A method for identifying a motion disturbance source of a tool changing mechanism of a machine tool

Abstract

The present application relates to a kind of machine tool tool changing mechanism movement disturbance source identification method, it is related to numerical control machine tool and automation equipment disturbance diagnosis field, the present application is arranged in the key position of tool changing mechanism current, vibration and laser displacement sensor, synchronous acquisition multi-source signal and with displacement signal as benchmark segment tool changing event, construct the state database containing single disturbance such as normal and cam wear, gear pitting, motor drive drift, multi-dimensional feature vector is extracted from vibration, current signal, for unknown sample, using ant colony algorithm to optimize the adaptive variational modal decomposition of parameter is combined with independent component analysis to vibration signal is carried out blind source separation, obtain independent candidate source signal, respectively to candidate source signal and original current signal extraction feature, with disturbance sample in database matching, according to matching distance comprehensive determination single or complex disturbance type.The present application realizes the adaptive separation and identification of tool changing mechanism movement disturbance source, solves the problem that disturbance source is difficult to distinguish under complex disturbance mode.

Description

Technical Field

[0001] This invention relates to the field of disturbance diagnosis of CNC machine tools and automated equipment, and in particular to a method for identifying motion disturbance sources in machine tool tool changing mechanisms. Background Technology

[0002] As a core functional unit of CNC machine tools, the tool changing mechanism's motion accuracy and operational reliability directly determine the success rate of tool changes and the surface quality of the workpiece. During actual operation, the tool changing mechanism is subjected to alternating loads and complex friction over extended periods. Its key internal transmission components, including cams, gears, and the drive system, are highly susceptible to performance degradation, manifesting as various disturbances such as cam profile wear, gear pitting, and motor drive parameter drift. Even more challenging is the fact that these disturbances often occur concurrently, forming complex composite disturbance signals that pose a serious threat to the reliable operation of the equipment.

[0003] In the prior art, patent CN110442085A discloses a comprehensive performance and reliability testing device for tool magazines and automatic tool changers. This device utilizes multiple detection elements, including laser displacement sensors, vibration acceleration sensors, temperature sensors, noise sensors, and current sensors, to achieve multi-dimensional performance monitoring of the tool magazine and automatic tool changer. While this solution integrates multiple sensors, it remains at the level of comprehensive evaluation of the overall equipment status and cannot solve the problem of difficulty in distinguishing between various disturbance sources coupled together under complex disturbances.

[0004] Patent CN112077667A discloses a method for detecting the operating status of an automatic tool changer. It uses a laser displacement sensor and a vibration acceleration sensor to collect the tool change positioning accuracy and vibration signals, extracts energy entropy features using integrated empirical mode decomposition, and employs support vector machines for classification and diagnosis, achieving intelligent assessment of the tool changer's status. However, this method essentially classifies the overall health status of the equipment or common single faults. When the tool changer simultaneously suffers from cam wear and gear pitting, the collected vibration signal is a strongly coupled result of the two effects. Existing technology cannot effectively separate which component originates from the cam and which from the gear from this mixed signal, resulting in diagnostic results that can only indicate abnormalities but cannot provide a definite fault severity level, thus limiting its practical engineering value.

[0005] In addition, some studies employ adaptive frequency band deconvolution-based methods for early fault detection in automatic tool changers, using envelope spectrum analysis to extract fault features after removing interference components through spectral separation and reconstruction. Other studies utilize multi-component sparse separation algorithms for fault diagnosis of automatic tool changers, separating impact and harmonic components from vibration response signals using morphological component analysis. While these methods achieve signal decomposition to some extent, their core remains focused on feature separation within a single channel, failing to address the technical challenge of independently identifying each disturbance source under complex disturbances in the tool changer mechanism. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method for identifying motion disturbance sources of a machine tool tool changing mechanism, which solves the problem that the existing technology is difficult to effectively separate and accurately identify the type and intensity of each disturbance source when faced with multiple motion disturbance sources of the tool changing mechanism, especially compound disturbances.

[0007] The technical solution adopted by this invention to solve its technical problem is: a method for identifying motion disturbance sources in a machine tool tool changing mechanism, comprising the following steps: Step 1: Install current sensors, acceleration sensors, and laser displacement sensors at key parts of the tool changing mechanism to synchronously collect multi-source signals during the tool changing process.

[0008] Step 2: Under normal and artificially created single disturbance conditions, execute a complete tool changing cycle, collect corresponding multi-source signals, and calibrate each tool changing event based on the displacement signal to build a benchmark database containing normal and single disturbance conditions.

[0009] Step 3: For each tool change event sample in the benchmark database, extract feature vectors including time domain, frequency domain and envelope spectrum from the vibration and current signals to form a disturbance sample library.

[0010] Step 4: For unknown samples acquired online, the vibration signal is processed using an adaptive variational mode decomposition method (ACO-VMD-ICA) with parameters optimized by ant colony optimization to obtain independent candidate source signals. VMD is used to construct a virtual multi-channel observation matrix from the one-dimensional observation signal to solve the underdetermined blind source separation problem of ICA when the number of source signals exceeds the number of sensors.

[0011] Step 5: Extract specific features from the candidate vibration source signal and the original current signal respectively, and match them with feature samples in the disturbance sample library. Based on the minimum comparison logic of multiple standardized Euclidean distances, comprehensively determine the type and intensity of disturbance, such as no disturbance, single disturbance, or complex disturbance. When a candidate source is determined to be cam wear, further compare its standardized Euclidean distance with cam wear samples of different intensities, and take the intensity level with the smallest distance as the final output disturbance intensity of the source.

[0012] The ACO-VMD-ICA method proposed in step four of the above scheme is designed to solve the problem of strong coupling of vibration signals under complex disturbances. Its separation effect is not a simple superposition of VMD and ICA, but rather, through adaptive optimization, it can decompose the mixed signal into candidate sources with clear physical meanings. This provides a more reliable basis for feature matching in step five. Step five uses the minimum value comparison of multiple standardized Euclidean distances to automatically determine whether the disturbance is undisturbed, a single disturbance, or a complex disturbance, and provides the corresponding disturbance type. Overall, the steps work together to form a complete technical solution, the effect of which is difficult to achieve using existing technologies alone. Therefore, it has outstanding substantive features and significant progress.

[0013] The beneficial effect of this invention is that it solves the defects existing in the prior art. 1. This invention diagnoses the operating status of the tool changing mechanism by simultaneously acquiring three signals: current, vibration, and displacement. This overcomes the shortcomings of single signals being susceptible to fluctuations in operating conditions and interference from environmental noise, and significantly improves the accuracy and reliability of disturbance identification under complex operating conditions.

[0014] 2. To address the problem of multi-source coupling in vibration signals, this invention employs an adaptive variational mode decomposition (VMD) method with ant colony optimization of parameters, combined with independent component analysis (ICA). The decomposition parameters are automatically determined using the average envelope entropy as the objective function. VMD upscales the one-dimensional vibration signal into a virtual multi-channel observation matrix, and ICA then separates the independent source components. This method can decompose mixed signals into physically meaningful candidate sources even under underdetermined conditions where the number of sensors is less than the number of source signals. Combined with single disturbance samples, it overcomes the limitation of traditional methods in distinguishing coupled disturbance sources.

[0015] 3. This invention pre-constructs a benchmark database containing single disturbance samples, extracts features from each candidate source and performs standardized Euclidean distance matching with different types of disturbance samples, and uses multiple sets of minimum value comparison logic to automatically determine the specific type of each source as normal, single disturbance or compound disturbance, thus realizing independent identification of multiple disturbance sources such as cam wear and gear pitting.

[0016] 4. The sample library contains pre-stored disturbance intensity labels. During online diagnosis, the system can not only identify the disturbance type through feature matching, but also directly output the intensity level of the corresponding disturbance based on the matching distance. Attached Figure Description

[0017] Figure 1 This is an overall flowchart of a method for identifying motion disturbance sources in a machine tool tool changing mechanism according to the present invention.

[0018] Figure 2(a) is an example of the time domain diagram of the original mixed vibration signal, Figure 2(b) is an example of the time domain diagram of the original cam wear fault vibration signal, and Figure 2(c) is an example of the time domain diagram of the original gear pitting fault mixed vibration signal.

[0019] Figure 3 This is an example of a time-domain plot of two candidate source signals separated by the ACO-VMD-ICA method of this invention.

[0020] Figure 4(a) shows an example of the power spectrum of the original mixed vibration signal, Figure 4(b) shows an example of the envelope spectrum of candidate source 1, and Figure 4(c) shows an example of the envelope spectrum of candidate 2. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to the accompanying drawings and preferred embodiments. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0022] like Figure 1 The method for identifying motion disturbance sources in a machine tool tool changing mechanism, as shown, includes the following steps: Step 1: Install current sensors, acceleration sensors, and laser displacement sensors at key parts of the tool changing mechanism to synchronously collect multi-source signals during the tool changing process.

[0023] The current sensor is installed on the U-phase power line of the tool changer motor to collect the drive current signal. The triaxial accelerometer is attached to the bearing housing near the cam follower to collect X, Y, and Z-axis vibration signals. , , The laser displacement sensor is fixed to the machine tool bed by a bracket, with the optical path parallel to the direction of movement of the driven slider and aligned with its side to collect displacement signals. This displacement signal completely records the entire process of cam motion.

[0024] Step 2: Under normal and artificially created single disturbance conditions, execute a complete tool changing cycle, collect corresponding multi-source signals, and calibrate each tool changing event based on the displacement signal to build a benchmark database containing normal and single disturbance conditions.

[0025] In this step, different degrees of cam wear, gear pitting, and motor drive drift are set as single disturbance sources. For each disturbance, the tool change action is executed multiple times and multi-source signals are collected simultaneously to form disturbance sample data; at the same time, the start and end points of each tool change action are calibrated based on the displacement signal S(t) to segment independent tool change event samples.

[0026] Step 3: For each tool change event sample in the benchmark database, extract feature vectors including time domain, frequency domain and envelope spectrum from the vibration and current signals to form a disturbance sample library.

[0027] The extracted feature vector F includes: Vibration signal feature extraction: Synthetic vibration amplitude And calculate its time-domain characteristics, including the root mean square value. , cliff The root mean square (RMS) Characterizing the magnitude of vibrational energy, kurtosis It reflects the strength of the signal impact component and is a dimensionless index, among which This represents the average value of the vibration amplitude.

[0028] Envelope spectral characteristics: The envelope signal is obtained by performing a Hilbert transform on A(t) and then performing spectral analysis to extract frequencies such as the rotation frequencies of a cam and a gear. and its multipliers The amplitude at the point, and the gear meshing frequency and its sideband energy near the gear meshing frequency The total energy within the frequency band.

[0029] Current signal feature extraction: Extracting the amplitude of the current signal. Calculate the maximum current value and the maximum slope of the current rise during the tool change phase .

[0030] Step 4: For unknown samples acquired online, the vibration signal is processed using an adaptive variational mode decomposition method (ACO-VMD-ICA) with parameters optimized by ant colony optimization to obtain independent candidate source signals. VMD is used to construct a virtual multi-channel observation matrix from the one-dimensional observation signal to solve the underdetermined blind source separation problem of ICA when the number of source signals exceeds the number of sensors.

[0031] The ACO-VMD-ICA method is used to process the vibration signal A(t) of the unknown sample, specifically including: Vibration signal preprocessing: A(t) is detrended, and the linear trend term is fitted using the least squares method. Subtract and get Then remove the DC component. That is, to vibration amplitude at each sampling point Calculate the average value and subtract it; finally, use a bandpass filter to preserve the target frequency band. The components within, output preprocessed signal .

[0032] Adaptive optimization of ACO-VMD decomposition parameters: Let the average envelope entropy be the input, and let it be the objective function. ,in It is the number of modes in the VMD decomposition, i.e., the number of intrinsic mode functions. It is a quadratic penalty factor in VMD, controlling the bandwidth of each mode. For the first The envelope entropy of each intrinsic mode function (EMF) is used. The average envelope entropy is chosen as the optimization objective because it is highly sensitive to periodic impact signals caused by wear and pitting faults, guiding the algorithm to decompose fault-related components. Compared to information entropy or permutation entropy commonly used in traditional VMD optimization, the average envelope entropy guides the decomposition direction more stably, especially when the composite disturbance contains multiple impact sources with different periods. It can more evenly decompose the modes corresponding to each source, rather than excessively fragmenting a single impact signal. The calculation process is as follows: [The text abruptly ends here, likely due to an incomplete translation or source material.] Performing the Hilbert transform yields the analytic signal, and taking the modulus gives the envelope sequence. , Normalize the envelope into a probability distribution. Then the envelope entropy .

[0033] Using ant colony algorithm , Internal search enables Minimum optimal parameters The parameters for the ant colony algorithm are set as follows: number of ants. Each ant independently selects a set in each iteration. Parameters; pheromone volatile factor Controlling the rate of decay of historical pheromone concentration: The larger the value, the faster the historical pheromone evaporates, and the more the algorithm tends to explore new paths; heuristic factor Control the relative importance of heuristic information in path selection; maximum number of iterations. Each ant selects a group based on its current pheromone concentration and heuristic information. It then calls VMD decomposition to calculate the corresponding average envelope entropy as a path quality evaluation. The output is the globally optimal parameter that minimizes the average envelope entropy. Finally, a virtual multi-channel observation matrix is ​​obtained. This step is crucial for solving the underdetermined blind source separation problem in ICA. It involves using VMD to upscale the one-dimensional vibration signal collected by a single sensor to a multi-dimensional signal, thus satisfying the ICA requirement that the number of observed signals be no less than the number of source signals, thereby separating multiple independent sources.

[0034] ICA blind source separation: for The process involves centering the values ​​of each row by subtracting the mean, then whitening to remove correlations between components, and finally calculating the covariance matrix. , The observed matrix is ​​centered, and N is the number of sampling points for each channel. Eigenvalue decomposition... , Covariance matrix The eigenvector matrix obtained after eigenvalue decomposition Covariance matrix The eigenvalue diagonal matrix and whitening matrix obtained after eigenvalue decomposition Obtaining whitening signal The FastICA algorithm was then used to estimate the unmixing matrix. The independent source component matrix is ​​obtained. Calculate the permutation entropy and kurtosis of each independent source component, and set the permutation entropy threshold based on historical experience. Components with permutation entropy exceeding a set threshold are classified as random noise and eliminated. The remaining components are used as candidate disturbance source signals, denoted as... .

[0035] Step 5: Extract specific features from the candidate vibration source signal and the original current signal respectively, and match them with feature samples in the disturbance sample library. Based on the minimum comparison logic of multiple standardized Euclidean distances, comprehensively determine the type and intensity of disturbance, such as no disturbance, single disturbance, or complex disturbance. When a candidate source is determined to be cam wear, further compare its standardized Euclidean distance with cam wear samples of different intensities, and take the intensity level with the smallest distance as the final output disturbance intensity of the source.

[0036] The specific operation of source signal matching and disturbance comprehensive judgment is as follows: From the disturbance sample library constructed in step three, samples of cam wear, motor drive drift, and gear pitting are selected, and their corresponding disturbance intensity labels are recorded. For each type of disturbance, specific feature vectors are extracted as matching samples, including cam wear features. ; The amplitude at the cam rotation frequency. The amplitude at the second harmonic. The amplitude at the third harmonic. Cumberliness; gear pitting characteristics , The amplitude at the gear's rotational frequency. The amplitude at the second harmonic. The amplitude at the third harmonic. Sideband energy; motor-driven drift characteristics , This is the maximum value of the motor current. The maximum rising slope of the motor current; and these features are also extracted from normal samples, and the mean is calculated for each feature vector. and standard deviation , used for standardized Euclidean distance calculation; For each candidate source signal Extract corresponding features to construct vectors ,calculate Standardized Euclidean distance from various perturbed samples and normal samples ,in For each feature vector in a certain class of samples, Let be the mean vector of all feature vectors in a certain class of samples. This is the standard deviation vector of each feature vector in a certain class of samples; including the distance to the cam wear sample. , to the normal sample distance corresponding to the cam feature Distance to gear pitting sample , to the normal sample distance corresponding to the gear feature Distance to motor drift sample Distance to normal samples corresponding to motor features .

[0037] Take the minimum value among all the above distances , , If the minimum value corresponds to a normal state, the source is determined to be without disturbance; if it corresponds to cam wear, gear pitting, or motor drive drift, it is determined to be the corresponding disturbance type, and the corresponding disturbance intensity is recorded. If multiple source signals or current signals are simultaneously determined to be abnormal states, they are collectively determined to be composite disturbances, and each determined disturbance type and its intensity are output.

[0038] The following is a detailed explanation using a specific simulation case.

[0039] Step 1: Sensor Setup and Signal Acquisition

[0040] Install the following sensors at key locations in the tool changing mechanism: Install a current sensor on the U-phase power line of the tool changing motor to collect the motor drive current signal. .

[0041] The triaxial accelerometer is magnetically attached to the bearing housing near the cam follower and is used to collect vibration signals in the X, Y, and Z directions. .

[0042] The laser displacement sensor is fixed to the machine tool bed by an independent sensor bracket. The optical path is parallel to the movement direction of the driven slider and aligned with the reflective sticker on the side of the slider, for non-contact acquisition of displacement signals from the actuator. .

[0043] Step 2: Construct a baseline database of normal and disturbed states.

[0044] When the tool changing mechanism is in normal condition, it performs a complete tool changing cycle, synchronously collects all sensor signals, repeats this process multiple times, obtains a set of healthy sample data, and extracts feature vectors according to the method in step three, storing them in the database as normal state samples.

[0045] For various disturbance states, signal model simulation is used to generate single disturbance signal samples with clear fault characteristics: vibration and displacement signals containing 0.1mm and 0.2mm cam profile wear characteristics, vibration signals containing 0.2mm and 0.5mm gear tooth surface pitting characteristics, and motor drive drift current signals generated by setting current loop gain offsets of +5% and -5% are generated respectively through dynamic simulation models. Each disturbance and its simulation signals of different intensities are simulated multiple tool changing operations to form disturbance sample data.

[0046] Each of the above disturbance samples carries a quantified disturbance intensity label when it is entered into the database: cam wear is wear width: 0.1mm, wear width: 0.2mm; gear pitting is pitting diameter: 0.2mm, pitting diameter: 0.5mm; motor drive drift is current loop gain offset: +5%, current loop gain offset: -5%.

[0047] Using the displacement signal S(t) as a reference, the start and end points of each tool change cycle are calibrated, and the continuous signal stream is divided into independent tool change event samples.

[0048] Step 3: Multidimensional Feature Extraction

[0049] For each tool change event sample, the time-domain features, envelope spectrum features, and current signal features of the vibration signal are extracted to construct a feature vector library.

[0050] Step 4: Separate and identify signals from unknown samples.

[0051] To verify the core capabilities of this invention, an unknown composite disturbance sample was constructed using simulation: the wear intensity label of the cam corresponds to a wear width of 0.2 mm, and the characteristic frequency... For gear pitting, the strength label corresponds to a pitting diameter of 0.5mm and a meshing frequency. Frequency conversion The vibration signals are linearly superimposed and synthesized, as shown in Figure 2. In the mixed vibration signal of this sample, the two fault characteristics are superimposed and modulated.

[0052] Following the aforementioned method, the vibration signal underwent detrending, DC removal, and bandpass filtering preprocessing. Then, the ACO-VMD-ICA method was used to separate candidate source 1 and candidate source 2. Figure 3 As shown.

[0053] Step 5: Source signal matching and disturbance comprehensive judgment

[0054] Candidate Source 1: As shown in Figure 4(b), its envelope spectrum exhibits significant peaks at 12 Hz, 24 Hz, and 36 Hz, with a kurtosis of 5.08 and a sideband energy of 0.039. Standardized Euclidean distance calculation results: the distance to the cam wear sample is 3.42, the distance to the gear pitting sample is 7.65, and the distance to the normal sample is 5.87. The minimum distance corresponds to cam wear, thus classifying it as cam wear type. Further comparison with distances to cam wear samples of different intensities shows that the closest distance is 3.42 to the sample with a wear width of 0.2 mm, and 7.89 to the sample with a wear width of 0.1 mm. Therefore, the intensity level is determined to be a 0.2 mm wear width.

[0055] Candidate Source 2: As shown in Figure 4(c), its envelope spectrum shows a main peak at 320 Hz, with obvious sidebands at 300 Hz and 340 Hz, a kurtosis of 3.41, and a sideband energy of 0.086. Normalized Euclidean distance: The distance to the gear pitting sample is 2.56, the distance to the cam wear sample is 6.73, and the distance to the normal sample is 4.92. It is determined to be of the gear pitting type. Further comparison with pitting samples of different intensities shows that the closest distance is 2.56 to the sample with a pitting diameter of 0.5 mm, and the closest distance is 5.18 to the sample with a pitting diameter of 0.2 mm. Therefore, the intensity level is determined to be 0.5 mm pitting diameter.

[0056] Overall judgment: Since both candidate sources are judged to be in an abnormal state, the system outputs a composite disturbance: the intensity level of cam wear: 0.2mm wear width + the intensity level of gear pitting: 0.5mm pitting diameter.

[0057] Although the composite disturbance demonstrated in this embodiment consists of two types of mechanical disturbances, it should be understood that the method of this invention can also determine the presence of motor-driven drift in the tool changing mechanism under test. Specifically: among the independent candidate sources separated in step four, if a candidate source has the smallest standardized Euclidean distance between its feature vector and the feature vector in the motor-driven drift sample library after feature extraction, then the source is determined to be of the motor-driven drift type. Simultaneously, in the disturbance sample library constructed in step three, the motor-driven drift samples have intensity labels such as current loop gain offset +5% or -5%. In step five, by comparing the distance between the source and samples with different intensity labels, the intensity level that best matches the source can be output. For example, if the current feature of a candidate source is closest to the current loop gain offset +5% sample, then it is determined to be a motor-driven drift intensity: +5% offset. This determination process follows the same logic as the determination of cam wear and gear pitting described above.

[0058] Comparative Verification: As a contrast, the traditional method, which directly extracts the overall energy entropy of the mixed signal and uses support vector machine classification, can only output that the tool changing mechanism is in an abnormal state, but cannot distinguish the specific faulty component and its intensity. In contrast, this invention successfully decouples and individually identifies the complex disturbance, and provides an intensity level consistent with the simulation settings, demonstrating its significant technological advancement.

[0059] The above description is only a specific embodiment of the present invention. Various examples and illustrations do not constitute a limitation on the substantive content of the present invention. Those skilled in the art can make modifications or variations to the above-described specific embodiments after reading the specification without departing from the substance and scope of the invention.

Claims

1. A method for identifying motion disturbance sources in a machine tool tool changing mechanism, characterized in that: Includes the following steps, Step 1: Install current sensors, acceleration sensors, and laser displacement sensors at key parts of the tool changing mechanism to synchronously collect multi-source signals during the tool changing process; Step 2: Under the normal state and under various artificially created single disturbance states of the tool changing mechanism, execute the complete tool changing action cycle, collect the corresponding multi-source signals, and use the displacement signal as a reference to calibrate each tool changing event, and build a reference database that includes the normal state and various single disturbance states. Step 3: For each tool change event sample in the benchmark database, extract feature vectors containing time domain, frequency domain, and envelope spectrum from the vibration and current signals to form a disturbance sample library; Step 4: For unknown samples acquired online, the vibration signal is processed by adaptive variational mode decomposition with ant colony algorithm to optimize parameters, combined with independent component analysis, to obtain independent candidate source signals; wherein, variational mode decomposition is used to construct a virtual multi-channel observation matrix from the one-dimensional observation signal. Step 5: Extract specific features from the candidate vibration source signal and the original current signal respectively, and match them with the feature samples in the disturbance sample library. Based on the minimum value comparison logic of multiple standardized Euclidean distances, comprehensively determine the type and intensity of no disturbance, single disturbance, or compound disturbance. When a candidate source is determined to be cam wear, further compare its standardized Euclidean distance with cam wear samples of different intensities, and take the intensity level with the smallest distance as the final output disturbance intensity of the source.

2. The method for identifying motion disturbance sources in a machine tool tool changing mechanism as described in claim 1, characterized in that: In step one, a current sensor is installed on the U-phase power line of the tool changer motor to collect the drive current signal I(t); a three-axis accelerometer is attached to the bearing housing near the cam follower to collect the X, Y, and Z-axis vibration signals Ax(t), Ay(t), and Az(t); a laser displacement sensor is fixed to the machine tool bed by a bracket, with the optical path parallel to the movement direction of the follower slider and aligned with its side to collect the displacement signal S(t). This displacement signal completely records the entire process of the cam movement.

3. The method for identifying motion disturbance sources in a machine tool tool changing mechanism as described in claim 1, characterized in that: In step two, different degrees of cam wear, gear pitting, and motor drive drift are set as single disturbance sources. For each disturbance, the tool change action is performed multiple times and multi-source signals are collected simultaneously to form disturbance sample data. At the same time, the starting point and ending point of each tool change action are calibrated based on the displacement signal S(t) to segment independent tool change event samples.

4. The method for identifying motion disturbance sources in a machine tool tool changing mechanism as described in claim 1, characterized in that: In step three, the extracted feature vector F includes: Vibration signal feature extraction: Synthetic vibration amplitude And calculate its time-domain characteristics, including the root mean square value. , cliff , where the root mean square Characterizing the magnitude of vibrational energy, kurtosis It reflects the strength of the signal impact component and is a dimensionless index, among which This represents the average amplitude of the vibration. Envelope spectral characteristics: The envelope signal is obtained by performing a Hilbert transform on A(t) and then subjected to spectral analysis to extract the rotation frequencies of the cam and gear, respectively. and its multipliers The amplitude at the point, and the gear meshing frequency and its sideband energy near the gear meshing frequency The total energy within the frequency band; Current signal feature extraction: Extract the amplitude I(t) of the current signal and calculate the maximum current value. and the maximum slope of the current rise during the tool changing phase .

5. The method for identifying motion disturbance sources in a machine tool tool changing mechanism as described in claim 1, characterized in that: Step four involves using an ant colony optimization algorithm to optimize parameters, followed by adaptive variational mode decomposition combined with independent component analysis to process the vibration signal A(t) of the unknown sample. Specifically, this includes: Vibration signal preprocessing: A(t) is detrended, and the linear trend term is fitted using the least squares method. Subtract and get Then remove the DC component. That is, to vibration amplitude at each sampling point Calculate the average value and subtract it; finally, use a bandpass filter to preserve the target frequency band. The components within, output preprocessed signal ; Adaptive optimization of variational mode decomposition parameters: Let the average envelope entropy be the input, and let it be the objective function. ,in It is the number of modes in the VMD decomposition, i.e., the number of intrinsic mode functions. It is a quadratic penalty factor in VMD, controlling the bandwidth of each mode. For the first The envelope entropy of each intrinsic mode function is calculated as follows: Performing the Hilbert transform yields the analytic signal, and taking the modulus gives the envelope sequence. , Normalize the envelope into a probability distribution. Then the envelope entropy ; Using ant colony algorithm , Internal search enables Minimum optimal parameters The output is the globally optimal parameter that minimizes the average envelope entropy, ultimately yielding the virtual multi-channel observation matrix. ; Independent component analysis blind source separation: for The process involves centering the values ​​of each row by subtracting the mean, then whitening to remove correlations between components, and finally calculating the covariance matrix. , The observed matrix is ​​centered, and N is the number of sampling points for each channel. Eigenvalue decomposition... , Covariance matrix The eigenvector matrix obtained after eigenvalue decomposition Covariance matrix The eigenvalue diagonal matrix and whitening matrix obtained after eigenvalue decomposition It transforms the observed signal into a whitened signal in which the variance of each component is 1 and they are uncorrelated. The FastICA algorithm is used to estimate the unmixing matrix W, and the independent source component matrix S=WZ is obtained. The permutation entropy of each independent source component is calculated. Components with permutation entropy higher than a set threshold are judged as random noise and eliminated. The remaining components are used as candidate disturbance source signals.

6. The method for identifying motion disturbance sources in a machine tool tool changing mechanism as described in claim 5, characterized in that: The parameters of the ant colony algorithm are set as follows: number of ants M=20, each ant independently selects a set of (K,α) parameters in each iteration; pheromone evaporation factor ρ=0.5, which controls the decay rate of historical pheromone concentration: the larger ρ is, the faster the historical pheromone evaporates, and the more the algorithm tends to explore new paths; heuristic factor β=2, which controls the relative importance of heuristic information in path selection; maximum number of iterations Tmax=30; each ant selects a set of (K,α) based on the current pheromone concentration and heuristic information, and calls variational mode decomposition to calculate the corresponding average envelope entropy as a path quality evaluation.

7. The method for identifying motion disturbance sources in a machine tool tool changing mechanism as described in claim 5, characterized in that: The permutation entropy threshold is set to Components with permutation entropy higher than 0.8 are classified as random noise and removed.

8. The method for identifying motion disturbance sources in a machine tool tool changing mechanism as described in claim 1, characterized in that: The specific operation of source signal matching and disturbance comprehensive determination in step five is as follows: From the disturbance sample library constructed in step three, samples of cam wear, motor drive drift, and gear pitting are selected, and the corresponding disturbance intensity labels are recorded. For each type of disturbance, specific feature vectors are extracted as matching samples, including cam wear features. , The amplitude at the cam rotation frequency. The amplitude at the second harmonic. The amplitude at the third harmonic. Cumberliness; gear pitting characteristics , The amplitude at the gear's rotational frequency. The amplitude at the second harmonic. The amplitude at the third harmonic. Sideband energy; motor-driven drift characteristics , This is the maximum value of the motor current. This represents the maximum rising slope of the motor current. These features are also extracted from normal samples, and the mean is calculated for each feature vector. and standard deviation , used for standardized Euclidean distance calculation; For each candidate source signal Extract corresponding features to construct vectors ,calculate Standardized Euclidean distance from various perturbed samples and normal samples ,in For each feature vector in a certain class of samples, Let be the mean vector of all feature vectors in a certain class of samples. This is the standard deviation vector of each feature vector in a certain class of samples; including the distance to the cam wear sample. , to the normal sample distance corresponding to the cam feature Distance to gear pitting sample , to the normal sample distance corresponding to the gear feature Distance to motor drift sample Distance to normal samples corresponding to motor features ; Take the minimum value among the above distances, that is , , If the minimum value corresponds to the normal state, the source is determined to be without disturbance; if it corresponds to cam wear, gear pitting, or motor drive drift, it is determined to be the corresponding disturbance type, and the corresponding disturbance intensity is recorded; if multiple source signals or current signals are determined to be abnormal at the same time, they are comprehensively determined to be a composite disturbance, and each determined disturbance type and its intensity are output.

9. The method for identifying motion disturbance sources in a machine tool tool changing mechanism as described in claim 1, characterized in that: The composite disturbances include the combination of cam wear and gear pitting, the combination of cam wear and motor drive drift, the combination of gear pitting and motor drive drift, and the combination of the three disturbances.

10. The method for identifying motion disturbance sources in a machine tool tool changing mechanism as described in claim 1, characterized in that: The disturbance intensity includes: the wear width of cam wear, the pitting diameter of gear pitting, and the current loop gain offset of motor drive drift.

Images