), effectively guiding the diversion of cutting fluid and preventing adhesion, thereby extending the service life of the gripper; during the transfer process, the gripper maintains an inclination angle of 3° - 5°, basically eliminating the problem of cutting fluid dripping onto the guide rail and reducing the wear rate of the guide rail. At the same time, a pressure sensor is built into the gripper to dynamically control the grasping force (5 - 8 N), avoiding over-tight deformation or over-loose detachment, and significantly improving the grasping reliability.

[0024] V. Intelligent monitoring of tool status and rapid response to abnormalities, reducing downtime losses Multi-modal data fusion monitoring: The time-frequency domain features of vibration (band energy ratio), current (harmonic distortion rate), and temperature (change rate) are fused (generating multi-dimensional feature vectors), combined with a random forest classifier (the initial training data n ≥ 2000 groups), and the accuracy of tool abnormality recognition is significantly improved compared with traditional single sensors; the dynamic evolution of the model (identifying new types of abnormalities) avoids missed detections and further reduces the defective product rate.

[0025] Precise adjustment of subdivision anomalies: According to the anomaly type (initial wear / medium wear / severe wear, etc.), adopt the "minimum intervention-quality first-efficiency balance" strategy. For example, for initial wear, only adjust the feed rate and increase the cutting fluid flow (without affecting efficiency); for severe wear, only stop the abnormal tool (other tools continue to run). Compared with the traditional "speed reduction and shutdown" strategy, the effective operation time of the production line is increased, and the tool replacement cost is reduced.

[0026] In summary, through multi-dimensional technological innovation, the present invention realizes high precision, high efficiency, and high reliability in the production of components of a three-cylinder compressor, providing key technical support for the large-scale manufacturing of high-efficiency compressors.

[0027] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. Description of the Drawings

[0028] Figure 1 It is a flowchart of the production method steps of components of a high-efficiency three-cylinder compressor according to an embodiment of the present invention; Figure 2 It is a flowchart of the conveying process of a workpiece facing a robotic arm according to an embodiment of the present invention; Figure 3 It is a flowchart of the monitoring process of the cooperation state between a workpiece and a mandrel according to an embodiment of the present invention; Figure 4 It is a flowchart of the conveying process of a workpiece facing a lathe according to an embodiment of the present invention. Specific Embodiments

[0029] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0030] At the same time, it should be understood that for the convenience of description, the dimensions of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0031] The description of at least one exemplary embodiment below is actually only illustrative and in no way limits the present application and its application or use.

[0032] For technologies, methods, and devices known to those of ordinary skill in the relevant art, detailed discussions may not be made, but where appropriate, the said technologies, methods, and devices should be regarded as part of the specification.

[0033] Embodiment 1. The production system for high - energy - efficiency three - cylinder compressor components provided by the present invention includes: a disc magazine, a power turret CNC lathe, a truss robot, a turning and transfer device, a gripper, a workpiece positioning module, and a control system.

[0034] The disc magazine is used to place and position the cylinders for the three - cylinder compressor to be processed, and automatically supply materials to the truss robot. The power turret CNC lathe is configured to integrate multi - process machining functions, and can perform machining of the flat surface, positioning holes, threaded holes, lateral suction holes, spring holes, and enthalpy - increasing holes of the cylinders for the three - cylinder compressor. The truss robot is cooperatively arranged with the disc magazine, the power turret CNC lathe, and the turning and transfer device, and is used for the automatic transfer of workpieces. The turning and transfer device is arranged between adjacent processing procedures, and is configured to adjust the direction of the workpiece within a preset time to shorten the subsequent processing preparation time. The gripper is installed at the end of the truss robot and is used to grab the workpiece, and its structural design can reduce the damage of the cutting fluid to the workpiece and the gripper itself. The workpiece positioning module integrates a laser displacement sensor, an industrial camera, and a vision processing unit, and is installed on the truss robot and the turning and transfer device. It is configured to perform multi - source data error characteristic modeling, dynamic adjustment of Kalman filter parameters, adaptive correction of reflectivity compensation parameters, and verification of positioning confidence, and output the final positioning coordinates and angle compensation values to identify the position of the workpiece. The control system is communicatively connected to the disc magazine, the power turret CNC lathe, the truss robot, the turning and transfer device, the gripper, and the workpiece positioning module, and is configured to perform fusion analysis including time - frequency domain feature fusion of multi - modal data and dynamic evolution of the random forest model to obtain the tool usage status. When detecting an abnormal tool usage status (such as wear, overheating), it adjusts the processing parameters to ensure product quality.

[0035] In a possible embodiment, the disc magazine is configured with a mandrel, a rotary indexing device, and a contact sensor; the contact sensor is integrated at the key position where the side of the mandrel contacts the inner hole of the workpiece, and the sampling frequency ≥10Hz to ensure real - time capture of the contact state between the workpiece and the mandrel.

[0036] In a possible embodiment, the workpiece positioning module is equipped with a laser displacement sensor and an industrial camera. The laser displacement sensor is used to scan the surface of the workpiece to obtain the initial spatial coordinates and contour features of the workpiece. The industrial camera is used to collect images of the workpiece within the pre - positioning area.

[0037] In a possible embodiment, the truss robot is equipped with joint servo motors for dynamically adjusting the posture of the gripper.

[0038] In a possible embodiment, the jaw adopts a composite structure of a corrosion-resistant alloy matrix (such as 316L stainless steel) + an elastic sealing layer (such as fluororubber). A diversion groove (groove depth 0.5 - 0.7 mm, groove width 1 - 2 mm) is arranged on the inner side of the jaw, and an oil-repellent coating (contact angle ≥ 110°) is wrapped on the outside. When grasping, the jaw guides the cutting fluid remaining on the surface of the workpiece to flow to both sides through the diversion groove, and the oil-repellent coating prevents the cutting fluid from adhering to the surface of the jaw.

[0039] In a possible embodiment, a contact probe is built into the power turret numerical control lathe for on-line detection of the flatness of the machined plane and the positional accuracy of each hole.

[0040] In a possible embodiment, the turning and transfer device is equipped with a rotating platform and a pneumatic clamping assembly. The pneumatic clamping assembly is installed on the rotating platform and has an inclined shaft built in.

[0041] In a possible embodiment, the turret of the power turret numerical control lathe is evenly distributed with 6 independent tool positions (which can be expanded according to processing requirements) along the circumference, and a face turning tool (such as a carbide turning tool), a positioning hole drill bit (φ3 mm high-speed steel drill bit), a thread hole tap (M6×1), a lateral suction hole drill bit (φ5 mm deep hole drill), a spring hole milling cutter (R2 mm ball-end milling cutter), and an enthalpy-increasing hole reamer (φ8H7) are respectively installed. The installation angles and positions of each tool are pre-adjusted through three-dimensional simulation (such as the angle between the cutting plane of the turning tool and the axis of the drill bit is 90°) to ensure no spatial interference during synchronous processing.

[0042] In a possible embodiment, the positioning platform is built in with a groove adapted to the outer contour of the workpiece and contact positioning blocks (the material is polyurethane to avoid scratching the workpiece).

[0043] In a possible embodiment, the power turret numerical control lathe is equipped with a vibration sensor, a current sensor, and a temperature sensor. Among them, the vibration sensor is installed on the tool holder of the turret, the current sensor is connected in series in the circuit of the turret drive motor, and the temperature sensor is embedded at the joint of the tool tip and the tool holder.

[0044] Example 2, referring to Figure 1 the production method step diagram of high-efficiency three-cylinder compressor parts, the present invention provides a Figure 1 production method of high-efficiency three-cylinder compressor parts as shown in Step S1, feeding and positioning: Manually place the cylinder for the three-cylinder compressor in the disc magazine in a preset direction, position the workpiece through the core shaft of the disc magazine, and automatically supply the workpiece to the truss robot arm. Step S2, workpiece position finding and transfer: After quickly determining the position of the workpiece by using the workpiece position finding module, the truss robot arm uses the jaw to grasp the workpiece and transfer it to the power turret numerical control lathe. Step S3, multi-process integrated machining: The turning of the plane, positioning holes, threaded holes, lateral suction holes, spring holes, and enthalpy-increasing holes are performed by a power turret CNC lathe; Step S4, inter-process direction adjustment: Between adjacent machining processes, the workpiece direction is adjusted by a turning transfer device within a preset time to provide an adapted orientation for subsequent machining; Step S5, machining process monitoring: The control system monitors the tool usage status in real time. When abnormalities (such as wear, temperature exceeding the threshold) are detected, the machining parameters (such as feed rate, rotational speed) are adjusted to ensure the quality stability of the product; Step S6, automated cycle: Repeat the above steps to complete the automated machining of the cylinder for a three-cylinder compressor, without manual intervention in the middle.

[0045] In the above Step S1, when positioning the workpiece and automatically feeding it to the truss robot arm, it includes: By inserting the mandrel into the inner hole of the workpiece, radial positioning (restricting horizontal offset) and circumferential positioning (restricting rotational angle deviation) of the workpiece are achieved simultaneously, fixing the workpiece on the surface of the disk magazine in a preset machining direction, avoiding the operation of manually adjusting the direction twice; Through the rotary indexing device configured on the disk magazine, according to the grasping cycle of the truss robot arm, the positioned workpieces are conveyed to the grasping area of the truss robot arm at a preset rhythm (such as every 15 - 20 seconds / piece); During the feeding process, the cooperation state between the workpiece and the mandrel is monitored by a contact sensor (such as a pressure sensor). If it is detected that the workpiece is not fully sleeved on the mandrel, the disk magazine is alarmed and the feeding is paused to avoid the risk of defective products caused by unpositioned workpieces entering the machining process; Among them, the grasping area is pre-programmed and matched with the movement trajectory of the truss robot arm, and the truss robot arm can directly perform the grasping action without additional path adjustment.

[0046] Refer to Figure 2 the conveying flow chart of the workpiece facing the robot arm. In a possible embodiment, when being conveyed to the grasping area of the truss robot arm, it includes: Through the real-time interaction between the magazine controller and the control system, the actual grasping cycle T1 of the truss robot arm is obtained; Compare the difference ΔT (ΔT = T1 - T0) between the actual grasping cycle T1 and the theoretical grasping cycle T0 pre-stored in the magazine controller, and dynamically adjust the rotational speed of the rotary indexing device so that the matching error between the actual feeding rhythm T2 and the actual grasping cycle T1 is ≤ ±1 second, avoiding the accumulation or waiting of feeding caused by the temporary acceleration or deceleration of the truss robot arm (such as after fault recovery); Define the grasping area of the truss robot arm as the zero-position grasping point (X3, Y3, Z3) and pre-programmatically bind it to the initial reset position of the truss robot arm and the starting coordinates of the movement trajectory. When the disc hopper is feeding, the positioned workpiece is pre-sent to the cache position of the grabbing area by a beat cycle (such as 15 seconds) in advance through the rotary indexing device. After the truss robot arm completes the grabbing of the previous workpiece and returns to the initial position, the workpiece is pushed from the cache position to the zero-position grabbing point through the rotary indexing device, realizing zero-wait synchronization of the truss robot arm grabbing as soon as it is in place; Among them, N workpiece placement positions are evenly distributed on the silo surface along the circumferential direction of the disc silo, and each placement position corresponds to a unique number (1-N); the silo controller records the number of the currently fed workpiece (such as No. k) through a counter, and dynamically plans the feeding queue according to the grasping sequence of the truss robot arm (such as sequential grasping 1→2→3... or interval grasping 1→3→5...), ensuring that there is always and only one workpiece to be grasped in the grasping area, avoiding the risk of misgrasping by the robot arm due to multiple workpieces entering the grasping area at the same time.

[0047] Background description: In the automated production process of three-cylinder compressor parts (such as cylinders), the coordination between the feeding system and the gripping action of the robotic arm is a key link that affects the production line efficiency and processing stability. In traditional production technology, the feeding rhythm of the disc silo and the gripping cycle of the truss robotic arm mainly rely on fixed parameter pre-settings and lack a dynamic matching mechanism, leading to the following problems: disconnection between the feeding and gripping cycles, significant loss of production line efficiency, and weak abnormal response capabilities. Based on this: In a possible embodiment, a pre-compensation mechanism based on a periodic fluctuation trend is used to dynamically adjust the rotation speed of the rotary indexing device, including: Establish a historical grasping cycle database to record the actual grasping cycle sequence T1(n) within M consecutive cycles (n=1,2,…,M, M≥5); The periodic fluctuation trend value ΔT_trend is calculated through the sliding window algorithm to reflect the long-term change trend of the crawling cycle; When ΔT (current cycle difference) and ΔT_trend have the same sign, pre-compensation adjustment is triggered: If ΔT>+1 second and ΔT_trend>0 (grasping cycle continues to extend), the rotation speed of the rotary indexing device is reduced by a factor of k1 (0.8≤k1<1), and the pre-delivery time of the next workpiece is extended by ΔT_trend×t1 (t1 is the time compensation coefficient, unit: seconds / trend unit); If ΔT<-1 second and ΔT_trend<0 (grasping cycle continues to shorten), increase the speed of the rotary indexing device by a factor of k2 (1 <k2≤1.2),并将下一工件的预送时间缩短|ΔT_trend|×t2(t2为时间补偿系数,单位:秒 / 趋势单位); Among them, k1, k2, t1, and t2 are compensation parameters pre-calibrated according to the mechanical characteristics of the device (such as the maximum acceleration and deceleration of the rotary indexing device and the motion inertia of the truss robotic arm), and are determined by fitting experimental data.

[0048] In the embodiments of the present invention, it needs to be further explained that the above pre-compensation mechanism reduces the matching error between the actual feeding beat T2 and the actual grasping cycle T1, and shortens the response time of the system to sudden cycle fluctuations (such as the accelerated operation after the robotic arm temporarily fails and recovers).

[0049] Refer to Figure 3 the flowchart of the workpiece and mandrel mating state monitoring. In a possible embodiment, when monitoring the mating state of the workpiece and the mandrel, it includes: The control system pre-stores the standard pressure threshold P0 (such as 5 - 8N, set according to the workpiece weight and the interference fit tolerance between the inner hole and the mandrel) matching the cylinder of the three-cylinder compressor; When the disk magazine starts, first calibrate the zero point of the contact sensor through no-load testing (no workpiece is sleeved) to eliminate the baseline drift caused by environmental vibration or the self-weight of the mandrel; When the workpiece is sleeved on the mandrel, the contact pressure value P is collected in real time through the contact sensor. If P ≥ P0, it is determined that the workpiece is fully sleeved. If P < P0 (such as P ≤ 3N), it is determined that the workpiece is not fully sleeved; Among them, when it is detected that P < P0, the magazine controller triggers a three-level response mechanism: First-level audible and visual alarm: The three-color indicator light (red / yellow / green) on the top of the disk magazine switches to red flashing, and at the same time the buzzer emits a high-frequency alarm (frequency 800 - 1000Hz, duration ≥ 2 seconds) to prompt the operator to check; Second-level feeding pause: Immediately stop the operation of the rotary indexing device to prevent unpositioned workpieces from entering the grasping area of the truss robotic arm; Third-level robotic arm linkage: Send a pause grasping instruction to the truss robotic arm through an industrial communication bus (such as PROFINET). After receiving the instruction, the truss robotic arm stops the current grasping action and stays in place to avoid mis-grasping unpositioned workpieces; After the operator confirms that the workpiece is abnormally sleeved (such as the workpiece is tilted or the inner hole is blocked) and re-places it, a reset instruction is triggered through the operation panel of the disk magazine; The pressure value P is re-collected through the contact sensor. If P ≥ P0, the alarm is turned off, the operation of the rotary indexing device is restored, and a resume grasping instruction is sent to the truss robotic arm to achieve a quick restart after the exception is handled.

[0050] In the above step S2, when using the workpiece positioning module to quickly determine the workpiece position, it includes: The surface of the workpiece is scanned by a laser displacement sensor equipped in the workpiece positioning module to obtain the initial spatial coordinates (X0, Y0, Z0) and contour features of the workpiece, quickly narrowing the positioning range to within ±2 mm to provide a pre-positioning area for visual fine positioning; An industrial camera equipped in the workpiece positioning module is used to collect images of the workpiece within the pre-positioning area (acquisition frequency ≥ 30 frames per second), and a feature recognition algorithm (such as edge detection based on HOG features + SIFT feature point matching) is run through the built-in vision processing unit to identify the key positioning features of the workpiece (including inner hole edges, end face marking lines, preset positioning holes), and calculate the accurate coordinates (X1, Y1, Z1) of the workpiece and the inclination angle θ of the workpiece; The initial spatial coordinates (X0, Y0, Z0) and the accurate coordinates (X1, Y1, Z1) are input into a data fusion algorithm (such as Kalman filtering), combined with the workpiece material reflectivity compensation parameter (for the reflective interference caused by the residual cutting fluid), and the final positioning coordinates (X, Y, Z) and the angle compensation value φ are output; The final positioning coordinates (X, Y, Z) and the angle compensation value φ are transmitted to the control system through the workpiece positioning module to control the truss robot arm to dynamically adjust the grasping path to ensure the precise alignment of the gripper with the part of the workpiece to be grasped.

[0051] Background description: In the automated production of components such as cylinders of a three-cylinder compressor, the precise positioning of the workpiece is the core prerequisite for ensuring subsequent machining accuracy (such as flatness, hole position accuracy). In traditional technologies, workpiece positioning mainly relies on a single sensor (such as a laser displacement sensor or an industrial camera). The cylinders of a three-cylinder compressor are mostly made of aluminum alloy (high reflectivity) or cast iron (low reflectivity), and the surface often remains with residual cutting fluid after machining (the transparent liquid is easy to form specular reflection). When the laser irradiates the workpiece with high reflectivity or the reflective area of the cutting fluid, the sensor is prone to signal saturation due to excessive echo; while for materials with low reflectivity (such as cast iron), due to weak echo, signal loss is likely to occur. Visual positioning depends on the key features on the surface of the workpiece (such as inner hole edges, marking lines), but the residual cutting fluid will blur the feature boundaries; at the same time, the surface contrast of materials with low reflectivity (such as cast iron) is low, which is likely to lead to failure in feature extraction. Therefore, the existing workpiece positioning is difficult to meet the manufacturing requirements of high-efficiency compressors in terms of positioning accuracy and stability. Based on this: In a possible embodiment, when outputting the final positioning coordinates (X, Y, Z) and the angle compensation value φ, it includes: Modeling of the error characteristics of multi-source data: Pre-establish an error characteristic database for the laser displacement sensor and the industrial camera, where: The error model of the laser displacement sensor is: ΔL = k3·d + k4·ρ + εL (where d is the measured distance, ρ is the reflectivity of the workpiece surface, k3 and k4 are calibration coefficients, and εL is the first random noise); The error model of the industrial camera is: ΔC = m1·θ + m2·σ1 + εC (where θ is the inclination angle of the workpiece, σ1 is the light reflection intensity of the cutting fluid, m1 and m2 are calibration coefficients, and εC is the second random noise); Among them, the error model is determined by fitting through repeated positioning experiments (n≥50 times) of standard parts (such as aluminum alloy cylinders with known surface reflectivity).

[0052] Dynamic adjustment of Kalman filter parameters: Input the initial spatial coordinates (X0, Y0, Z0) and the precise coordinates (X1, Y1, Z1) as state variables into the Kalman filter. The state vector is defined as [X, Y, Z, φ] (where φ is the angle compensation value); Dynamically adjust the process noise covariance matrix Q according to the current workpiece material (such as aluminum alloy / cast iron): If the material is high-reflectivity aluminum alloy (ρ≥0.6), increase the noise variance of the Z-axis in Q (σ²_Z = 0.05mm²) to suppress the laser overexposure error; if it is low-reflectivity cast iron (ρ≤0.3), increase the noise variance of the X-Y plane (σ²_XY = 0.03mm²) to compensate for the fuzzy visual feature recognition. Dynamically correct the measurement noise covariance matrix R according to the residual degree of the cutting fluid: Calculate the proportion S of the reflective area (S = number of reflective pixels / total number of pixels) through the gray histogram of the industrial camera image. If S≥30% (severe reflection), increase the noise variance of the visual coordinates in R (σ²_C = 0.02mm²); if S≤10% (slight reflection), decrease the noise variance of the laser coordinates in R (σ²_L = 0.01mm²).

[0053] Adaptive correction of reflectivity compensation parameters: The reflectivity ρ of the workpiece material is calculated through the echo intensity I of the laser displacement sensor: ρ = α·I + β (where α and β are sensor calibration coefficients); For the reflection interference caused by the residual cutting fluid, collect the RGB image of the workpiece surface through the industrial camera, extract the average gray value G of the blue channel (the characteristic color of the cutting fluid), establish the compensation coefficient γ = 1 - k5·G (k5 is an empirical coefficient, with a value range of 0.01 - 0.03), and the final reflectivity compensation parameter is ρ' = ρ·γ; If the positioning errors (position deviations detected after actual grasping) of three consecutive workpieces are all >0.03mm, trigger the self-learning of compensation parameters: Correct the values of α, β, and k5 through the least squares method to improve the matching degree of ρ' and the actual reflection effect to ≥95%.

[0054] Position confidence verification: When the final positioning coordinates (X, Y, Z) and the angle compensation value φ are output, the confidence index C is synchronously calculated based on the positioning error output by the error model; If C < 0.8 (low confidence), secondary positioning is triggered: the laser displacement sensor rescans the workpiece surface (the scanning density is increased by 50%), the industrial camera switches to the near-focus mode (the resolution is increased from 1 million pixels to 2 million pixels), and the data fusion process is re-executed until C ≥ 0.8 and then the result is output.

[0055] In the embodiments of the present invention, it needs to be further explained that the "data fusion algorithm and the dynamic correction process of the reflectivity compensation parameter" proposed in this embodiment realizes the deep fusion of laser and vision data by establishing a multi-sensor error model, dynamically adjusting the filtering parameters, adaptively correcting the reflectivity compensation value, and introducing a confidence verification mechanism, and solves the problem of insufficient accuracy caused by light reflection interference, material differences, and cutting fluid residues in traditional positioning technologies.

[0056] Refer to Figure 4 the conveying flow chart of the workpiece facing the lathe. In step S2 above, when transferring to the power turret lathe, it includes: After the truss robot arm receives the final positioning coordinates (X, Y, Z) and the angle compensation value φ output by the workpiece positioning module, it dynamically adjusts the gripper posture through the joint servo motor to make the grasping surface of the gripper fully fit with the part of the workpiece to be grasped (such as the outer cylindrical surface of the cylinder). At the same time, the pressure sensor built in the gripper monitors the grasping force in real time. When the grasping force reaches the preset threshold (such as 5 - 8N), the gripper is locked to avoid workpiece deformation caused by over-tightening or dropping caused by over-loosening; After the truss robot arm grasps the workpiece, it transfers to the power turret lathe along the preset lifting - translation - lowering path. During the transfer process, the gripper maintains an inclination angle of 3° - 5° (the drainage groove faces downwards), so that the residual cutting fluid continuously drains to the workshop waste liquid collection system (integrated with the production line) through the drainage groove and from the gripper surface, avoiding the cutting fluid dripping onto the machine tool guide rail during the transfer process; Before the truss robot arm reaches the lathe station, it communicates with the control system in real time through the industrial bus (such as PROFINET) to obtain the position coordinates (X2, Y2, Z2) of the current loading port of the lathe and the fixture opening state; The truss robot arm fine-tunes the gripper position according to the obtained information, accurately places the workpiece in the lathe fixture, and then the gripper releases and returns to the initial position, completing the automated loading without manual intervention.

[0057] In step S3 above, during multi-process integrated machining, it includes: Based on the structural characteristics of the cylinder for a three-cylinder compressor (such as the end face diameter and the distribution positions of various holes), the control system pre-programs and generates a master-slave parallel machining path: taking the operation of turning the plane as a benchmark (machining time T3 = 8 - 10 seconds), the operations of positioning holes and threaded holes start 1 - 2 seconds after the turning tool contacts the workpiece end face (machining time T4 = 6 - 8 seconds), so that the machining time of the operation of turning the plane overlaps with that of the operations of positioning holes and threaded holes by ≥ 4 seconds; the lateral operations of the lateral suction holes, spring holes, and enthalpy-increasing holes start when 50% of the plane turning is completed, realizing the time coordination of multi-directional machining; According to the different process requirements of turning and drilling, the control system adjusts the parameters in real-time and synchronously: when turning the plane, the rotational speed is maintained at 1500 rpm and the feed rate is 0.1 mm / r; when drilling, the rotational speed is maintained at 800 rpm and the feed rate is 0.05 mm / r; If it is detected that the hardness fluctuation of the workpiece material exceeds the threshold (fed back by the cutting force sensor), the feed rates of the turning tool and the drill bit are synchronously adjusted (such as when the hardness increases by 10%, the feed rate decreases by 0.01 mm / r) to ensure that the surface roughness of the turned plane and the position tolerance of the drilled holes meet the standards synchronously; After the machining is completed, the flatness of the turned plane (tolerance ≤ 0.05 mm) and the position of each hole (such as the relative position tolerance between the threaded hole and the lateral suction hole ≤ 0.1 mm) are on-line detected by the contact probe built in the lathe. If the tolerance is exceeded, the abnormal tool number is automatically recorded and the compensation program is triggered to realize the closed-loop precision control of synchronous machining.

[0058] Background description: Traditional detection mostly adopts "single-point sampling" or "random sampling in key areas" (such as only detecting 4 points at the center and four corners of the plane), and the covered area is less than 30% of the workpiece surface. For areas such as the cylinder end face that are prone to deformation due to cutting force (such as local protrusions at the edge due to tool deflection), single-point detection cannot capture the true deviation of the flatness (the actual maximum deviation may occur in the untested edge area), resulting in a large error in flatness evaluation and unable to provide an accurate basis for tool compensation. In addition, traditional detection only collects the center point coordinates of key hole positions such as threaded holes and lateral suction holes, ignoring the position deviations in different directions around the holes (such as the offset of one side of the hole due to drill bit yaw). For example, if the actual deviation of the hole position is "0.12 mm to the left and 0.03 mm to the right", only measuring the center point will result in the wrong conclusion of "deviation 0.075 mm" (the actual position tolerance exceeds the standard), resulting in the compensation program not being triggered or being wrongly adjusted, and there is still a risk of hole position deviation in subsequent machining. The detection path of the traditional contact probe is a fixed parameter and cannot adapt to cylinder workpieces of different sizes. Based on this: In a possible embodiment, when on-line detecting the flatness of the turned plane and the position of each hole by adopting a composite path combining global grid scanning and local feature fine measurement, it includes: Global scanning: Taking the center of the turning plane as the origin, collect the height data of N points (N≥20) at a grid spacing of 5mm×5mm (covering a plane area ≥90%), and calculate the flatness (fitting the reference plane by the least squares method and taking the maximum and minimum height differences); Local precise measurement: For each hole position (such as threaded hole, lateral suction hole), evenly distribute 4 detection points around the hole, collect the hole position coordinates (X_i, Y_i), and calculate the position tolerance (by the root mean square value of the coordinate deviation in the reference coordinate system); The detection path is pre-programmed and generated by the control system, and the grid spacing is dynamically adjusted according to the workpiece size (such as the plane diameter D) (spacing 3mm when D≤50mm, spacing 5mm when D>50mm) to ensure the balance between detection coverage and efficiency.

[0059] In the embodiments of the present invention, it needs to be further explained that the above composite detection path combining global grid scanning and local feature precise measurement ensures the comprehensiveness of flatness evaluation through grid scanning covering more than 90% of the plane area, captures the true deviation of the position tolerance through precise measurement of multiple points around the hole, and dynamically adjusts the detection parameters according to the workpiece size, thoroughly solving the problems of incomplete coverage and poor adaptability of traditional detection, and providing accurate data support for closed-loop compensation.

[0060] Background description: In traditional technologies, when it is detected that the flatness or the position tolerance of the hole exceeds the tolerance, compensation is only triggered through simple threshold judgment (such as "stop when exceeding the tolerance"), without analyzing the specific type of the out-of-tolerance (such as is the flatness out-of-tolerance due to tool wear or abnormal cutting parameters? Is the single-hole position tolerance out-of-tolerance due to drill bit deviation or fixture looseness?). This "one-size-fits-all" compensation method often leads to: When the flatness exceeds the tolerance, the drill bit parameters are wrongly adjusted (unrelated to the turning tool), and the compensation is ineffective; When the single-hole position tolerance exceeds the tolerance, the entire set of tools is blindly replaced (actually only a single drill bit is offset), resulting in waste of resources; When the multi-hole correlation exceeds the tolerance, the fixture is not calibrated (actually the fixture is loose), and subsequent workpieces continue to exceed the tolerance.

[0061] In a possible embodiment, a closed-loop precision control for synchronous machining is implemented by adopting multi-class compensation based on out-of-tolerance, including: When the flatness exceeds the tolerance (>0.05mm): It is determined that there is tool wear or abnormal cutting parameters, automatically retrieve the historical machining data of the turning tool (such as cumulative cutting time, cutting depth), calculate the tool compensation value Δd through a linear regression model, and increase the Z-axis compensation value of the turning tool by Δd; When the single-hole position tolerance exceeds the tolerance (>0.1mm): It is determined that the positioning of the corresponding drill bit is offset, calculate the X / Y-axis compensation amounts Δx, Δy of the drill bit through coordinate transformation, and adjust the initial positioning coordinates of the drill bit; Multi-hole position tolerance correlation out-of-tolerance (≥2 hole position deviation directions are the same): It is determined that the lathe fixture is loose or the reference coordinate system is offset, triggering the fixture calibration program (repositioning the fixture zero point through a standard part), and recording the fixture number and calibration time.

[0062] In the embodiments of the present invention, it needs to be further explained that the above-mentioned multi-category compensation based on out-of-tolerance classifies out-of-tolerance into three categories: flatness out-of-tolerance, single-hole position tolerance out-of-tolerance, and multi-hole correlation out-of-tolerance, and triggers corresponding compensation, thereby improving compensation accuracy, resource utilization rate, and quality consistency, and solving the core contradiction that the compensation strategy does not match the problem in the traditional technology.

[0063] In step S4 above, when adjusting the workpiece direction, it includes: After the previous process (such as OP1) is completed, the workpiece is placed on the positioning platform of the turning transfer device by the truss robot. Through the pre-programmed coordinate matching between the truss robot and the positioning platform, the initial position of the workpiece on the turning transfer device is fixed, providing a reference for direction adjustment. Through the real-time communication between the controller of the turning transfer device and the lathes of the previous and subsequent processes (such as OP1 lathe, OP2 lathe), the processing requirement data of the subsequent process (such as the target angle θ1 required by the subsequent process) is obtained, and the initial orientation data (such as the current rotation angle θ0) identified by the workpiece positioning module is synchronously received, and the required adjustment amount Δθ = θ1 - θ0 is calculated; (if OP2 needs to perform machining of the flat surface, positioning holes, threaded holes, lateral suction holes, spring holes, and enthalpy-increasing holes under the reverse plane of the workpiece, the target adjustment angle is 180° (rotation around the Z axis)); The workpiece is fixed (to prevent rotation deviation) by the pneumatic clamping component on the rotating platform equipped with the turning transfer device (clamping force 10 - 15N, response time ≤0.3 seconds), and the servo motor drives the rotating platform according to Δθ, so that the workpiece completes the direction adjustment within a preset time (such as ≤5 seconds). If tilt adjustment is required (such as OP3 needs to machine a 45° inclined hole), the built-in tilt axis of the pneumatic clamping component moves synchronously to achieve multi-dimensional orientation adaptation of horizontal rotation + tilt. After the adjustment is completed, the industrial camera equipped with the workpiece positioning module is used to collect images of the preset marks (such as the crosshair marked by laser) on the workpiece surface, and the actual angle θ2 of the marks is calculated through visual algorithms. If the actual angle θ2 is less than or equal to the target angle θ1 by the set value (such as |θ2 - θ1| ≤ 0.5°), the adjustment is determined to be qualified. If it is out-of-tolerance, the device automatically triggers secondary adjustment (repeating the above steps) until the accuracy meets the standard, avoiding subsequent machining deviation caused by adjustment errors. The adjustment time (≤5 seconds) of the U-turn transfer device is pre-matched with the processing cycle of the previous and next processes (such as 15 seconds for OP1 processing and 15 seconds for OP2 processing). After the adjustment is completed, the truss robot arm immediately grabs the workpiece and transfers it to the subsequent process; if the previous process is completed ahead of schedule (such as OP1 only takes 12 seconds), the U-turn transfer device automatically enters the standby cache mode (keeping the workpiece orientation unchanged) and waits for the truss robot arm to trigger the transfer command to avoid equipment idling or workpiece backlog due to beat fluctuations; If the workpiece is detected to be loose during the adjustment process (feedback by the pressure sensor of the pneumatic clamping component, the pressure value is <8N) or the visual verification is out of tolerance for three consecutive times (|θ2-θ1|>0.5°), the U-turn transfer device will immediately trigger an alarm (three-color light flashes red + buzzer), and suspend the previous process processing and the adjustment work of the U-turn transfer device through the control system to prevent the workpiece that has not been adjusted into place from flowing into the subsequent links; after the operator checks for abnormalities (such as failure of the pneumatic clamping component, failure of the rotating platform, mark wear) and repairs them, the U-turn transfer device will be restarted to automatically clear the error record and resume the normal adjustment process.

[0064] In the above step S5, real-time monitoring of the tool usage status and adjustment of the processing parameters include: The control system collects tool status data in real time through multiple types of sensors integrated in the turret: Vibration sensor: monitors high-frequency vibration signals during tool cutting and identifies abnormal vibrations caused by tool wear; Current sensor: monitors the motor load current and reflects changes in tool cutting resistance (such as tool wear, chip blockage or abnormal workpiece hardness); Temperature sensor: monitors the cutting temperature of the tool to prevent tool annealing failure caused by excessive cutting heat; The control system uses machine learning models (such as random forest classifiers) to fuse and analyze vibration, current, and temperature data and output tool status labels; The control system subdivides the abnormality types based on the tool status labels output by the machine learning model and the sensor data features (such as vibration frequency band energy, current harmonic distortion rate, and temperature change rate); According to the subdivided abnormal types, the control system dynamically adjusts the processing parameters to maintain the processing quality.

[0065] Background description: In the precision machining of three-cylinder compressor cylinder parts, real-time and accurate monitoring of tool status is the core link to ensure machining quality (such as flatness, hole position accuracy) and production line stability. In traditional technologies, tool status monitoring mainly relies on a single sensor (such as a current sensor) or simple data superposition analysis, resulting in low accuracy in tool status recognition. Based on this: In a possible embodiment, when performing fusion analysis, it includes multi-modal data time-frequency domain feature fusion and dynamic evolution of the random forest model; Among them, for multi-modal data time-frequency domain feature fusion: cross-scale feature extraction is performed on the original data of vibration, current, and temperature sensors: Vibration signal: Decompose it into 6 frequency bands from 10 - 500 Hz through continuous wavelet transform (CWT), and extract the energy ratio of each frequency band (reflecting the high-frequency impact of tool edge micro-cracking) and kurtosis value (reflecting the non-Gaussian characteristics of abnormal impact); Current signal: Extract the root mean square value of the low-frequency component from 0 - 10 Hz (reflecting the motor load fluctuation caused by tool wear) and the harmonic distortion rate of the medium-frequency component from 10 - 50 Hz (reflecting the periodic load change caused by chip jamming) through STFT (Short-Time Fourier Transform); Temperature signal: Calculate Δ_T / Δt (temperature change rate) and T_max (highest temperature) through a sliding window (window length 60 s), reflecting the degree of cutting heat accumulation (tool annealing risk); Finally, a fusion vector of multi-dimensional features (such as vibration frequency band energy ratio, current harmonic distortion rate, temperature change rate, etc.) is generated as the input of the random forest classifier.

[0066] Among them, for the dynamic evolution of the random forest model: The initial model is trained based on historical data (n ≥ 2000 groups, covering all stages of the tool life cycle), including 100 decision trees, and each tree corresponds to an abnormal mode (initial wear, medium wear, severe wear, chip entanglement, workpiece material abnormality); When it is detected that the "unrecognized state" is output continuously for 3 times (that is, the model cannot match the known abnormal modes), trigger model evolution: Screen 10% of the unrecognized data (samples with the largest feature differences) through the active learning algorithm and push them to the operator for abnormal type annotation (such as "tool coating peeling off"); Update the random forest in an incremental learning manner (retain the original tree structure, add 20 new trees for new abnormal classification), and eliminate redundant features (features with importance < 0.05) through feature importance ranking (Gini index); Through this mechanism, the model has the ability to identify new abnormal types, and the model volume remains stable.

[0067] In the embodiments of the present invention, it needs to be further explained that in the above fusion analysis process, through multi-time scale feature extraction, the model can capture the full-dimensional features of the tool state (such as the vibration high-frequency energy ratio and the current harmonic distortion rate), and the recognition accuracy is greatly improved; through model dynamic evolution, the recognition ability of new abnormal types (such as coating peeling) is obtained, avoiding the increase in the defective product rate caused by missed detection, and solving the technical problems of "one-sided feature extraction, poor model adaptability, and fuzzy abnormal attribution" in traditional tool state monitoring.

[0068] Background description: In the precision machining of cylinder parts of a three-cylinder compressor, the adjustment strategy of machining parameters when a tool anomaly occurs is the core contradiction point in balancing quality stability and production line efficiency. In traditional technologies, when a tool anomaly (such as increased vibration or current) is detected, only a unified adjustment strategy of "reducing speed and stopping" or "changing the tool" is triggered through simple threshold judgment, without distinguishing the specific type of anomaly. Traditional adjustments only focus on a single parameter (such as reducing the rotational speed) and do not consider the synergistic effect of multiple parameters (such as the linked adjustment of the feed rate, cutting depth, and cooling parameters), resulting in the technical problems of "excessive intervention affecting efficiency" or "insufficient intervention leading to quality out of control". Based on this: In a possible embodiment, the abnormal types can be subdivided into: Initial wear (label A): The tool has micro-cracking (the energy ratio of the 100 - 500 Hz frequency band of vibration > 15%), and the cutting force fluctuation is small (the root mean square value of the current ≤ 11 A); Moderate wear (label B): The tool is dull (the energy ratio of the 50 - 100 Hz frequency band of vibration > 25%), and the motor load increases (the root mean square value of the current is 11 - 12 A); Severe wear (label C): The tool wear amount > 0.3 mm (the temperature change rate ΔT / Δt > 2 °C / s), and the cutting heat increases significantly (the temperature T > 180 °C); Chip blockage (label D): The chip winds around the tool head (the current harmonic distortion rate > 15%), and the vibration signal shows periodic impacts (the energy ratio of the 10 - 50 Hz frequency band of vibration > 30%); Insufficient cutting fluid (label E): Cooling failure (the temperature T > 200 °C), and the feedback value of the cutting fluid flow sensor < 80% of the rated flow rate.

[0069] In a possible embodiment, the control system dynamically adjusts the machining parameters according to the principle of minimum intervention - quality priority - efficiency balance, including: Initial wear (label A): Only adjust the feed rate F of the corresponding tool, keep the rotational speed S and the cutting depth ap unchanged, delay wear without reducing efficiency; synchronously increase the cutting fluid flow rate Q to 110% of the rated flow rate, and reduce the accumulation of cutting heat by strengthening cooling.

[0070] Medium wear (Label B): Coordinately adjust the feed rate and rotational speed to reduce the cutting load; adjust the cutting depth ap to 90% of the rated cutting depth to reduce the single-cutting amount; trigger the turret pre-tightening program to increase the hydraulic clamping force and prevent the tool from loosening and aggravating wear.

[0071] Severe wear (Label C): Immediately stop the current tool machining (rotational speed S = 0), but keep other tools (such as an unworn drill bit) running (independent control of multiple turrets); adjust the machining parameters of the unworn tool (such as the drill bit rotational speed S = 800 rpm → 700 rpm, feed rate F = 0.05 mm / r → 0.04 mm / r) to avoid the interruption of the entire process due to the abnormality of a single tool.

[0072] Chip clogging (Label D): Pause the cutting fluid supply (to prevent chip adhesion), start the reverse rotation program (S = -500 rpm, for 2 seconds), and throw off the entangled chips by reverse rotation; adjust the chip conveyor rotational speed (from 100 rpm to 150 rpm) to enhance the chip removal ability; after resuming machining, reduce the feed rate by 15% to reduce the chip generation amount.

[0073] Insufficient cutting fluid (Label E): Immediately turn off the cutting fluid pump (to prevent dry cutting from damaging the tool), start the standby cutting fluid tank switching program (switching time ≤ 2 seconds); adjust the rotational speed (such as the lathe tool rotational speed S = 1500 rpm → 1200 rpm) to reduce the cutting heat generation rate.

[0074] In the embodiments of the present invention, it needs to be further explained that the above dynamic adjustment process accurately maps the abnormal characteristics and adjustment strategies through "subdividing the abnormal types + three principles of adjustment", and guides the decision-making with a clear priority, solving the core contradictions of "inaccuracy, non-coordination, and unscientificness" in parameter adjustment in the traditional technology.

[0075] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A production system for high-energy-efficiency three-cylinder compressor components, characterized in that, Including: A disk silo for placing and positioning the cylinders of a three-cylinder compressor to be processed and automatically feeding the truss robot arm; A power turret CNC lathe configured to integrate multi-process machining functions; A truss robot arm, cooperatively arranged with the disk silo, the power turret CNC lathe, and the turning and transfer device, for automatic transfer of workpieces; A turning and transfer device, arranged between adjacent processing procedures, configured to adjust the direction of the workpiece within a preset time; Jaws, installed at the end of the truss robot arm, for gripping the workpiece, and the structural design can reduce the damage of cutting fluid to the workpiece and the jaws themselves; A workpiece positioning module, integrating a laser displacement sensor, an industrial camera, and a vision processing unit, and installed on the truss robot arm and the turning and transfer device, configured to perform multi-source data error characteristic modeling, dynamic adjustment of Kalman filter parameters, adaptive correction of reflectivity compensation parameters, and verification of positioning confidence, and output the final positioning coordinates and angle compensation values to identify the position of the workpiece; A control system, communicatively connected to the disk silo, the power turret CNC lathe, the truss robot arm, the turning and transfer device, the jaws, and the workpiece positioning module, configured to perform fusion analysis including multi-modal data time-frequency domain feature fusion and dynamic evolution of a random forest model to obtain the tool usage status and adjust the processing parameters.

2. A production method for high-energy efficiency three-cylinder compressor parts based on the production system described in claim 1, characterized in that, Including the following steps: Manually place the cylinders of the three-cylinder compressor in the disk silo in a preset direction, position the workpiece through the core shaft of the disk silo, and automatically feed the truss robot arm; After quickly determining the position of the workpiece using the workpiece positioning module, use the jaws of the truss robot arm to grip the workpiece and transfer it to the power turret CNC lathe; Perform machining of the flat surface, positioning holes, threaded holes, lateral suction holes, spring holes, and enthalpy-increasing holes through the power turret CNC lathe; Between adjacent processing procedures, adjust the direction of the workpiece within a preset time through the turning and transfer device to provide an adapted orientation for subsequent processing; The control system based on multi-modal data monitors the tool usage status in real time through a machine learning model and dynamically adjusts the processing parameters to complete the automatic processing of the cylinders of the three-cylinder compressor.

3. The production method of the high-energy efficiency three-cylinder compressor parts according to claim 2, characterized in that, When positioning the workpiece and automatically feeding the truss robot arm, it includes: By inserting the core shaft into the inner hole of the workpiece, simultaneously realizing the radial positioning and circumferential positioning of the workpiece, and fixing the workpiece on the surface of the disk silo in a preset processing direction; Through the rotary indexing device configured in the disk silo, according to the gripping cycle of the truss robot arm, convey the positioned workpiece to the gripping area of the truss robot arm at a preset beat; During the feeding process, monitor the mating state of the workpiece and the core shaft through a contact sensor. If it is detected that the workpiece is not fully sleeved on the core shaft, trigger an alarm of the disk silo and pause the feeding; Wherein, the contact sensor is integrated at the key position where the side of the core shaft contacts the inner hole of the workpiece.

4. The production method of the high-energy efficiency three-cylinder compressor component according to claim 3, characterized in that, When conveying to the gripping area of the truss robot arm, it includes: Through the real-time interaction between the silo controller and the control system, obtain the actual gripping cycle T1 of the truss robot arm; Compare the difference ΔT between the actual grasping cycle T1 and the theoretical grasping cycle T0 pre-stored in the bin controller, and adopt a pre-compensation mechanism based on the cycle fluctuation trend to dynamically adjust the rotation speed of the rotary indexing device, so that the matching error between the actual feeding rhythm T2 and the actual grasping cycle T1 is lower than the set value; Define the grasping area of the truss manipulator as the zero-position grasping point and pre-program it to be bound with the initial reset position of the truss manipulator and the starting coordinate of the motion trajectory; Through the rotary indexing device, the positioned workpiece is pre-fed to the buffer position in the grasping area one beat cycle in advance. After the truss manipulator completes the grasping of the previous workpiece and returns to the initial position, the workpiece is pushed from the buffer position to the zero-position grasping point by the rotary indexing device.

5. The production method of the high energy efficiency three-cylinder compressor parts according to claim 3, characterized in that, When monitoring the mating state of the workpiece and the mandrel, it includes: Pre-store the standard pressure threshold P0 matching the cylinder of the three-cylinder compressor in the control system; When the disk bin starts, calibrate the zero point of the contact sensor through no-load testing; When the workpiece is sleeved on the mandrel, the contact pressure value P is collected in real time through the contact sensor. If P≥P0, it is determined that the sleeve is fully inserted. If P<P0, it is determined that the sleeve is not fully inserted; Among them, when it is detected that P<P0, the bin controller triggers a three-level response mechanism.

6. The production method of the high energy efficiency three-cylinder compressor parts according to claim 2, characterized in that, When using the workpiece positioning module to quickly determine the workpiece position, it includes: Use a laser displacement sensor to scan the surface of the workpiece to obtain the initial spatial coordinates and contour features of the workpiece, and provide a pre-positioning area for visual fine positioning; Use an industrial camera to collect images of the workpiece in the pre-positioning area, and run a feature recognition algorithm through the visual processing unit to identify the key positioning features of the workpiece, and calculate the accurate coordinates and the inclination angle of the workpiece; Input the initial spatial coordinates and accurate coordinates into the data fusion algorithm, perform multi-source data error characteristic modeling, dynamic adjustment of Kalman filter parameters, adaptive correction of reflectivity compensation parameters, and verification of positioning confidence, and output the final positioning coordinates and angle compensation values; Transmit the final positioning coordinates and angle compensation values to the control system through the workpiece positioning module to control the truss manipulator to dynamically adjust the grasping path.

7. The production method of the high-energy efficiency three-cylinder compressor component according to claim 6, characterized in that When transferring to a power turret lathe, it includes: After the truss manipulator receives the final positioning coordinates and angle compensation values, dynamically adjust the jaw posture through the joint servo motor; After the truss manipulator grasps the workpiece, it transfers to the power turret lathe along the preset lifting-translation-lowering path. During the transfer process, the jaws maintain the preset inclination angle; Before the truss manipulator reaches the lathe station, communicate with the control system in real time to obtain the position coordinates of the current loading port of the lathe and the opening state of the fixture; The truss manipulator fine-tunes the jaw position according to the obtained information and accurately places the workpiece in the lathe fixture.

8. The production method of the high-energy efficiency three-cylinder compressor parts according to claim 2, characterized in that, When performing multi-process integrated machining, it includes: Based on the structural characteristics of the cylinder of the three-cylinder compressor, the control system pre-programs and generates a master-slave parallel machining path; For the different process requirements of turning and drilling, the control system synchronously adjusts the parameters in real time; After the machining is completed, use the contact probe built into the lathe to perform on-line detection of the flatness of the turned plane and the position accuracy of each hole by using a composite path of global grid scanning combined with local feature fine measurement. If the tolerance is exceeded, trigger various compensation programs based on the tolerance.

9. The production method of the high-energy efficiency three-cylinder compressor parts according to claim 2, characterized in that, When adjusting the workpiece orientation, it includes: After the previous process is completed, the workpiece is placed on the positioning platform of the turning transfer device by the truss robot arm; Through the pre-programmed coordinate matching between the truss robot arm and the positioning platform, the initial position of the workpiece on the turning transfer device is fixed, providing a reference for orientation adjustment; Through the real-time communication between the controller of the turning transfer device and the lathes of the previous and subsequent processes, the target angle θ1 required for the subsequent process is obtained, and the current rotation angle θ0 identified by the workpiece positioning module is synchronously received, and the required adjustment amount Δθ = θ1 - θ0 is calculated; The workpiece is fixed by the pneumatic clamping component on the rotating platform equipped with the turning transfer device, and the servo motor drives the rotating platform according to Δθ to complete the orientation adjustment of the workpiece within the preset time; If tilt adjustment is required, the tilt axis built in the pneumatic clamping component moves synchronously; After the adjustment is completed, the industrial camera equipped with the workpiece positioning module is used to collect images of the preset marks on the surface of the workpiece, and the actual angle θ2 of the marks is calculated through the vision algorithm; If the actual angle θ2 is lower than the set value compared with the target angle θ1, it is determined that the adjustment is qualified. If it is out of tolerance, the device automatically triggers secondary adjustment until the accuracy meets the standard.

10. The production method of the high energy efficiency three-cylinder compressor parts according to claim 2, characterized in that, When the tool usage status is monitored in real time and the machining parameters are dynamically adjusted, it includes: The control system collects the tool status data in real time through various types of sensors integrated in the turret; The control system performs fusion analysis on the vibration, current, and temperature data through the machine learning model, including multi-modal data time-frequency domain feature fusion and dynamic evolution of the random forest model, and outputs the tool status label; Based on the tool status label output by the machine learning model and combined with the sensor data characteristics, the control system classifies the abnormal types in detail; For the classified abnormal types, the control system dynamically adjusts the machining parameters according to the principle of minimum intervention - quality priority - efficiency balance to maintain the machining quality.

Images