Numerical control machine tool real-time sensing data compensation analysis method based on end-cloud collaboration

Abstract

The invention provides a numerical control machine tool real-time sensing data compensation analysis method based on end-cloud collaboration in the technical field of numerical control machine tool compensation algorithms, which comprises the following steps of: acquiring and caching multi-dimensional sensing data through an edge end high-speed signal acquisition card, and regularly reading the sensing data in batches by a cloud server; the optimal sampling frequency and period are designed to eliminate the inherent frequency error between the edge end and the cloud server, the clock communication error caused by decimal truncation is considered, compensation is started when the accumulative error exceeds a set range, and the influence of the cloud server clock jitter and the communication priority on the data acquisition quality is considered to adjust the compensation in real time. And finally, multi-scale unified multi-source sensing data is formed through fusion of a Lagrange interpolation algorithm, multi-source data analysis of real-time error compensation is realized, and a basis is provided for subsequent process analysis and health management of a numerical control machine tool.

Description

Technical Field

[0001] The present invention relates to the technical field of compensation algorithms for CNC machine tools, and specifically to a real-time sensor data compensation and analysis method for CNC machine tools based on end-cloud collaboration. Background Art

[0002] With the rapid improvement of the integration degree of CNC machine tools and the complexity of processing tasks, under the premise of ensuring computing efficiency and scheduling capabilities, it is a research focus to overcome the influence of communication fluctuations of multi-source signals of CNC machine tools under real-time high-speed transmission conditions, enhance the numerical accuracy and timing consistency in the process of CNC machine tool sensor data acquisition, and make the sensor data truly and comprehensively reflect the operation process of CNC machine tool equipment in complex manufacturing systems.

[0003] At present, the quality of sensor data acquisition for CNC machine tools is usually assumed to be stable and reliable, but this assumption depends largely on the matching of the CNC machine tool signal sampling card clock and the cloud server computing task signal clock. With the continuous increase in the sampling frequency of CNC machine tools and the continuous increase in computing task resource consumption, CNC machine tools are currently prone to significant sampling errors in the actual sampling process due to factors such as task priority, inherent frequency, and rounding operations, which makes the machine tool sensor data under multi-channel significantly distorted and difficult to cope with the complex and changeable actual industrial conditions and process optimization needs. Existing methods usually assume the integrity and authenticity of the machine tool sensor data acquisition quality, and then guide the process optimization and equipment monitoring during the machine tool processing process, without fully considering the data sensing errors caused by the dynamic relationship between sampling frequency, system load, and data real-time.

[0004] Therefore, it is urgent to integrate the multi-source sensor data of CNC machine tools through the real-time data acquisition capability of the edge, and perform unified compensation for clock accuracy errors, critical periodic errors, and communication priority errors. Then, the Lagrangian difference algorithm is used at the edge to build a multi-dimensional sensor data set with a unified main frequency, and establish a real-time association mapping relationship between labels of specific inputs and outputs to support the multi-dimensional sensor analysis of CNC machine tools in the cloud server, and provide a real and reliable data source support for the subsequent CNC system process optimization and machine tool equipment health management. Summary of the invention

[0005] The purpose of the present invention is to provide a real-time sensor data compensation and analysis method for CNC machine tools based on end-cloud collaboration, which integrates multi-source sensor data of CNC machine tools through the real-time data acquisition capability of the edge, collects and compensates multi-dimensional sensor information of cutting force, position, and speed in real time, and effectively maps and analyzes sensor data from different channel sources.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is to provide a method for real-time sensor data compensation and analysis of CNC machine tools based on end-cloud collaboration, which is characterized by comprising the following steps:

[0007] Step S1: The high-speed signal acquisition card at the edge of the CNC machine tool equipment issues a real-time sampling instruction. After the sensor data is written into the edge sampling cache, the timer at the computing resource of the cloud server in the workshop will periodically issue a communication instruction to read the sampling cache in batches. The amount of cache data read in each batch during this process is defined as N, and the initial sampling frequency of the high-speed signal acquisition card at the edge is f. s , then the timer period at the cloud server is

[0008] Step S2: Design single sampling data volume index N c , so that the maximum cache capacity N can be N c divisibility;

[0009] Step S3: Minimum resolution around the data refresh cycle of the edge high-speed signal acquisition card Minimum resolution of the communication cycle with the cloud server timer The inherent frequency difference between the two makes it difficult to design the backup resolution of the high-speed signal acquisition card. Alternate resolution with cloud server timer With the initial sampling frequency as f s The optimal frequency Initial input, iterative update Until the backup resolution of the high-speed signal acquisition card is satisfied Integer requirement;

[0010] Step S4: When the Under the constraints of integers, further iterations are performed Enables alternate resolution of cloud server timer It also satisfies the integer requirement. When searching for all and After finding the optimal frequency set that all meet the integer requirement, select the smallest one As sampling frequency to eliminate the inherent frequency error between the edge high-speed signal acquisition card and the cloud server timer;

[0011] Step S5: Around The accumulated clock error caused by the decimal truncation error of the high-speed signal acquisition card at the edge under the sampling frequency is used to calculate the optimal frequency at the edge. The optimal collection cycle under Define the actual communication collection period at the edge due to automatic truncation of the decimal part as T k′ , then the edge clock resolution error is obtained And define the edge clock resolution cumulative error after n rounds of communication cycles

[0012] Step S6: Defining an additional delay amount that can correct the clock cumulative accuracy error When the cumulative error is first detected Beyond After the error compensation is started, the current cycle clock time is changed from T now Update to T now + To eliminate the accumulated error, thus completing the error compensation of the edge clock accuracy channel;

[0013] Step S7: Further consider the instability error caused by the excessive jitter amplitude of the timer system clock at the cloud server, and record the issuance time T of the cloud server sampling task start With end time T end , define the jitter range of the cloud server timer clock as , then when the cloud server communication program first obtains the edge cache sampling data, the remaining sampling time of the sampling period is T remain =Tc-(T end -T start ), and the remaining time of the current cycle T remain Determined as the communication delay T for the next cycle delay ;

[0014] Step S8: Further compare T delay and The relationship between the size of the two is used to evaluate whether the communication delay is within the compensable allowable range. If the judgment condition is not met, the delay time of the next cycle is adjusted cyclically to keep the communication delay correction within the preset range; if the judgment condition is not met, that is, This indicates that the current delay time is insufficient for effective compensation, which will automatically trigger a skip report and terminate the error compensation process at the current stage, thereby completing the error compensation of the critical periodic channel;

[0015] Step S9: For the clock communication error caused by insufficient priority of the communication collection task at the edge end, record the communication start time of each task cycle at the edge end The start time of the previous task cycle And get the running time interval between two consecutive sampling programs like This indicates that the current communication clock has a priority error due to the preemption of other high-priority tasks, so before the next communication cycle starts, the machine tool sensor data collection amount is adjusted to mN c,in Indicates a rounding down operation, thereby completing the error compensation of the communication priority channel;

[0016] Step S10: While ensuring the accuracy of real-time compensation of sensor data sampling, the sensor data of machine tools of different scales are adjusted to a unified reference scale for correction and fusion, and the sensor sample points (t0, y0) and (t1, y1) for which real-time error compensation has been achieved are supplemented with the arbitrary interpolation point time step label t through the first-order Lagrange interpolation function;

[0017] Step S11: by increasing the time scale frequency and simulation sampling density of different sensor data, the machine tool sensor data of each time scale reaches the main frequency time scale, and unified fusion on the time scale is achieved based on the consistent sampling frequency, so that the multi-source sensor data after real-time error compensation is collaboratively calculated;

[0018] Step S12: Use the time dimension as a link to associate the sensor data of the same time scale to form an associated data set containing multi-dimensional information, and select the link data dimension as the analysis benchmark to construct an associated analysis data set with this dimension as the target label to support subsequent cloud server data applications.

[0019] Furthermore, the sampling buffer in step S1 adopts a ring buffer structure to achieve continuous data collection and storage.

[0020] Furthermore, the edge high-speed signal acquisition card in step S1 has multi-channel acquisition capability and uses TCP / IP protocol for data transmission to support simultaneous data acquisition of multiple sensors.

[0021] Furthermore, the communication interface between the cloud server computing resources in the workshop and the CNC machine tools in step S1 is OPC UA.

[0022] Furthermore, the optimal acquisition period in step S5 is is 200μs.

[0023] Furthermore, the cloud server timer clock jitter range in step S7 is The value is [10,20]μs.

[0024] Furthermore, the cyclic adjustment process includes dynamically adjusting the delay time to adapt to real-time changes in communication conditions, and the preset jump range is set to [20, 40] μs.

[0025] Furthermore, the increase in the time scale frequency in step S111 is based on system performance requirements and sensor characteristics, and uses interpolation or resampling technology to improve the time resolution of the data.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] 1. Taking into full consideration the dynamic relationship between sampling frequency, system load and data real-time, the multi-source sensor data of CNC machine tools is integrated through the real-time data acquisition features of the edge and the high-speed signal acquisition card to perform unified compensation for clock accuracy error, critical periodicity error and communication priority error.

[0028] 2. A real-time correlation mapping relationship between specific inputs and outputs is established through a one-dimensional difference algorithm, and a multi-dimensional sensor data set design method is constructed. The multi-dimensional sensor data can be integrated into the cloud server analysis module to provide a real and reliable data source support for the health management analysis of CNC machine tools and the process optimization decision-making of CNC systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a flow chart of the real-time sensor data compensation and analysis method for CNC machine tools in an embodiment of the present invention.

[0030] Figure 2 It is a schematic diagram of the overall process of multi-dimensional correlation fusion analysis of real-time sensor data of CNC machine tools in an embodiment of the present invention.

[0031] Figure 3 This is a comparison chart of the effects of the real-time sensor data compensation strategy for CNC machine tools in an embodiment of the present invention.

[0032] Figure 4 This is a diagram showing the effect of analyzing the sensor data of a CNC machine tool in an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0034] See also Figure 1 and Figure 2 The present invention provides a method for real-time sensor data compensation and analysis of CNC machine tools based on end-cloud collaboration, comprising the following steps:

[0035] S1: The high-speed signal acquisition card at the edge of the CNC machine tool equipment issues a real-time sampling command. After the sensor data is written into the edge sampling cache, the sampling cache adopts a ring cache structure to achieve continuous data collection and storage. The timer at the cloud server computing resource in the workshop will periodically issue communication instructions to read the sampling cache in batches. The amount of cache data read in each batch during this process is defined as N, and the initial sampling frequency of the edge high-speed signal acquisition card is f s , then the timer period at the cloud server is The communication interface between the cloud server computing resources and the CNC machine tools in the workshop is Object Linking and Embedding for Process Control (OPC UA). The edge high-speed signal acquisition card has multi-channel acquisition capabilities and uses TCP / IP protocol for data transmission to support simultaneous data acquisition from multiple sensors.

[0036] S2: Design single sampling data volume index N c , ensuring that the maximum cache capacity N can be N c Divisible.

[0037] S3: Minimum resolution around the data refresh cycle of the edge high-speed signal acquisition card Minimum resolution of the communication cycle with the cloud server timer The inherent frequency difference between the two makes it difficult to design the backup resolution of the high-speed signal acquisition card. Alternate resolution with cloud server timer With the initial sampling frequency as f s The optimal frequency Initial input, iterative update Until the backup resolution of the high-speed signal acquisition card is satisfied Integer requirement.

[0038] S4: In meeting Under the constraints of integers, further iterations are performed Enables alternate resolution of cloud server timer It also satisfies the integer requirement. When searching for all and After finding the optimal frequency set that all meet the integer requirement, select the smallest one As the sampling frequency, it can eliminate the inherent frequency error between the edge high-speed signal acquisition card and the cloud server timer.

[0039] S5: Surrounded by The accumulated clock error caused by the decimal truncation error of the high-speed signal acquisition card at the edge under the sampling frequency is used to calculate the optimal frequency at the edge. The optimal collection cycle under Define the actual communication collection period at the edge due to automatic truncation of the decimal part as T k′ , then the edge clock resolution error is formed And define the edge clock resolution cumulative error after n rounds of communication cycles Preferably, it is 200 μs.

[0040] S6: Defines the amount of additional delay that can correct for clock accuracy errors When the cumulative error is first detected Beyond Then start the error compensation algorithm and change the current cycle clock time from T now Updated to To eliminate the accumulated error, thus completing the error compensation of the clock accuracy channel.

[0041] S7: Further consider the instability error caused by the excessive jitter of the timer system clock at the cloud server, and record the issuance time T of the cloud server sampling task start With end time T end , define the jitter range of the cloud server timer clock as When the cloud server communication program first obtains the edge cache sampling data, the remaining sampling time of the sampling cycle is T remain =T c -(T end -T start ), and the remaining time of the current cycle T remain Determined as the communication delay for the next cycle Cloud server timer clock jitter range The value is [10,20]μs.

[0042] S8: Further comparison with T delay and The relationship between the size of the two is used to evaluate whether the communication delay is within the compensable allowable range. If the condition is met, the delay time of the next cycle is adjusted cyclically. The cyclic adjustment process includes dynamically adjusting the delay time to adapt to the real-time changes in communication conditions. The preset jump range is set to [20,40]μs. The communication delay correction is kept within a reasonable jump range to achieve more accurate clock synchronization; if the judgment condition is not met, that is, This indicates that the current delay time is insufficient for effective compensation, which will automatically trigger a skip report and terminate the error compensation process at the current stage, thereby completing the error compensation of the critical periodic channel.

[0043] S9: In view of the clock communication error caused by insufficient communication acquisition task priority at the edge, record the communication start time of each task cycle at the edge The start time of the previous task cycle And get the running time interval between two consecutive sampling programs like This indicates that the current communication clock has a priority error due to the preemption of other high-priority tasks, so before the next communication cycle starts, the machine tool sensor data collection amount is adjusted to mN c,in Represents a round-down operation, thereby completing the error compensation of the communication priority channel.

[0044] S10: While ensuring the accuracy of real-time compensation of sensor data sampling, the sensor data of machine tools of different scales are adjusted to a unified reference scale for correction and fusion. Through the first-order Lagrangian interpolation function, the sensor sample points (t0, y0) and (t1, y1) that have achieved real-time error compensation are completed with the arbitrary interpolation point time step label t.

[0045] S11: By increasing the time scale frequency and simulation sampling density of different perception data, the machine tool sensor data of each time scale can reach the main frequency time scale, and unified fusion on the time scale is achieved based on the consistent sampling frequency, so that the multi-source sensor data after real-time error compensation can be collaboratively calculated.

[0046] S12: Using the time dimension as a link, the sensor data of the same time scale are mutually associated to form an associated data set containing multi-dimensional information, and the link data dimension is selected as the analysis benchmark to construct an associated analysis data set with this dimension as the target label.

[0047] The following uses the Kistler9171A dynamometer and high-speed signal card to monitor and collect the milling force generated during the milling process of TC4 titanium alloy as an example to further illustrate the specific implementation and effect of the invention.

[0048] In one embodiment, the spindle speed of the milling cutter used for TC4 titanium alloy is 1368 rpm, so the initial sampling frequency of the real-time sampling instruction issued by the high-speed signal acquisition card at the edge of the CNC machine tool equipment is set to f s =136.8Hz, and the amount of cache data read in each batch is N = 30 bytes, then the timer period T at the cloud server c =30 / 136.8=219ms. At the same time, set the single edge sampling data volume index N c = 3 bytes to ensure that N can be c Divisible.

[0049] Calculate the minimum resolution of the data refresh cycle of the edge high-speed signal acquisition card And set the minimum resolution of the cloud server timer communication cycle With the initial sampling frequency as f s =136.8Hz is the initial input of the optimal frequency, iterative update The optimal frequency is obtained when Make the backup resolution of the high-speed signal acquisition card meet For integer requirements, alternate resolution of the cloud server timer At the same time, it is calculated as 180μs; There is still a cumulative clock error caused by the decimal truncation error itself, and the optimal frequency at the edge is calculated. The optimal collection cycle under The actual communication collection period at the edge due to automatic truncation of the decimal part is T k′ = 180μs, so after n rounds of communication cycles, the edge clock resolution cumulative error is formed The unit is μs, which defines the additional delay that can correct the accumulated accuracy error of the clock. Each time the cumulative error is first detected After exceeding 100μs, error compensation is started and the current cycle clock time is changed from T now Update to T now +100μs to eliminate the accumulated error, thereby completing the error compensation of the edge clock accuracy channel.

[0050] Further considering the instability error caused by the excessive jitter of the timer system clock at the cloud server, the sampling task T start =T now +18.20ms and end time T end =T now +18.24ms as an example, the jitter range of the cloud server timer clock is Then the remaining sampling time T of the cloud server communication program in this sampling period remain =180-(18.24-18.20)*1000=140μs, and the remaining time of the current cycle T remain = 140μs is determined as the communication delay T of the next cycle delay ; Further comparison T delay and The size relationship between them is found It is established, and a cyclic adjustment is performed on the delay time of the next cycle, and the communication delay is corrected from 140μs to be kept within the preset jump range and set to 30μs, thereby completing the error compensation of the critical periodic channel of the sampling task during the 100th communication cycle of the cloud server.

[0051] In view of the clock communication error caused by insufficient priority of communication acquisition tasks at the edge, the communication start time of the 15th task cycle at the edge is used. The start time of the previous task cycle Take the example and get the running time interval of two consecutive sampling programs. satisfy This indicates that the current communication clock has a priority error due to the preemption of other high-priority tasks, so before the next communication cycle starts, the machine tool sensor data collection amount is adjusted to N c =30 bytes, thereby completing the error compensation of the communication priority channel; finally, the sensor data of machine tools of different scales are adjusted to a unified reference scale for correction and fusion. Through the first-order Lagrange interpolation function, taking the time step label t = 10352 as an example,

[0052] The specific verification process includes collecting the milling force in a stable state during the milling process and storing it in the edge memory area, and using the simulated sending program in the high-speed signal acquisition card to output the shared memory signal to the shared memory area of ​​the industrial computer, which acts as a transit cache area after the high-speed signal acquisition card transmits it. Finally, a call instruction is issued through the cloud server to obtain three acquisition programs for real-time sensing data. Among them, program one does not apply the proposed real-time sensing data compensation method, program two only performs error compensation for the communication priority channel, and program three coordinates the deployment of error compensation for the clock accuracy channel, critical periodicity channel, and communication priority channel, thereby comparing and verifying the acquisition data quality and real-time compensation effect of the three programs.

[0053] See also Figure 3 , Figure 3 The compensation results of programs one, two and three for real-time sampling data are shown. The sampling data volume of program one is 24118, which is significantly lower than the 28938 sampling data volume of program two. In contrast, when program three simultaneously applies three types of channel real-time data compensation, the sampling volume of degradation-aware data is 28950 in total, which is exactly the same as the data volume in the shared memory at the edge. This proves that the proposed real-time data error compensation strategy can effectively eliminate the clock error, cycle error and task error in the sampling program, and avoid the loss of sampling data during the degradation process.

[0054] See also Figure 4 , Figure 4 A multi-dimensional machine tool sensor data analysis effect based on real-time data error compensation is provided. The main frequency time scale of the multi-dimensional data is 4kHz. The real-time position feedback data of the Y-axis inside the machine tool is processed, and the missing points of the two main time scale data are interpolated between every two data. The actual position of the Y-axis and the actual position of the Z-axis are selected as target labels, and the size dimensions of the processed weakly rigid thin-walled parts under the Y-axis change are generated as a benchmark; at this time, the milling force information and workpiece deformation information data are correlated and fused through the same time scale, and the final multi-dimensional degradation perception correlation fusion analysis forms the spatial domain and position dimension sensor data analysis information for the Y-axis deformation, which provides data support and basic models for further equipment health management and system intelligent operation and maintenance.

[0055] The present invention proposes a real-time sensor data compensation and analysis method for CNC machine tools based on end-cloud collaboration, which aims to improve the real-time collection and numerical accuracy of sensor data. First, multi-dimensional data is collected and cached by a high-speed signal acquisition card on the edge, and the cloud server reads the sensor data in batches at a regular interval. Then, the optimal sampling frequency and period are designed to eliminate the inherent frequency error between the edge and the cloud server. Subsequently, the clock communication error caused by decimal truncation is considered, and the compensation algorithm is started when the cumulative error exceeds the set range. The influence of cloud server clock jitter and communication priority on the sensor data collection quality is also considered, and real-time adjustment and compensation are performed. Finally, the sensor data from different sources are fused through the Lagrange interpolation algorithm and the multi-dimensional scaling unified method to realize the collaborative calculation of multi-source data after real-time error compensation, which provides a basis for subsequent process analysis and health management of CNC machine tools.

[0056] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art may make some modifications and improvements without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be based on the definition of the claims.

Claims

1. A method for real-time sensor data compensation and analysis of CNC machine tools based on end-cloud collaboration, characterized in that: The following steps are involved: Step S1: The high-speed signal acquisition card at the edge of the CNC machine tool equipment issues a real-time sampling instruction. After the sensor data is written into the edge sampling cache, the timer at the computing resource of the cloud server in the workshop will periodically issue a communication instruction to read the sampling cache in batches. The amount of cache data read in each batch during this process is defined as N, and the initial sampling frequency of the high-speed signal acquisition card at the edge is f. s , then the timer period at the cloud server is Step S2: Design single sampling data volume index N c , so that the maximum cache capacity N can be N c divisibility; Step S3: Minimum resolution around the data refresh cycle of the edge high-speed signal acquisition card Minimum resolution of the communication cycle with the cloud server timer The inherent frequency difference between the two makes it difficult to design the backup resolution of the high-speed signal acquisition card. Alternate resolution with cloud server timer With the initial sampling frequency as f s is the optimal frequency f s * Initial input, iterative update f s * =f s * +1, until the spare resolution of the high-speed signal acquisition card is satisfied Integer requirement; Step S4: When the Under the constraint of integer, further iterative update f s * =f s * +1, made the cloud server timer have alternate resolution It also satisfies the integer requirement. When searching for all and After finding the optimal frequency set that all meet the integer requirement, select the smallest f s * As sampling frequency to eliminate the inherent frequency error between the edge high-speed signal acquisition card and the cloud server timer; Step S5: Around f s * The accumulated clock error caused by the decimal truncation error of the high-speed signal acquisition card at the edge under the sampling frequency is used to calculate the optimal frequency f at the edge. s * The optimal collection cycle under Define the actual communication collection period at the edge due to automatic truncation of the decimal part as T k′ , then the edge clock resolution error is obtained And define the edge clock resolution cumulative error after n rounds of communication cycles Step S6: Defining an additional delay amount that can correct the clock cumulative accuracy error When the cumulative error is first detected Beyond After the error compensation is started, the current cycle clock time is changed from T now Update to T now + To eliminate the accumulated error, thus completing the error compensation of the edge clock accuracy channel; Step S7: Further consider the instability error caused by the excessive jitter amplitude of the timer system clock at the cloud server, and record the issuance time T of the cloud server sampling task start With end time T end , define the jitter range of the cloud server timer clock as When the cloud server communication program first obtains the edge cache sampling data, the remaining sampling time of the sampling cycle is T remain =T c -(T end -T start ), and the remaining time of the current cycle T remain Determined as the communication delay T for the next cycle delay ; Step S8: Further compare T delay and The relationship between the size of the two is used to evaluate whether the communication delay is within the compensable allowable range. If the judgment condition is not met, the delay time of the next cycle is adjusted cyclically to keep the communication delay correction within the preset range; if the judgment condition is not met, that is, This indicates that the current delay time is insufficient for effective compensation, which will automatically trigger a skip report and terminate the error compensation process at the current stage, thereby completing the error compensation of the critical periodic channel; Step S9: For the clock communication error caused by insufficient priority of the communication collection task at the edge end, record the communication start time of each task cycle at the edge end The start time of the previous task cycle And get the running time interval between two consecutive sampling programs like This indicates that the current communication clock has a priority error due to the preemption of other high-priority tasks, so before the next communication cycle starts, the machine tool sensor data collection amount is adjusted to mN c ,in Indicates a rounding down operation, thereby completing the error compensation of the communication priority channel; Step S10: While ensuring the accuracy of real-time compensation of sensor data sampling, the sensor data of machine tools of different scales are adjusted to a unified reference scale for correction and fusion, and the sensor sample points (t0, y0) and (t1, y1) for which real-time error compensation has been achieved are supplemented with the arbitrary interpolation point time step label t through the first-order Lagrange interpolation function; Step S11: by increasing the time scale frequency and simulation sampling density of different sensor data, the machine tool sensor data of each time scale can reach the main frequency time scale, and unified fusion on the time scale is achieved based on the consistent sampling frequency, so that the multi-source sensor data after real-time error compensation can be collaboratively calculated; Step S12: Use the time dimension as a link to associate the sensor data of the same time scale to form an associated data set containing multi-dimensional information, and select the link data dimension as the analysis benchmark to construct an associated analysis data set with this dimension as the target label to support subsequent cloud server data applications.

2. The method according to claim 1, characterized in that The sampling buffer in step S1 adopts a ring buffer structure to achieve continuous data collection and storage.

3. The method according to claim 1, characterized in that The edge high-speed signal acquisition card in step S1 has multi-channel acquisition capability and uses TCP / IP protocol for data transmission to support simultaneous data acquisition of multiple sensors.

4. The method according to claim 1, characterized in that: The communication interface between the cloud server computing resources in the workshop and the CNC machine tool in step S1 is OPC UA.

5. The method according to claim 1, characterized in that The optimal acquisition cycle in step S5 is 200μs.

6. The method according to claim 1, characterized in that The cloud server timer clock jitter range in step S7 The value is [10,20]μs.

7. The method according to claim 1, characterized in that The cyclic adjustment process in step s8 includes dynamically adjusting the delay time to adapt to real-time changes in communication conditions, and the preset jump range is set to [20, 40] μs.

8. The method according to claim 1, characterized in that: The increase in the time scale frequency in step S11 is based on system performance requirements and sensor characteristics, and uses interpolation or resampling technology to improve the time resolution of the data.

Images