A control method and system for a carving and milling machine based on multi-axis vibration correlation analysis

Abstract

The application provides a kind of engraving and milling machine control method and system based on multi-axis vibration correlation analysis, it is related to data analysis processing technical field.The method comprises the following steps: collecting numerical control execution data and processing response data in the processing process of engraving and milling machine and constructing execution fragment sequence, extracting the fragment instruction matrix and instruction response matrix of each execution fragment, generating a plurality of fragment instruction type sets by classifying the instructions of a plurality of execution fragments, constructing the reference response template of each fragment instruction type set to identify a plurality of abnormal response fragments, respectively taking each abnormal response fragment as the endpoint to perform a pre-sequencing backtracking along the execution fragment sequence to identify a state write-in fragment, determining a state trigger fragment according to the abnormal response fragment and the state write-in fragment, constructing a working condition abnormal evolution chain and extracting control call information for engraving and milling machine control strategy generation.The application realizes the improvement of control effect in the processing process of engraving and milling machine.

Description

Technical Field

[0001] This invention relates to the field of data analysis and processing technology, and in particular to a control method and system for a milling and engraving machine based on multi-axis vibration correlation analysis. Background Technology

[0002] CNC engraving and milling machines are widely used in mold processing, precision parts machining, craft engraving, and complex surface machining. During the CNC engraving and milling process, the high-speed rotation of the spindle, the periodic cutting of the tool, and the coordinated operation of multiple feed axes all contribute to machine tool vibration. Vibration not only affects the surface quality, dimensional accuracy, and tool life, but can also lead to local overcutting, chatter marks, chipping, or even machining interruption. Therefore, effective monitoring and control of vibration during the CNC engraving and milling process is necessary.

[0003] Current CNC engraving machine control technologies typically identify abnormal states during machining by collecting vibration signals, spindle load signals, servo current signals, or position following error signals. Control is then implemented through adjustments such as feed rate adjustment, spindle speed adjustment, or alarm shutdown. These technologies mostly focus on analyzing a single moment, a single segment, or a single characteristic quantity, making control decisions primarily based on vibration amplitude, frequency characteristics, or threshold exceedances. While they can reflect the current machining state to some extent, they lack in-depth analysis of the formation process, transmission relationships, and subsequent manifestation patterns of vibration anomalies under multi-axis linkage conditions.

[0004] Especially during continuous machining, certain abnormal conditions do not occur immediately in the current execution segment, but may be gradually introduced by previous machining segments and triggered and manifested in subsequent specific segments. Existing technologies typically struggle to effectively identify such preceding influence relationships and clearly distinguish the writing, transmission, and triggering stages of abnormal conditions. This results in control systems often only making passive adjustments after the abnormality manifests, hindering more targeted early intervention. Therefore, it is necessary to propose a new control method and system for CNC engraving machines to improve the analysis capabilities and control effectiveness for multi-axis vibration-related abnormal conditions. Summary of the Invention

[0005] This invention proposes a control method and system for a CNC engraving and milling machine based on multi-axis vibration correlation analysis, aiming to solve at least one of the technical problems mentioned in the background art.

[0006] To achieve the above objectives, the first aspect of the present invention provides a control method for a CNC engraving and milling machine based on multi-axis vibration correlation analysis, comprising: Collect CNC execution data and machining response data during the CNC milling process, and extract multiple execution segments with sequential relationships from the CNC execution data to construct an execution segment sequence; Extract the segment instruction matrix and instruction response matrix for each execution segment based on the CNC execution data and machining response data, and classify the multiple execution segments according to the multiple segment instruction matrices to generate multiple segment instruction type sets; A reference response template for each instruction type set is constructed based on multiple instruction response matrices. The response residual trajectory of each execution segment is extracted based on the multiple reference response templates, and multiple abnormal response segments are identified based on the response residual trajectory. Each exception response segment is used as the endpoint to perform a pre-order backtracking along the execution segment sequence to identify the state write segment, and the state trigger segment is determined based on the exception response segment and the state write segment. An abnormal working condition evolution chain is constructed based on the state writing fragment, state trigger fragment, and abnormal response fragment. Control call information is extracted from the abnormal working condition evolution chain for use in the generation of the CNC engraving machine control strategy.

[0007] Preferably, the process of extracting the response residual trajectory of each execution segment based on multiple reference response templates, and identifying multiple abnormal response segments based on the response residual trajectory includes: To construct the reference response template for the fragment instruction type set, the centrality parameter of each execution fragment in the fragment instruction type set is calculated. Based on the centrality parameter, multiple candidate normal fragments are selected. The reference response template for the fragment instruction type set is generated by fusing the instruction response matrices corresponding to multiple candidate normal fragments in the fragment instruction type set. The instruction response matrix of each execution segment in the fragment instruction type set is subtracted element by element from the reference response template to obtain multiple element response residual values. The multiple response residual values ​​corresponding to each execution segment are combined to construct the response residual trajectory of the execution segment. For each response residual trajectory, calculate the average residual value. Based on the multiple average residual values, determine the average response threshold for detecting abnormal responses in multiple execution segments within each segment instruction type set. Execution segments with average residual values ​​greater than the average response threshold corresponding to their respective segment instruction type set are marked as abnormal response segments.

[0008] Preferably, performing pre-order backtracking along the execution segment sequence, with each exception response segment as the endpoint, to identify the state-written segment includes: Multiple execution segments are classified according to multiple response residual trajectories to generate multiple segments response type sets. Each execution segment is encoded according to the segment instruction type set and the segment response type set to determine the instruction code and status code of each execution segment. Based on multiple candidate normal segments, identify multiple state transition pairs in the execution segment sequence, extract the state transition triplet of each state transition pair, including the state codes of the two preceding and following candidate normal segments and the instruction code of the following candidate normal segment; A normal state transition library is constructed based on multiple state transition triples. Prequence backtracking is performed on each abnormal response segment according to the normal state transition library and the execution segment sequence to identify the state write segment corresponding to each abnormal response segment.

[0009] Preferably, the pre-order backtracking of each abnormal response segment based on the normal state transition library and the execution segment sequence to identify the state write segment corresponding to each abnormal response segment includes: Construct a normal state set for each abnormal response segment, including extracting the state codes corresponding to multiple candidate normal segments in the segment instruction type set according to the segment instruction type set to which the abnormal response segment belongs, and combining multiple state codes to obtain the normal state set of the abnormal response segment. Multiple candidate preceding segments are identified for each abnormal response segment in the constructed execution segment sequence. State replacement detection is performed on the multiple candidate preceding segments in reverse order to identify the state writing segments in the multiple candidate preceding segments. The state replacement detection for any candidate preceding segment includes: determining the replacement state code of the current candidate preceding segment based on the instruction code of the current candidate preceding segment and the state code of the previous execution segment in the execution segment sequence, and replacing the state code of the current candidate preceding segment. After traversing to the abnormal response segment, if the state code of the abnormal response segment after being replaced belongs to the normal state set of the abnormal response segment, then the candidate preceding segment is recorded as a candidate state and written into the segment. After performing state replacement detection on multiple candidate preceding segments, the candidate state writing segment that is furthest from the abnormal response segment is recorded as the state writing segment corresponding to the abnormal response segment.

[0010] Preferably, extracting control call information from the abnormal operating condition evolution chain for use in the generation of the CNC engraving and milling machine control strategy includes: Based on the instruction codes corresponding to the state writing segment and the state triggering segment in the abnormal working condition evolution chain, the categories of the preceding writing instruction and the working condition triggering instruction are determined, and the control call information corresponding to the abnormal working condition evolution chain is obtained. The control call information is then bound to the preset control strategy corresponding to the abnormal working condition evolution chain. During the controlled processing of the CNC engraving and milling machine, if the segment to be analyzed is detected to match any working condition trigger instruction category, and the segment to be analyzed has a preceding segment that matches the preceding instruction category written in the control call information to which the working condition trigger instruction category belongs, the preset control strategy corresponding to the control call information is invoked to control the CNC engraving and milling machine for processing.

[0011] Preferably, the segment instruction matrix and instruction response matrix for each execution segment are extracted based on the CNC execution data and machining response data. Multiple execution segments are then classified according to the multiple segment instruction matrices to generate multiple sets of segment instruction types, including: Extract the three-axis command position increment sequence and spindle speed sequence for each execution segment from the CNC execution data, and construct the segment command matrix for each execution segment based on the three-axis command position increment sequence and spindle speed sequence; Based on the machining response data, extract the triaxial vibration acceleration sequence, triaxial servo current sequence, and spindle load sequence for each execution segment, and combine the triaxial vibration acceleration sequence, triaxial servo current sequence, and spindle load sequence to construct the instruction response matrix for each execution segment; Based on the fragment instruction matrix, a distance matrix is ​​constructed corresponding to multiple execution fragments. Based on the distance matrix, multiple execution fragments are clustered to generate multiple fragment instruction type sets.

[0012] A second aspect of the present invention provides a control system for a CNC engraving and milling machine based on multi-axis vibration correlation analysis, comprising: The data acquisition module is used to collect CNC execution data and machining response data during the CNC milling process, and extract multiple execution segments with sequential relationships from the CNC execution data to construct an execution segment sequence; The instruction classification module is used to extract the segment instruction matrix and instruction response matrix of each execution segment based on the CNC execution data and machining response data, and to classify multiple execution segments according to multiple segment instruction matrices to generate multiple segment instruction type sets; The anomaly identification module is used to construct a reference response template for each instruction type set based on multiple instruction response matrices, extract the response residual trajectory of each execution segment based on multiple reference response templates, and identify multiple abnormal response segments based on the response residual trajectory. The pre-order backtracking module is used to perform pre-order backtracking along the execution segment sequence with each exception response segment as the endpoint to identify the state write segment, and determine the state trigger segment based on the exception response segment and the state write segment. The control call module is used to construct an abnormal working condition evolution chain based on the state write fragment, state trigger fragment, and abnormal response fragment, and extract control call information from the abnormal working condition evolution chain for use in the generation of the engraving and milling machine control strategy.

[0013] The present invention has the following beneficial effects: This invention divides the execution segments in a continuous machining process and combines CNC execution data and machining response data to construct reference response templates, identify abnormal response segments, trace state writing segments, and construct an abnormal working condition evolution chain. This allows for further identification of the preceding writing relationships and subsequent triggering relationships of abnormal working conditions during the machining process, building upon existing methods that only detect and control single abnormal moments or segments. This provides more targeted control call information for the CNC engraving and milling machine control system, improving the ability to analyze the formation process of abnormal working conditions in complex multi-axis linkage machining, enhancing the timeliness and rationality of subsequent control strategy calls, and ultimately improving the control effect during CNC engraving and milling machine machining. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating a milling machine control method based on multi-axis vibration correlation analysis in an embodiment of the present invention.

[0015] Figure 2 This is a schematic diagram of the structure of a milling machine control system based on multi-axis vibration correlation analysis in an embodiment of the present invention. Detailed Implementation

[0016] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention.

[0017] One aspect of this invention provides a control method for a CNC engraving and milling machine based on multi-axis vibration correlation analysis. This method treats the continuous machining process of the CNC engraving and milling machine as a continuous evolutionary process composed of multiple sequentially connected execution segments. Data is collected synchronously from both the CNC execution process and the machining response process. By jointly analyzing the sequential relationships of the execution segments and the corresponding response changes, the method identifies the writing, transmission, and manifestation relationships of abnormal working conditions during the machining process, thereby providing a more forward-looking basis for subsequent control strategy invocation. Please refer to [link to relevant documentation]. Figure 1 The method specifically includes the following steps: Step S1: Collect CNC execution data and machining response data during the CNC milling process, and extract multiple execution segments with sequential relationships from the CNC execution data to construct an execution segment sequence.

[0018] In this step, the CNC execution data is used to characterize the execution commands issued by the CNC system during the engraving and milling machine's machining process and their actual execution status. It includes at least the command position information for each feed axis, the actual position information for each feed axis, the servo current information for each feed axis, the spindle speed information, the spindle phase information, and the current CNC execution block number information. Preferably, each feed axis includes the X-axis, Y-axis, and Z-axis. The machining response data is used to characterize the actual machining response generated by the engraving and milling machine when executing the corresponding machining command. It includes at least X-axis vibration acceleration data, Y-axis vibration acceleration data, Z-axis vibration acceleration data, and spindle load data. In some embodiments, the machining response data may further include feedback quantities related to the servo drive state to enhance the characterization capability of changes in the machining state.

[0019] It is worth noting that before analyzing the collected data, the CNC execution data and machining response data are first synchronized in time, so that the CNC execution process information and machining response process information at the same sampling moment correspond one-to-one. The synchronized data can simultaneously reflect what machining action the engraving and milling machine performed at the current moment, and what response changes this machining action caused in terms of machine tool structure, spindle load, and multi-axis vibration.

[0020] It is worth noting that this invention does not directly make a holistic judgment on the entire continuous machining data, but rather extracts multiple execution segments with a clear sequential relationship from the CNC execution data. An execution segment refers to a local continuous instruction interval with relatively complete execution significance during the machining process. Its extraction is based on at least one of the following: changes in the CNC execution block number, changes in the movement direction of any feed axis, changes in the spindle speed, and interruptions in the continuity of machining instructions. Through this method, the originally continuous machining process can be divided into multiple execution segments arranged sequentially in time, forming an execution segment sequence. Since the duration and number of sampling points of each segment are different, the length of each segment is first standardized. Each segment is then uniformly resampled from the start to the end point into a fixed length of L sampling points according to the segment's progress. The above-mentioned data preprocessing process is a well-known technique to those skilled in the art, and this embodiment does not specifically limit the specific data cleaning, alignment, or other preprocessing procedures.

[0021] Step S2: Extract the segment instruction matrix and instruction response matrix of each execution segment based on the CNC execution data and machining response data, and classify the multiple execution segments according to the multiple segment instruction matrices to generate multiple segment instruction type sets.

[0022] In this step, for each execution segment, a segment instruction matrix is ​​constructed to characterize the execution mode of the segment, and an instruction response matrix is ​​constructed to characterize the processing response of the segment. The execution mode and execution result corresponding to the segment are distinguished at the data level. Based on this, the execution segments are classified according to the segment instruction matrices of multiple execution segments, resulting in multiple sets of segment instruction types. This allows segments with similar execution modes to be grouped into the same type set, so that subsequent response comparisons are performed between execution segments of the same type, rather than directly comparing segments with different motion modes or different processing stages.

[0023] As an optional implementation process, the above-mentioned process of extracting the fragment instruction matrix and instruction response matrix of the execution fragments, and classifying multiple execution fragments into multiple fragment instruction type sets, specifically includes: Extract the three-axis command position increment sequence and spindle speed sequence for each execution segment from the CNC execution data, and construct the segment command matrix for each execution segment based on the three-axis command position increment sequence and spindle speed sequence.

[0024] For any execution segment, the X-axis, Y-axis, and Z-axis command position data between the start and end sampling points of the execution segment are first acquired. Using the command position at the start of the execution segment as a reference, the X-axis, Y-axis, and Z-axis command position increments of each sampling point within the execution segment relative to the start time are calculated, thus forming a three-axis command position increment sequence. Simultaneously, the spindle speed sampling values ​​within the corresponding time range of the execution segment are extracted to form a spindle speed sequence. Subsequently, the X-axis, Y-axis, and Z-axis command position increment sequences, along with the spindle speed sequence, are combined to construct the segment command matrix corresponding to the execution segment.

[0025] Based on the machining response data, extract the triaxial vibration acceleration sequence, triaxial servo current sequence, and spindle load sequence for each execution segment, and combine the triaxial vibration acceleration sequence, triaxial servo current sequence, and spindle load sequence to construct the instruction response matrix for each execution segment.

[0026] Specifically, for any execution segment, the X-axis vibration acceleration sample values, Y-axis vibration acceleration sample values, and Z-axis vibration acceleration sample values ​​within the corresponding time range of the execution segment are extracted to form a three-axis vibration acceleration sequence; the X-axis servo current sample values, Y-axis servo current sample values, and Z-axis servo current sample values ​​within the corresponding time range of the execution segment are extracted to form a three-axis servo current sequence; and the spindle load sample values ​​within the corresponding time range of the execution segment are extracted to form a spindle load sequence. The three-axis vibration acceleration sequences, three-axis servo current sequences, and spindle load sequences are then combined to form the instruction response matrix corresponding to the execution segment.

[0027] Based on the fragment instruction matrix, a distance matrix is ​​constructed corresponding to multiple execution fragments. Based on the distance matrix, multiple execution fragments are clustered to generate multiple fragment instruction type sets.

[0028] Specifically, for any two execution segments, the segment instruction matrix can be expanded into a two-dimensional vector, and a distance metric, such as calculating the Euclidean distance between the vectors, can be used to obtain the segment instruction distance between any two execution segments, thereby constructing a distance matrix that includes the segment instruction distance between any two execution segments.

[0029] Based on this, multiple execution segments are clustered according to the distance matrix to group execution segments with similar execution methods into the same instruction type set. The clustering process can employ a hierarchical clustering algorithm based on the distance matrix, which progressively merges the segment distances between execution segments to form multiple instruction type sets. In other embodiments, conventional clustering algorithms such as K-means clustering or spectral clustering can also be used, as long as they can classify multiple execution segments based on the distance matrix. This embodiment does not specifically limit the algorithm used for clustering. The above method allows for the initial classification of multiple execution segments at the execution method level.

[0030] Step S3: Construct a reference response template for each instruction type set based on multiple instruction response matrices, extract the response residual trajectory of each execution segment based on multiple reference response templates, and identify multiple abnormal response segments based on the response residual trajectory.

[0031] In this step, for each set of instruction fragment types, a corresponding reference response template is constructed using representative response data from that set. A reference response basis matching the execution method is established for each type of execution fragment. Then, the response of the actual execution fragment is compared with the reference response template to extract the corresponding response residual trajectory. The response residual trajectory is used to characterize the degree of additional deviation of the current execution fragment relative to normal fragments of the same type, thereby reflecting whether the fragment has deviated from the normal processing state.

[0032] Furthermore, the present invention does not judge anomalies based solely on a single-point deviation at a certain moment, but rather combines the changes in the response residual trajectory throughout the entire segment execution process to comprehensively judge multiple execution segments, thereby identifying multiple abnormal response segments. The purpose is to determine the location where the abnormal working condition finally manifests, that is, to find out which segments in the continuous processing chain have shown obvious abnormal responses.

[0033] As an optional implementation process, the construction of a reference response template for a set of fragment instruction types, and the process of identifying multiple abnormal response fragments, specifically include: To construct the reference response template for the fragment instruction type set, the centrality parameter of each execution fragment in the fragment instruction type set is calculated. Based on the centrality parameter, multiple candidate normal fragments are selected. The reference response template for the fragment instruction type set is generated by fusing the instruction response matrices corresponding to the multiple candidate normal fragments in the fragment instruction type set.

[0034] For multiple execution segments within the same instruction type set, the segment response distance between any two execution segments is first calculated based on the instruction response matrix corresponding to each execution segment. This calculation can be performed in the same way as the segment instruction distance calculation, and will not be elaborated here. Then, the centrality parameter of the execution segment is determined based on the segment response distances between each execution segment and the remaining execution segments in the same instruction type set. In this embodiment, the median of the multiple segment response distances of an execution segment is preferably used as the centrality parameter of that execution segment. This parameter reflects the degree of concentration of the instruction response matrix corresponding to the execution segment within its instruction type set. A smaller centrality parameter indicates a smaller difference in response between the execution segment and most other execution segments of the same type, and a closer approximation to the central response pattern of that type of execution segment.

[0035] Based on this, multiple execution segments within the same instruction type set are filtered according to the centrality parameter of each execution segment, and execution segments whose centrality parameter meets preset conditions are selected as candidate normal segments. For example, execution segments with a centrality parameter less than a preset deviation threshold are marked as candidate normal segments. After determining multiple candidate normal segments for each instruction type set, the instruction response matrices corresponding to the multiple candidate normal segments are fused to obtain a reference response template for that instruction type set. Preferably, the fusion process can use a median fusion method to reduce the impact of individual deviation segments on the reference response template, thereby enabling the generated reference response template to more stably reflect the typical response process of the instruction type set under normal processing conditions.

[0036] The instruction response matrix of each execution segment in the fragment instruction type set is subtracted element by element from the reference response template to obtain multiple response residual values. The multiple response residual values ​​corresponding to each execution segment are combined to construct the response residual trajectory of the execution segment.

[0037] Specifically, for any execution segment, based on the reference response template corresponding to its instruction type set, the instruction response matrix of the execution segment is subtracted element-by-element from the corresponding reference response template at the same response channel and the same segment process position to obtain the response residual value corresponding to each element in the instruction response matrix of the execution segment. Multiple response residual values ​​are then combined to construct the response residual trajectory of the execution segment, which reflects the deviation of the execution segment from the normal response template of the same type throughout the entire execution process.

[0038] For each response residual trajectory, calculate the average residual value. Based on the multiple average residual values, determine the average response threshold for detecting abnormal responses in multiple execution segments within each segment instruction type set. Execution segments with average residual values ​​greater than the average response threshold corresponding to their respective segment instruction type set are marked as abnormal response segments.

[0039] Specifically, for any execution segment, the average residual value corresponding to multiple response residual values ​​in the response residual trajectory is calculated as a measure of the overall deviation of the execution segment. Subsequently, within each segment instruction type set, the average residual value of multiple execution segments in the set is statistically analyzed, and the corresponding average response threshold is determined based on the distribution of the average residual values ​​of that type of execution segment. Preferably, the average response threshold can be determined based on the median and dispersion of the average residual values ​​of multiple execution segments within the segment instruction type set; for example, the sum of the median of the average residual values ​​and two or three times their standard deviation can be used as the average response threshold. When the average residual value of an execution segment is greater than the average response threshold corresponding to its segment instruction type set, it indicates that the execution segment has deviated significantly from the normal response template of the same type, and thus the execution segment is marked as an abnormal response segment.

[0040] Step S4: Perform a pre-order backtracking along the execution segment sequence with each abnormal response segment as the endpoint to identify the state writing segment, and determine the state trigger segment based on the abnormal response segment and the state writing segment.

[0041] In this step, the present invention does not simply regard the abnormal response segment as the starting point of the abnormality, but rather as the endpoint of the abnormal condition manifestation, and performs a sequential backtracking analysis along the execution segment sequence according to the chronological relationship. In this way, it is possible to identify which preceding execution segments have already had a sustained impact on the subsequent processing state before the abnormality manifests, and identify the segments with the initial impact as state-writing segments. State-writing segments do not necessarily exhibit obvious abnormalities immediately upon their own execution, but their execution results will change the response basis of subsequent segments, making certain subsequent segments more prone to abnormal manifestation.

[0042] After identifying the state write segment, the execution relationship between the exception response segment and the state write segment is further combined to determine the state trigger segment. The state trigger segment characterizes which specific type or execution location of the segment activates the potential condition and causes it to manifest as a detectable exception, given that a preceding abnormal condition has already been written and propagated along the subsequent processing steps. By distinguishing between state write segments and state trigger segments, the two different questions of how an exception is embedded in the processing and how it is ultimately activated and manifested can be separated, thus making subsequent control more targeted.

[0043] As an optional implementation process, the preceding backtracking along the execution segment sequence, with each exception response segment as the endpoint, to identify the state-written segment specifically includes: Multiple execution segments are classified according to multiple response residual trajectories to generate multiple segments response type sets. Each execution segment is encoded according to the segment instruction type set and the segment response type set to determine the instruction code and status code of each execution segment.

[0044] The specific method for classifying the responses of multiple execution segments after obtaining the response residual trajectories can refer to the aforementioned instruction classification process. The segment response distance between any two execution segments can be calculated based on the response residual trajectories, thereby constructing a distance matrix that includes the segment response distances between any two execution segments. Multiple execution segments are then clustered based on this distance matrix to group execution segments with similar response residual trajectory variations into the same segment response type set.

[0045] After clustering, each execution segment has a set of instruction types and a set of response types. Further, the set of instruction types to which the execution segment belongs is used as the instruction code for that execution segment, and the set of response types to which the execution segment belongs is used as the state code for that execution segment. In this way, execution segments can be discretized from both the execution method and response state dimensions.

[0046] Based on multiple candidate normal segments, identify multiple state transition pairs in the execution segment sequence and extract the state transition triplet for each state transition pair.

[0047] In this process, adjacent segments that are both candidate normal segments are selected from the execution segment sequence and identified as state transition pairs. For any state transition pair, a state transition triple is extracted based on the instruction code and state code of the execution segment. Specifically, this includes the state codes of the two preceding and following candidate normal segments and the instruction code of the subsequent candidate normal segment. For example, the state transition triple (A, b, C) represents the relationship that, under normal processing conditions, when the preceding execution segment is in state code A, the subsequent execution segment with instruction code b typically evolves to state code C. By extracting multiple state transition triples from multiple candidate normal segments, the evolution pattern of the response state along the execution segment sequence during normal processing can be obtained.

[0048] A normal state transition library is constructed based on multiple state transition triples. Prequence backtracking is performed on each abnormal response segment according to the normal state transition library and the execution segment sequence to identify the state write segment corresponding to each abnormal response segment.

[0049] This involves categorizing and statistically analyzing multiple state transition triples based on the state code of the preceding segment and the instruction code of the following segment to establish a normal state transition library. This library records at least one normal successor state code or set of normal successor state codes given the state code of the preceding segment and the instruction code of the following segment. Using this normal state transition library, during subsequent pre-execution backtracking, state replacement verification can be performed on each candidate pre-execution segment in the execution segment sequence, and it can be determined whether a particular pre-execution segment has a state writing effect on the formation of subsequent abnormal response segments, thereby identifying the state writing segments corresponding to each abnormal response segment.

[0050] As an optional implementation process, a pre-processing backtracking is performed on each abnormal response segment based on the normal state transition library and the execution segment sequence to identify the state write segment corresponding to each abnormal response segment. Specifically, this includes: Construct a normal state set for each abnormal response segment, including extracting the state codes corresponding to multiple candidate normal segments in the segment instruction type set according to the segment instruction type set to which the abnormal response segment belongs, and combining multiple state codes to obtain the normal state set of the abnormal response segment.

[0051] For any abnormal response fragment, firstly, its fragment instruction type set is determined. Then, based on multiple candidate normal fragments selected from this fragment instruction type set, the corresponding status codes for these candidate normal fragments are extracted. The extracted status codes are then deduplicated and combined to form the normal status set for the abnormal response fragment. It should be noted that since the fragment instruction type set and the fragment response type set are obtained independently, multiple candidate normal fragments in the same fragment instruction type set can correspond to multiple different status codes. Therefore, this invention does not use a single normal status code to represent a fragment instruction type set, but rather uses a normal status set to represent multiple normal response states that are allowed to occur under normal processing conditions for this type of execution fragment.

[0052] Multiple candidate preceding segments are identified for each abnormal response segment in the constructed execution segment sequence. State replacement detection is performed on the multiple candidate preceding segments in reverse order to identify the state write segments in the multiple candidate preceding segments.

[0053] Specifically, taking a certain abnormal response segment as the endpoint, multiple execution segments within a preset number or range are selected as candidate preceding segments in the execution segment sequence, and state replacement detection is performed sequentially in reverse order of distance from the abnormal response segment. By first detecting the candidate preceding segments close to the abnormal response segment, and then gradually tracing back to earlier execution segments, it is possible to progressively determine whether the abnormal condition has been written into a certain preceding segment and passed to the abnormal response segment along the subsequent execution process.

[0054] In this embodiment, the state replacement detection for any candidate preceding segment includes: determining the replacement state code of the current candidate preceding segment based on the instruction code of the current candidate preceding segment and the state code of the previous execution segment in the execution segment sequence, and replacing the state code of the current candidate preceding segment; then, using the replaced state code of the current candidate preceding segment as the preceding state, and combining it with the instruction code of the next execution segment in the execution segment sequence, determining the replacement state code of the next execution segment based on the normal state transition library, and replacing the state code of the next execution segment; then, performing the same process on the subsequent multiple execution segments after the candidate preceding segment, until the abnormal response segment is reached. In other words, the state replacement detection does not directly replace the abnormal response segment itself, but rather starts from a certain candidate preceding segment and, according to the normal evolution relationship described by the normal state transition library, recursively replaces the state codes of subsequent execution segments segment by segment to simulate the normal state evolution process that subsequent segments should exhibit if the candidate preceding segment has not been written into the abnormal condition.

[0055] In this embodiment, the method for determining the replacement state code of the current candidate preceding segment based on the normal state transition library is as follows: based on the instruction code of the current candidate preceding segment and the state code of the previous execution segment, query multiple state codes that the current candidate preceding segment may transition to under this combination, and select the one that appears most frequently as the replacement state code of the current candidate preceding segment.

[0056] After traversing to an exception response segment, if the state code after the exception response segment is replaced belongs to the normal state set of that exception response segment, then the candidate preceding segment is recorded as a candidate state write segment. In other words, if recursively replacing normal states starting from a certain candidate preceding segment eventually allows the exception response segment to return to the normal state range allowed by its instruction type set, then it indicates that the state anomaly before that candidate preceding segment has a decisive influence on the subsequent exception manifestation. Therefore, that candidate preceding segment can be regarded as a segment that has a writing effect on the formation of the exception response segment.

[0057] After performing state replacement detection on multiple candidate preceding segments, the candidate state writing segment that is furthest from the abnormal response segment is recorded as the state writing segment corresponding to the abnormal response segment.

[0058] The significance of the above state replacement detection is that if multiple candidate preceding segments all satisfy the condition that the abnormal response segment can return to the normal state set after replacement, it means that the abnormal condition has been written into the subsequent processing process at least from the earliest candidate preceding segment that satisfies the condition. Therefore, the candidate state writing segment that is furthest from the abnormal response segment is selected as the final state writing segment, so as to identify the earliest writing position of the abnormal condition in the execution segment sequence.

[0059] Based on this, the determination of the state triggering segment is specifically to write the candidate state closest to the exception response segment as the state triggering segment, which represents the last key triggering segment before the exception is finally manifested.

[0060] Step S5: Construct an abnormal working condition evolution chain based on the state writing fragment, state trigger fragment, and abnormal response fragment, and extract control call information from the abnormal working condition evolution chain for use in the generation of the engraving and milling machine control strategy.

[0061] In this step, based on the aforementioned analysis results, the state writing segment, state triggering segment, and abnormal response segment are linked to construct an abnormal operating condition evolution chain. This chain characterizes the entire process of an abnormal operating condition being introduced from a preceding segment, gradually propagating during subsequent processing, and ultimately being triggered and manifested in a specific segment. Unlike existing technologies that rely solely on passive control based on current vibration values ​​or individual abnormal segments, this invention makes the formation process of abnormal operating conditions explicit through the abnormal operating condition evolution chain. This allows the control system to not only identify where the abnormality is, but also where it begins to form and where it is most likely to be triggered, thus providing a more forward-looking basis for control intervention.

[0062] As an optional implementation process, extracting control call information from the abnormal operating condition evolution chain for use in the generation of the CNC engraving and milling machine control strategy specifically includes: Based on the instruction codes corresponding to the state writing segment and the state triggering segment in the abnormal operating condition evolution chain, the categories of the preceding writing instruction and the operating condition triggering instruction are determined, and the control call information corresponding to the abnormal operating condition evolution chain is obtained. The control call information is then bound to the preset control strategy corresponding to the abnormal operating condition evolution chain.

[0063] For any abnormal operating condition evolution chain, the instruction code corresponding to the state write segment in the chain is first extracted and identified as the preceding write instruction category. Simultaneously, the instruction code corresponding to the state trigger segment in the chain is extracted and identified as the operating condition trigger instruction category. Since the state write segment represents the initial execution type when an abnormal operating condition is introduced into the processing procedure, and the state trigger segment represents the key execution type when the abnormal operating condition is activated and tends to become explicit, the preceding write instruction category and the operating condition trigger instruction category together constitute the control call information for this abnormal operating condition evolution chain.

[0064] Furthermore, control call information corresponding to the same abnormal evolution chain of working conditions can be bound to a set of pre-configured control strategies. These pre-configured control strategies are used to execute corresponding controls on the engraving and milling machine when the same or similar working condition evolution conditions are detected to reappear during subsequent processing. The aforementioned pre-configured control strategies can be preset according to different application requirements, preferably including at least one of the following: feed parameter adjustment strategy, spindle speed adjustment strategy, trajectory smoothing parameter adjustment strategy, alarm prompt strategy, and processing pause strategy. By binding control call information to pre-configured control strategies, the aforementioned abnormal evolution chain analysis results of this invention can be further transformed into decision-making basis for actual control calls, enabling the results of the preceding working condition analysis to directly serve subsequent processing control.

[0065] During the controlled processing of the CNC engraving and milling machine, if the segment to be analyzed is detected to match any working condition trigger instruction category, and the segment to be analyzed has a preceding segment that matches the preceding instruction category written in the control call information to which the working condition trigger instruction category belongs, the preset control strategy corresponding to the control call information is invoked to control the CNC engraving and milling machine for processing.

[0066] Specifically, during the subsequent processing of the engraving and milling machine, newly generated execution segments can be continuously identified for segment instructions. It can be determined whether the currently analyzed execution segment belongs to a recorded condition trigger instruction category. If the determination is yes, a preceding sequence check is performed along the execution segment sequence preceding the currently analyzed execution segment to determine if a preceding execution segment matching the corresponding preceding instruction category has already appeared before the currently analyzed execution segment. When both conditions are met—that the currently analyzed execution segment belongs to a condition trigger instruction category and that a preceding execution segment matching the preceding instruction category exists—it indicates that the condition anomaly evolution chain corresponding to the control call information has again met the formation conditions, thereby triggering the invocation of the preset control strategy bound to the control call information. In this way, the present invention does not passively adjust after the anomaly manifests, but rather, once the preceding writing conditions and trigger conditions matching the existing condition anomaly evolution chain are identified during subsequent processing, the corresponding control strategy can be invoked in advance, thereby achieving proactive control against the risk of anomaly condition recurrence.

[0067] It is worth noting that the above-mentioned matching refers to the instruction code of the execution segment to be analyzed being consistent with the recorded working condition trigger instruction category. Preferably, the matching of the preceding execution segment refers to the existence of an execution segment with the same category as the preceding written instruction within a preset segment range before the execution segment to be analyzed. By limiting the joint occurrence relationship between the working condition trigger instruction category and the preceding written instruction category, this invention can further transform the working condition anomaly evolution chain from offline analysis results into online control call conditions. This allows the control system not only to know how existing abnormal working conditions are formed, but also to use the identified working condition anomaly evolution chain to intervene in advance against the recurrence of similar working conditions during subsequent processing, improving the timeliness and targeting of the CNC engraving machine control.

[0068] For example, when the invoked preset control strategy is a parameter adjustment strategy, the feed parameters, spindle parameters, or trajectory smoothing parameters can be switched; when the invoked preset control strategy is an early warning or protection strategy, abnormal early warning information can be output, or the engraving and milling machine can be controlled to perform deceleration, pause, or shutdown. It should be noted that this invention does not limit the generation of underlying control quantities directly from the abnormal working condition evolution chain. Instead, it preferably extracts control call information from the abnormal working condition evolution chain. That is, after corresponding abnormal working conditions occur in the historical control process, what strategies were used to handle them and restore normal operation can be collected and summarized by experts. Based on the handling methods for abnormal working conditions in the historical control process, the information can be collected, organized, and configured to construct multiple preset control strategies for use in actual operation. The control system can invoke existing control strategies such as feed adjustment, spindle adjustment, trajectory smoothing, early warning output, or machining pause when the subsequent execution segment meets the corresponding call conditions, thereby achieving early intervention for abnormal working conditions. Therefore, this invention realizes the transformation from adjusting after the anomaly is manifested to calling up pre-control based on the anomaly evolution process, thereby improving the timeliness and effectiveness of control in the engraving and milling machine processing.

[0069] Another aspect of this invention provides a CNC engraving machine control system based on multi-axis vibration correlation analysis; please refer to [link to relevant documentation]. Figure 2 The system includes: The data acquisition module is used to collect CNC execution data and machining response data during the CNC milling process, and extract multiple execution segments with sequential relationships from the CNC execution data to construct an execution segment sequence; The instruction classification module is used to extract the segment instruction matrix and instruction response matrix of each execution segment based on the CNC execution data and machining response data, and to classify multiple execution segments according to multiple segment instruction matrices to generate multiple segment instruction type sets; The anomaly identification module is used to construct a reference response template for each instruction type set based on multiple instruction response matrices, extract the response residual trajectory of each execution segment based on multiple reference response templates, and identify multiple abnormal response segments based on the response residual trajectory. The pre-order backtracking module is used to perform pre-order backtracking along the execution segment sequence with each exception response segment as the endpoint to identify the state write segment, and determine the state trigger segment based on the exception response segment and the state write segment. The control call module is used to construct an abnormal working condition evolution chain based on the state write fragment, state trigger fragment, and abnormal response fragment, and extract control call information from the abnormal working condition evolution chain for use in the generation of the engraving and milling machine control strategy.

[0070] The above are merely specific embodiments of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A control method for a CNC engraving and milling machine based on multi-axis vibration correlation analysis, characterized in that, include: Collect CNC execution data and machining response data during the CNC milling process, and extract multiple execution segments with sequential relationships from the CNC execution data to construct an execution segment sequence; Extract the segment instruction matrix and instruction response matrix for each execution segment based on the CNC execution data and machining response data, and classify the multiple execution segments according to the multiple segment instruction matrices to generate multiple segment instruction type sets; A reference response template for each instruction type set is constructed based on multiple instruction response matrices. The response residual trajectory of each execution segment is extracted based on the multiple reference response templates, and multiple abnormal response segments are identified based on the response residual trajectory. Each exception response segment is used as the endpoint to perform a pre-order backtracking along the execution segment sequence to identify the state write segment, and the state trigger segment is determined based on the exception response segment and the state write segment. An abnormal working condition evolution chain is constructed based on the state writing fragment, state trigger fragment, and abnormal response fragment. Control call information is extracted from the abnormal working condition evolution chain for use in the generation of the CNC engraving machine control strategy.

2. The control method for a milling machine based on multi-axis vibration correlation analysis according to claim 1, characterized in that, The response residual trajectory of each execution segment is extracted based on multiple reference response templates, and multiple abnormal response segments are identified based on the response residual trajectory, including: To construct the reference response template for the fragment instruction type set, the centrality parameter of each execution fragment in the fragment instruction type set is calculated. Based on the centrality parameter, multiple candidate normal fragments are selected. The reference response template for the fragment instruction type set is generated by fusing the instruction response matrices corresponding to multiple candidate normal fragments in the fragment instruction type set. The instruction response matrix of each execution segment in the fragment instruction type set is subtracted element by element from the reference response template to obtain multiple element response residual values. The multiple response residual values ​​corresponding to each execution segment are combined to construct the response residual trajectory of the execution segment. For each response residual trajectory, calculate the average residual value. Based on the multiple average residual values, determine the average response threshold for detecting abnormal responses in multiple execution segments within each segment instruction type set. Execution segments with average residual values ​​greater than the average response threshold corresponding to their respective segment instruction type set are marked as abnormal response segments.

3. The control method for a milling machine based on multi-axis vibration correlation analysis according to claim 2, characterized in that, Each exception response segment is used as the endpoint for pre-order backtracking along the execution segment sequence to identify the state-written segments, including: Multiple execution segments are classified according to multiple response residual trajectories to generate multiple segments response type sets. Each execution segment is encoded according to the segment instruction type set and the segment response type set to determine the instruction code and status code of each execution segment. Based on multiple candidate normal segments, identify multiple state transition pairs in the execution segment sequence, extract the state transition triplet of each state transition pair, including the state codes of the two preceding and following candidate normal segments and the instruction code of the following candidate normal segment; A normal state transition library is constructed based on multiple state transition triples. Prequence backtracking is performed on each abnormal response segment according to the normal state transition library and the execution segment sequence to identify the state write segment corresponding to each abnormal response segment.

4. The control method for a milling machine based on multi-axis vibration correlation analysis according to claim 3, characterized in that, Based on the normal state transition library and execution fragment sequence, a pre-order backtracking is performed on each abnormal response fragment to identify the state write fragment corresponding to each abnormal response fragment, including: Construct a normal state set for each abnormal response segment, including extracting the state codes corresponding to multiple candidate normal segments in the segment instruction type set according to the segment instruction type set to which the abnormal response segment belongs, and combining multiple state codes to obtain the normal state set of the abnormal response segment. Multiple candidate preceding segments are identified for each abnormal response segment in the constructed execution segment sequence. State replacement detection is performed on the multiple candidate preceding segments in reverse order to identify the state writing segments in the multiple candidate preceding segments. The state replacement detection for any candidate preceding segment includes: determining the replacement state code of the current candidate preceding segment based on the instruction code of the current candidate preceding segment and the state code of the previous execution segment in the execution segment sequence, based on the normal state transition library, and replacing the state code of the current candidate preceding segment. The above process is used to replace the state codes of multiple subsequent execution segments in the execution segment sequence. After traversing to the exception response segment, if the state code of the exception response segment after being replaced belongs to the normal state set of the exception response segment, then the candidate preceding segment is recorded as a candidate state and written into the segment. After performing state replacement detection on multiple candidate preceding segments, the candidate state writing segment that is furthest from the abnormal response segment is recorded as the state writing segment corresponding to the abnormal response segment.

5. The control method for a milling machine based on multi-axis vibration correlation analysis according to claim 4, characterized in that, Extracting control call information from the abnormal operating condition evolution chain for use in the generation of CNC engraving and milling machine control strategies includes: Based on the instruction codes corresponding to the state writing segment and the state triggering segment in the abnormal working condition evolution chain, the categories of the preceding writing instruction and the working condition triggering instruction are determined, and the control call information corresponding to the abnormal working condition evolution chain is obtained. The control call information is then bound to the preset control strategy corresponding to the abnormal working condition evolution chain. During the controlled processing of the CNC engraving and milling machine, if the segment to be analyzed is detected to match any working condition trigger instruction category, and the segment to be analyzed has a preceding segment that matches the preceding instruction category written in the control call information to which the working condition trigger instruction category belongs, the preset control strategy corresponding to the control call information is invoked to control the CNC engraving and milling machine for processing.

6. The control method for a milling machine based on multi-axis vibration correlation analysis according to claim 1, characterized in that, Based on the CNC execution data and machining response data, extract the segment instruction matrix and instruction response matrix for each execution segment. Then, classify the multiple execution segments according to the multiple segment instruction matrices to generate multiple sets of segment instruction types, including: Extract the three-axis command position increment sequence and spindle speed sequence for each execution segment from the CNC execution data, and construct the segment command matrix for each execution segment based on the three-axis command position increment sequence and spindle speed sequence; Based on the machining response data, extract the triaxial vibration acceleration sequence, triaxial servo current sequence, and spindle load sequence for each execution segment, and combine the triaxial vibration acceleration sequence, triaxial servo current sequence, and spindle load sequence to construct the instruction response matrix for each execution segment; Based on the fragment instruction matrix, a distance matrix is ​​constructed corresponding to multiple execution fragments. Based on the distance matrix, multiple execution fragments are clustered to generate multiple fragment instruction type sets.

7. A control system for a CNC engraving and milling machine based on multi-axis vibration correlation analysis, characterized in that, The system is used to implement the CNC engraving machine control method based on multi-axis vibration correlation analysis as described in any one of claims 1-6, including: The data acquisition module is used to collect CNC execution data and machining response data during the CNC milling process, and extract multiple execution segments with sequential relationships from the CNC execution data to construct an execution segment sequence; The instruction classification module is used to extract the segment instruction matrix and instruction response matrix of each execution segment based on the CNC execution data and machining response data, and to classify multiple execution segments according to multiple segment instruction matrices to generate multiple segment instruction type sets; The anomaly identification module is used to construct a reference response template for each instruction type set based on multiple instruction response matrices, extract the response residual trajectory of each execution segment based on multiple reference response templates, and identify multiple abnormal response segments based on the response residual trajectory. The pre-order backtracking module is used to perform pre-order backtracking along the execution segment sequence with each exception response segment as the endpoint to identify the state write segment, and determine the state trigger segment based on the exception response segment and the state write segment. The control call module is used to construct an abnormal working condition evolution chain based on the state write fragment, state trigger fragment, and abnormal response fragment, and extract control call information from the abnormal working condition evolution chain for use in the generation of the engraving and milling machine control strategy.

Images