CN116079498A - Method for identifying abnormal signals of cutter - Google Patents

Abstract

The invention provides a method for identifying abnormal signals of a cutter, which comprises the following steps: firstly, selecting a primary section to be identified and a primary comparison section from an aligned cutter initial signal and a standard template signal, and then selecting an alignment point with the maximum Manhattan distance between the primary section and the primary comparison section as a starting point of the identification section, thereby being beneficial to the improvement of redundancy and the signal to noise ratio; and judging whether the difference identification section is abnormal according to the Manhattan operation result of the identification section and the comparison section and the three sigma principle, thereby ensuring the robustness of the identification result.

Description

Method for identifying abnormal signals of cutter

Technical Field

The invention relates to the field of industrial control, in particular to a method for identifying abnormal signals of a cutter.

Background

With the development of sensors, big data and edge computing devices, predictive maintenance (PHM) and real-time monitoring techniques for machine tools and tools are possible.

The current common technical architecture includes the following parts: data acquisition, data preprocessing, feature extraction, model construction and model iterative correction. The above-mentioned technical architecture generally supports various data sources and data types, such as spindle current load signals, vibration signals, sound signals, and image signals.

The data acquisition step acquires the data from a certain monitored object (taking a certain specific machine tool as an example) through a network card/serial port and other ways, and acquires other auxiliary information such as time, equipment running state and the like. After all the acquired data fields are fused, through a data preprocessing step, noise data is cleaned, and finally, the data are sent to a subsequent step for feature extraction and model construction.

Most of the feature extraction and model construction methods currently in common use are based on the morphology of time series signals and can be roughly classified into two categories:

(1) Classical signal processing scheme. For example, the EMD method finds the upper envelope curve and the lower envelope curve of the time sequence signal through three interpolation fitting, then iteratively searches the components of the original signal, splits the original signal into the accumulation of the corresponding components, stops iteration after reaching the convergence boundary, and then performs the corresponding feature extraction on the obtained components as the subsequent analysis index. The method needs to perform third-order polynomial interpolation, has low efficiency, and has huge calculation resource consumption and great difficulty in application to actual scenes under the condition of slightly long signal length or lower signal-to-noise ratio.

(2) Machine learning algorithms. Feature extraction and model construction of timing signals can be seen as classification problems in machine learning. Existing methods include Support Vector Machines (SVMs), convolutional Neural Networks (CNNs), etc., which mostly require perfectly aligned, equal length time series data as input. However, in the actual production environment, in consideration of factors such as network communication delay, calculation power of the acquisition module and the machine tool chip, it is impossible to acquire the time sequence signals at uniform intervals. Moreover, training of such models requires a large amount of data to ensure reliability.

Disclosure of Invention

In order to solve the problems, the invention provides a method for identifying abnormal signals of a cutter.

The main content of the invention comprises:

a method for identifying abnormal signals of a cutter comprises the following steps:

acquiring a cutter initial signal transmitted by a machine tool;

after aligning a cutter initial signal with a standard template signal, selecting a primary section to be identified in the cutter initial signal and a primary comparison section corresponding to the primary section to be identified in the standard template signal;

comparing the primary section to be identified with the primary comparison section to obtain an identification section and a comparison section;

and judging whether the identification section is abnormal or not according to the Manhattan operation result of the identification section and the comparison section and the three sigma principle.

Preferably, the manhattan operation result according to the primary segment to be identified and the comparison segment includes:

calculating corresponding Manhattan distances for each pair of the position points of the identification section and the comparison section, and summing to obtain a mean value to obtain a Manhattan distance mean value;

comparing the Manhattan mean value with a preset threshold value, and judging whether the difference identification section is abnormal or not according to a three-sigma principle.

Preferably, comparing the primary segment to be identified with the primary comparison segment to obtain an identification segment and a comparison segment, including:

calculating the Manhattan distance corresponding to each pair of position points of the primary section to be identified and the primary comparison section;

and selecting the biggest alignment point of the Manhattan record as the starting point of the identification segment, and selecting the end point of the primary segment to be identified as the end point of the identification segment.

Preferably, after aligning the tool initial signal with the standard template signal, selecting a primary segment to be identified in the tool initial signal, and a primary comparison segment corresponding to the primary segment to be identified in the standard template signal, including:

windowing and sliding are carried out on the cutter initial signal and the standard template signal, and the distance between sliding windows is calculated;

and taking the alignment point with the largest sliding window distance as the starting point of the primary section to be identified, and determining the end point of the primary section to be identified according to the size of the sliding window.

Preferably, the standard template signal is generated as follows:

creating an input data set, wherein the input data set is processing data of normal operation of a product in a set time, the product consists of N workpieces, and the processing data comprises processing signals of the N workpieces;

calculating the DTW distances among N workpieces to obtain a DTW distance matrix of the product;

calculating the sum of the DTW distances of each workpiece and the rest of the workpieces respectively, and marking the sum as a row DTW distance sum;

and taking a signal after the dynamic time warping of the workpiece corresponding to the minimum value in the N line DTW distance sums as a standard template signal.

Preferably, the tool initiation signal is aligned with the standard template signal by DTW.

The identification method of the cutter abnormal signal provided by the invention has the beneficial effects that: the aligned primary segment to be identified and the corresponding primary comparison segment of the standard template signal are aligned, and the alignment point with the maximum Manhattan distance between the primary segment to be identified and the corresponding primary comparison segment is selected as the starting point of the identification segment, so that redundancy is facilitated, and the signal to noise ratio is improved; and judging whether the difference identification section is abnormal according to the Manhattan operation result of the identification section and the comparison section and the three sigma principle, thereby ensuring the robustness of the identification result.

Drawings

FIG. 1 is a flow chart of the present invention.

Detailed Description

The technical scheme protected by the invention is specifically described below with reference to the accompanying drawings.

Please refer to fig. 1. The invention provides a method for identifying abnormal signals of a cutter, which comprises the following steps:

(1) Performing DTW alignment on the acquired cutter initial signal and a standard template signal;

(2) Windowing and sliding are carried out on the two aligned signals, and the distance between sliding windows is calculated;

taking the position with the largest sliding window distance as the starting point of a difference section, starting from the starting point, and then taking all signal data as a primary section to be identified, wherein a section of signal corresponding to the primary section to be identified of a standard template signal is taken as a primary comparison section; i.e. the bifurcation point of the two signals is selected. Since the monitoring object may perform multiple processing actions, the processing actions are represented by multiple similar subsequences in the signal, and damage to the monitoring object (tool) may occur on any subsequence. Therefore, the influence of the healthy subsequence can be removed by performing the starting point selection operation, and the signal-to-noise ratio is improved, so that the discrimination accuracy is improved.

(3) In the invention, after the primary segment to be identified is selected, a more accurate identification segment is also required to be constructed, specifically, the Manhattan distance of each pair of sites in the primary segment to be identified and the primary comparison segment is compared, the maximum value of the Manhattan distance is selected as a starting point, and the identification segment and the comparison segment are constructed and obtained.

(4) Judging whether the difference identification section is abnormal according to the Manhattan operation results of the identification section and the comparison section and the three sigma principle, specifically, calculating corresponding Manhattan distances for each pair of the positions of the identification section and the comparison section, and summing to obtain a mean value of the Manhattan distances; comparing the Manhattan mean value with a preset threshold value, and judging whether the difference identification section is abnormal or not according to a three-sigma principle.

The Manhattan mean value is selected to enable the calculation result to be more file, if the starting point of the identification segment determined in the last step is only identified as an abnormal point because of a single abnormal value in the signal, the influence of the abnormal value can be flattened to the whole comparison segment by averaging the subsequent pairing distances, and the influence is reduced. Meanwhile, if the real abnormality occurs at the tail of the signal to be judged, the Manhattan distance of the two partial signals is amplified because the number of the average value is small, so that a better judging effect can be obtained.

The invention also provides a standard template signal generation method, which specifically comprises the following steps:

creating an input data set, wherein the input data set is processing data of normal operation of a product in a set time, the product consists of N workpieces, and the processing data comprises processing signals of the N workpieces;

calculating the DTW distance between N workpieces to obtain a DTW distance matrix of the product, wherein the scale of the DTW distance matrix is N; if a certain product comprises three workpieces, the DTW matrix of the product is:

calculating the sum of the DTW distances of each workpiece and the rest of the workpieces respectively, and marking the sum as a row DTW distance sum;

and taking a signal after the dynamic time warping of the workpiece corresponding to the minimum value in the N line DTW distance sums as a standard template signal.

The foregoing description is only illustrative of the present invention and is not intended to limit the scope of the invention, and all equivalent structures or equivalent processes or direct or indirect application in other related technical fields are included in the scope of the present invention.

Claims (6)

1. A method of identifying a tool anomaly signal, comprising:

acquiring a cutter initial signal transmitted by a machine tool;

after aligning a cutter initial signal with a standard template signal, selecting a primary section to be identified in the cutter initial signal and a primary comparison section corresponding to the primary section to be identified in the standard template signal;

comparing the Manhattan distance of each pair of sites in the primary segment to be identified and the primary comparison segment, selecting the maximum value of the Manhattan distance as a starting point, and constructing and obtaining an identification segment and a comparison segment;

and judging whether the identification section is abnormal or not according to the Manhattan operation result of the identification section and the comparison section and the three sigma principle.

2. The method for identifying a tool abnormality signal according to claim 1, wherein determining whether the identified segment is abnormal according to a three sigma principle based on a manhattan operation result of a primary segment to be identified and a comparison segment comprises:

calculating corresponding Manhattan distances for each pair of the position points of the identification section and the comparison section, and summing to obtain a mean value to obtain a Manhattan distance mean value;

comparing the Manhattan mean value with a preset threshold value, and judging whether the difference identification section is abnormal or not according to a three-sigma principle.

3. The method for identifying a tool abnormality signal according to claim 1, wherein comparing the primary section to be identified with the primary comparison section to obtain the identification section and the comparison section, comprises:

calculating the Manhattan distance corresponding to each pair of position points of the primary section to be identified and the primary comparison section;

and selecting the alignment point with the maximum Manhattan distance as the starting point of the identification segment, and selecting the end point of the primary segment to be identified as the end point of the identification segment.

4. The method for identifying a tool abnormality signal according to claim 1, wherein, after aligning a tool initial signal with a standard template signal, selecting a primary segment to be identified in the tool initial signal, and a primary comparison segment corresponding to the primary segment to be identified in the standard template signal, comprising:

windowing and sliding are carried out on the cutter initial signal and the standard template signal, and the distance between sliding windows is calculated;

and taking the alignment point with the largest sliding window distance as the starting point of the primary section to be identified, and determining the end point of the primary section to be identified according to the size of the sliding window.

5. The method for identifying a tool abnormality signal according to claim 1, characterized in that the step of generating the standard template signal is as follows:

creating an input data set, wherein the input data set is processing data of normal operation of a product in a set time, the product consists of N workpieces, and the processing data comprises processing signals of the N workpieces;

calculating the DTW distances among N workpieces to obtain a DTW distance matrix of the product;

calculating the sum of the DTW distances of each workpiece and the rest of the workpieces respectively, and marking the sum as a row DTW distance sum;

and taking a signal after the dynamic time warping of the workpiece corresponding to the minimum value in the N line DTW distance sums as a standard template signal.

6. The method of claim 1, wherein the tool initiation signal is aligned with the standard template signal by DTW.

Images