An Adaptive Monitoring Method for Roughing Tool Wear of Titanium Alloy Aerospace Structural Components

Abstract

This invention belongs to the field of tool wear monitoring technology and discloses an adaptive monitoring method for tool wear during rough machining of titanium alloy aerospace structural parts. The steps are as follows: Step 1: Determine the tool wear monitoring threshold T0 based on the machine tool spindle load torque value; Step 2: Collect a set A of CNC machine tool spindle load torque values ​​during rough machining within a fixed period T1; Step 3: Perform low-pass filtering on the data in set A to obtain dataset B; Step 4: Perform kurtosis analysis and evaluation on set B to obtain dataset X; Step 5: Classify the data in dataset X using a support vector machine, and take the set with the largest number of data after classification to obtain dataset T; Step 6: Calculate the average torque value of dataset T; Step 7: Perform adaptive monitoring of tool wear based on the average torque value of each period according to the threshold T0.

Description

Technical Field

[0001] This invention belongs to the field of tool wear monitoring technology, specifically relating to an adaptive monitoring method for roughing tool wear of titanium alloy aerospace structural components. Background Technology

[0002] Titanium alloy aerospace structural components are critical structural products for aircraft. Their mechanical properties result in high cutting forces, leading to rapid tool wear and low machining efficiency, which severely impacts aircraft development cycle time. In traditional cutting processes, tool wear status is typically assessed based on the number of parts processed, cumulative machining time, part surface quality, abnormal cutting noises, and chip condition. These methods require operators with extensive practical experience and are not highly accurate. Therefore, exploring adaptive monitoring mechanisms for tool wear during the machining of aerospace titanium alloy structural components and developing reliable and effective tool wear monitoring technologies is of great significance for improving machining efficiency, enhancing part quality, reducing production costs, and extending the service life of machine tool systems.

[0003] Currently, most general monitoring systems have limited functionality, relying on preset alarm limits or monitoring by learning and fitting the machine tool spindle torque curve in the time domain. These methods require high product consistency and repeatability, making them only suitable for mass production. Aerospace structural components are typically produced in small batches with many varieties, and monitoring systems are incompatible with the characteristics of manufacturing complex aerospace structural components, failing to meet processing monitoring needs. Domestic and international research on monitoring the CNC machining process of multi-variety, small-batch aerospace structural components has largely focused on adding vibration and current sensors to the machine tool to collect high-frequency signals and monitor tool errors through frequency domain characteristics. This method requires a high degree of matching between machining characteristics and signals, necessitating machine tool modifications and the addition of additional sensors and computing equipment, resulting in high modification costs. Summary of the Invention

[0004] Purpose of the invention: To solve the above problems, this invention provides an adaptive monitoring method for tool wear during rough machining of titanium alloy aerospace structural components. By collecting and processing low-frequency machine tool torque data, the method analyzes and monitors the tool wear status during the rough machining stage of aerospace titanium alloy structural components, determines whether the tool status is abnormal, and issues an alarm. This monitoring method can effectively monitor tool breakage faults online.

[0005] An adaptive monitoring method for roughing tool wear of aerospace titanium alloy structural components, comprising the following steps:

[0006] Step 1: Determine the tool wear monitoring threshold T0 based on the machine tool spindle load torque value;

[0007] Step 2: Within a fixed period T1, collect the set A of CNC machine tool spindle load torque values ​​during the roughing process;

[0008] Step 3: Perform low-pass filtering on the data in set A to obtain dataset B;

[0009] Step 4: Perform kurtosis value analysis and evaluation on set B to obtain dataset X;

[0010] Step 5: Use a support vector machine to classify the dataset X, and take the set with the largest number of data after classification to obtain the dataset T;

[0011] Step Six: Calculate the average torque of dataset T.

[0012] Step 7: Average torque for each cycle based on threshold T0 Implement adaptive monitoring of tool wear.

[0013] Furthermore, in step one, within a fixed period T1, the spindle load torque value of the tool when the tool is in the middle wear stage is collected through the OPC UA server interface of the CNC machine tool, and the average value of the machine tool torque within this period is taken as the tool wear monitoring threshold T0.

[0014] Furthermore, in step two, set A = {a1, a2, ..., a...} i , ..., a k}, where k is the total amount of torque values ​​collected.

[0015] Furthermore, in step three, the low-pass filtering process is as follows:

[0016] Determine the values ​​of each data a in dataset A. i Does it satisfy the following formula:

[0017]

[0018] In the formula, K is the average value of dataset A. k The Schauville coefficient is related to the quantity of data in dataset A and is obtained from the Schauville coefficient table. σ is the load torque value a in dataset A. i The mean squared error;

[0019] If the above formula is satisfied, the corresponding data value is removed; otherwise, it is retained. All retained data form a dataset B, B = {b1, b2, ..., b...} j , ..., b p}, where p is the number of torque values ​​in dataset B.

[0020] Furthermore, in step four, the kurtosis value analysis and evaluation process is as follows:

[0021] Divide the data in dataset B into p / m groups;

[0022] The kurtosis value K for each set of data is calculated using the following formula:

[0023]

[0024] In the formula, m represents the number of data sets in each group. If the kurtosis value K of a data set is greater than the threshold, then the data set is cleared to zero. All data sets whose kurtosis values ​​do not exceed the threshold constitute the dataset X.

[0025] Furthermore, in step five, the Support Vector Machine (SVM) algorithm, based on binary classification, searches for a hyperplane (ω·x) + b = 0 that satisfies the classification requirements, where ω is the normal vector, b is the bias vector, and x is a function. This hyperplane then divides the data into two classes.

[0026] Furthermore, in step five, the specific process is as follows:

[0027] (1) Dataset X = {x1, x2, ..., x} j , ..., x n The hyperplane (ω·x) + b = 0 divides the dataset into two classes, and the sample dataset {(x)} is divided into two classes. j y i The group x, i = 1, 2, ..., n, consists of two categories. If x i If it belongs to category 1, it is marked as positive (y). j =+1); if x i If it belongs to category 2, it is marked as negative (y). i =-1), represented as:

[0028]

[0029] (2) Normalize the classification equation (ω·x) + b = 0, and the sample set (x) i y i ), y j ∈{+1, -1}, satisfying y j [(ω·x j [)+b]-1≥0;

[0030] (3) Any point x i The distance d to the hyperplane (ω·x)+b=0 is expressed as

[0031]

[0032] (4) When the classification margin d is maximized, the classification plane (ω·x) + b = 0 is the optimal plane. The optimal solution is obtained by solving the following optimization problem.

[0033]

[0034] The Lagrange multiplier method is used to solve this quadratic programming problem with linear constraints, as follows:

[0035]

[0036] Where α i It is a Lagrange multiplier and satisfies 0 ≤ α j ≤C, where C is the penalty factor, j = 1, 2, ..., n;

[0037] (5) Substitute the sample data (x) i y i Solving the above quadratic programming problem function yields the Lagrange multiplier vector.

[0038] (6) α * Substitute Get ω * ;

[0039] (7) Take α * A positive component Calculate b * =y j -ω * x j , get b * ;

[0040] (8) Substitute ω * b * The decision function is obtained as: f(x) = sgn[(ω * ·x i )+b * ];

[0041] (9) For the given sample x of unknown type i Calculate f(x) = sgn[(ω * ·x j )+b * This allows us to determine the category to which the sample belongs.

[0042] (10) After dividing the dataset X into two categories using a function, we take the set with larger data volume as the tool smooth cutting dataset T.

[0043] Furthermore, in step seven, if the average torque is within the threshold T0 ± 10% T0 range, the tool wear is normal; otherwise, stop the machine for inspection.

[0044] The beneficial effects of this application are as follows:

[0045] This invention proposes an adaptive monitoring method for rough machining tool wear of titanium alloy aerospace structural components. This method can calculate and monitor the tool wear state during the rough machining stage of aerospace titanium alloy structural components, set a tool wear monitoring threshold to monitor changes in machine tool torque, determine if the tool condition is abnormal, and trigger an alarm. This monitoring method can effectively monitor tool breakage faults online. Utilizing this adaptive rough machining tool wear monitoring method can effectively avoid damage to parts and machine tools caused by abnormal tool conditions, improve stability during machining, reduce scrap rates and production risks, and has advantages such as good versatility, low data requirements, simple algorithm, fast response speed, and zero-cost modification. Attached Figure Description

[0046] Figure 1 This is a flowchart of the adaptive monitoring method for roughing tool wear provided by the present invention.

[0047] Figure 2 This is a graph showing the torque data after the spindle load torque variable A has been processed by low-pass filtering.

[0048] Figure 3 This is a graph of torque data after the torque variable B has been processed by kurtosis value.

[0049] Figure 4 This is a graph of torque data after the torque variable X has been classified by a support vector machine. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0051] The terms "first," "second," etc., used in the specification and claims of this invention are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0052] An adaptive monitoring method for roughing tool wear of aerospace titanium alloy structural components includes the following steps:

[0053] Step 1: Within a fixed period T1, collect the spindle load torque value of the tool when the tool is in the middle wear stage, and take the average value of the machine tool torque within this period as the tool wear monitoring threshold T0.

[0054] Step 2: Within a fixed period T1, collect the set A of CNC machine tool spindle load torque values ​​A during the roughing process, where A = {a1, a2, ..., a...} i , ..., a k}, where k is the total amount of torque values ​​collected;

[0055] Step 3: Filter dataset A to obtain dataset B, where B = {b1, b2, ..., b} j , ..., b p}, where p is the number of torque values ​​after filtering;

[0056] Step 4: Perform kurtosis value analysis and evaluation on B to obtain dataset X, where X = {x1, x2, ..., x} j , ..., x n}, where n is the number of torque values ​​after kurtosis processing;

[0057] Step 5: Classify X using a support vector machine to obtain dataset T, where T = {t1, t2, ..., t} j , ..., t s}, where s is the number of torque values ​​after data classification, and s < n < p < k;

[0058] Step 6: Calculate the average torque after data classification.

[0059] Step 7: Calculate the average torque for each cycle based on the adaptive tool wear monitoring threshold T0. Implement adaptive monitoring of tool wear.

[0060] In step 1, the spindle load torque value of the CNC machine tool is collected through the OPC UA server interface of the CNC machine tool.

[0061] In step 3, filtering the spindle load torque value set A is performed to weaken and eliminate the influence of noise, thereby improving the measurement accuracy of the system. The filtering process uses the Schauville criterion to remove outliers in the machine tool torque. The actual filter algorithm used is as follows:

[0062]

[0063] In the formula, K is the average value of dataset A. k The Schauville coefficients are related to the number of data sets A (obtained from the Schauville coefficient table), and σ is the load torque value a in data set A.i The mean squared error.

[0064] Eliminate set A from those that satisfy the Schauville criterion. The sampled values ​​were collected and the storage unit was cleared. This process was repeated until all obvious interference items were removed, resulting in dataset B. After filtering, the sampled data was initially stabilized, avoiding monitoring errors caused by outliers.

[0065] In step 4, the kurtosis value processing involves analyzing and evaluating the kurtosis of the data set. As tool wear increases, the increase in cutting force caused by the tool gradually becoming duller changes gradually over time. At special locations such as the tool's entry, exit, and corners within the workpiece, the cutting amount experiences sudden increases and decreases, resulting in larger kurtosis values. In these cases, the data is useless for judging the tool wear state. Therefore, by analyzing and evaluating the data kurtosis, large kurtosis data is removed, retaining the stable cutting state. The p data points in dataset B are divided into p / m groups, and the kurtosis value processing calculation method is as follows:

[0066] Kurtosis value K:

[0067]

[0068] In the formula, m refers to the number of torques in the data set to be processed. If the kurtosis value K of the data set is greater than 1.6, the data set is cleared to zero. This process is repeated for each data set until all sets with large kurtosis values ​​are removed, and finally, the dataset X is obtained.

[0069] In step 5, within a fixed cycle time, the tool feed and retraction processes are shorter than the smooth cutting process. Therefore, the dataset X needs to be divided into two classes, and the set with the largest data volume is taken as the smooth cutting dataset. The support vector machine algorithm is based on binary classification, searching for a hyperplane (ω·x) + b = 0 that satisfies the classification requirements (where ω is the normal vector, b is the bias vector, and x is a function), such that this hyperplane divides the data into two classes.

[0070] Furthermore, the support vector machine data classification method is as follows:

[0071] (1) Dataset X = {x1, x2, ..., x} j , ..., x n The hyperplane (ω·x) + b = 0 divides the dataset into two classes, namely the sample dataset {(x)}. j y j The group x, i = 1, 2, ..., n, consists of two categories. If x i If it belongs to category 1, it is marked as positive (y). i =+1); if x i If it belongs to category 2, it is marked as negative (y). i =-1), can be represented as:

[0072]

[0073] (2) Normalize the classification equation (ω·x) + b = 0, and the sample set (x) i ,y i ), y j ∈{+1,-1}, satisfying y j [(ω·x j )+b]-1≥0.

[0074] (3) Any point x i The distance d to the hyperplane (ω·x)+b=0 can be expressed as:

[0075]

[0076] (4) When the classification margin d is maximized, the classification surface (ω·x) + b = 0 is the optimal plane, which can be obtained by solving the following optimization problem.

[0077]

[0078] The Lagrange multiplier method is used to solve this quadratic programming problem with linear constraints, namely...

[0079]

[0080] Where α i It is a Lagrange multiplier and satisfies 0 ≤ α j ≤C, where C is the penalty factor, j = 1, 2, ..., n.

[0081] (5) Substitute the sample data (x) j y j Solving the above quadratic programming problem function yields the Lagrange multiplier vector.

[0082] (6) α * Substitute Get ω * ;

[0083] (7) Take α * A positive component Calculate b * =y j -ω * x j , get b * .

[0084] (8) Substitute ω * b * The decision function is obtained as follows:

[0085] f(x)=sgn[(ω * ·x i )+b * ]

[0086] (9) For the given sample x of unknown type i We only need to calculate f(x) = sgn[(ω * ·x i )+b * This allows you to determine the category to which the sample belongs;

[0087] (10) After dividing the dataset X into two categories using a function, we take the set with the larger data volume as the tool smooth cutting dataset T.

[0088] Example

[0089] Taking the roughing process of the titanium alloy aerospace structural component cabin door rocker arm on a five-axis linkage PV2216 CNC machine tool as an example, the specific steps of the adaptive monitoring method for tool wear in roughing of titanium alloy aerospace structural components proposed in this invention are as follows:

[0090] 1) The PV2216 five-axis CNC machine tool uses the OPC UA server interface to read the spindle load torque variable. In this embodiment, a fixed time period of 5 minutes is used, and approximately 5000 torque data points are collected in each time period.

[0091] 2) The Schauville criterion is used to eliminate outliers in machine tool torque. The actual filter algorithm used is as follows:

[0092]

[0093] Eliminated according to the Schauville criterion Sample value X i If Then clear the memory cell, and check each item one by one until all obviously interfering items are removed. For example... Figure 2 As shown.

[0094] 3) Analyze and evaluate the kurtosis of the data, remove data with high kurtosis, and retain the smooth cutting state.

[0095] The kurtosis value processing and calculation method is as follows:

[0096] In this embodiment, to ensure the results are as accurate as possible, m = 10, meaning each data set contains 10 data points. The kurtosis value K for each data set is calculated. If the kurtosis value K > 1.6, the value for that data set is reset to zero. This process is repeated for each data set until all sets with large kurtosis values ​​are removed. The data processing results are as follows: Figure 3 As shown.

[0097] 4) Support Vector Machines (SVMs) divide the data into two classes based on consistency, thus obtaining the optimal classification function:

[0098] f(x) = 0.081x j -235.2

[0099] Based on the above data classification, the largest set is taken as the tool smooth cutting data, such as... Figure 4 As shown.

[0100] 5) Based on experimental results and processing experience, when the cutting time of the tool is between 15 and 20 minutes, all types of roughing tools are in the mid-term wear stage. The average value of the machine tool torque at this time is taken as the reference value T0 for tool wear monitoring. In this embodiment, T0 = 10.81.

[0101] a) Average value of computer tool torque data after processing

[0102]

[0103] Where s represents the number of torque value data.

[0104] b) Based on on-site processing experience, the processing torque fluctuation should not exceed 10% of the reference value.

[0105] c) The formula for calculating the adaptive tool wear monitoring threshold is:

[0106]

[0107] The monitoring threshold serves as the tolerance range boundary of the machining curve. In subsequent machining, the machining process is monitored based on the threshold and the tolerance range boundary. If abnormal situations such as missing tools, broken tools, broken edges, or burning tools occur during rough machining, the machining curve will definitely have a sudden change or a large difference compared to the normal machining process. When the monitored torque value exceeds the preset monitoring threshold, the monitoring system automatically identifies this sudden change and difference, and then issues an alarm and controls the machine tool to stop.

[0108] 6) During the monitoring period of the example, no false alarms occurred. After 60 minutes of rough machining, to verify the accuracy of the monitoring, the machine was stopped to check the tool status. Observation revealed discoloration in some areas of the tool face and slight damage at the rounded corner of one insert. The tool was in the normal wear stage, and the monitoring system did not display any wear characteristics, so rough machining continued. When the system alarmed after 105 minutes of rough machining, the tool was in a damaged state; two of the four inserts showed slight damage. The damaged areas were approximately 0.6 square millimeters and 0.2 square millimeters. At this time, the machined surface of the part was normal, and the machine tool was cutting smoothly.

[0109] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of the present invention is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0110] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art, under the guidance of the present invention, can make many other forms without departing from the spirit and scope of protection of the claims, and all such forms are within the protection scope of the present invention.

Claims

1. An adaptive monitoring method for roughing tool wear of aerospace titanium alloy structural components, characterized in that: The steps are as follows: Step 1: Determine the tool wear monitoring threshold T0 based on the machine tool spindle load torque value; Step 2: Within a fixed period T1, collect the set A of CNC machine tool spindle load torque values ​​during the roughing process; Step 3: Perform low-pass filtering on the data in set A to obtain dataset B; Step 4: Perform kurtosis value analysis and evaluation on set B to obtain dataset X; Step 5: Classify the dataset X using a support vector machine, and obtain the dataset T by taking the set with the largest number of data after classification; the specific process is as follows: (1) Dataset Hyperplane Divided into two categories, sample datasets It consists of two categories, if If it belongs to category 1, then it is marked as positive ( );if If it belongs to category 2, it is marked as negative ( ), represented as: (2) For classification equations Normalize the sample set , ,satisfy ; (3) Any point To the hyperplane distance Represented as (4) Classification interval At its maximum, the classification surface To find the optimal plane, the optimal solution is obtained by solving the following optimization problem. The Lagrange multiplier method is used to solve this quadratic programming problem with linear constraints, as follows: in It is a Lagrange multiplier and satisfies , It is a punishment factor. ; (5) Input sample data Solving the above quadratic programming problem function yields the Lagrange multiplier vector. ; (6) Substitute ,get ; (7) Take A positive component ,calculate ,get ; (8) Substitute The decision function is obtained as follows: ; (9) For the given samples of unknown type ,calculate This allows us to determine the category to which the sample belongs; (10) Use functions to convert the dataset After dividing the data into two categories, we selected the set with the larger data volume as the tool smooth cutting dataset. ; Step Six: Calculate the average torque of dataset T. ; Step 7: Average torque for each cycle based on threshold T0 Implement adaptive monitoring of tool wear.

2. The method according to claim 1, characterized in that: In step one, within a fixed period T1, the spindle load torque value of the tool is collected through the OPC UA server interface of the CNC machine tool when the tool is in the middle wear stage. The average value of the machine tool torque within this period range is taken as the tool wear monitoring threshold T0.

3. The method according to claim 2, characterized in that: In step two, the set k represents the total amount of torque values ​​collected.

4. The method according to claim 3, characterized in that: In step three, the low-pass filtering process is as follows: Determine the individual data in dataset A Does it satisfy the following formula: In the formula, It is a dataset The average value, The Chauville coefficients are related to the number of elements in dataset A and can be obtained from the Chauville coefficient table. The load torque value in dataset A The mean squared error; If the above formula is satisfied, the corresponding data value is removed; otherwise, it is retained. All retained data constitutes dataset B. , where p is the number of torque values ​​in dataset B.

5. The method according to claim 4, characterized in that: In step four, the kurtosis value analysis and evaluation process is as follows: Divide the data in dataset B into Group; The kurtosis value K for each set of data is calculated using the following formula: In the formula, The number of data sets is determined by the following: if the kurtosis value K of a data set is greater than the threshold, that data set is cleared to zero. The dataset consists of all data sets whose kurtosis values ​​do not exceed the threshold. .

6. The method according to claim 5, characterized in that: In step five, the support vector machine algorithm, based on binary classification, searches for a hyperplane that satisfies the classification requirements. ,in Let b be the normal vector and b be the bias vector. Let be a function that divides the data into two classes through the hyperplane.

7. The method according to claim 6, characterized in that: In step seven, if the average torque is within the threshold T0 ± 10% T0 range, the tool wear is normal; otherwise, stop the machine for inspection.

Images