Detection system and method for tool magazine lubrication system
By monitoring and analyzing the pressure, temperature and flow parameters of the tool magazine lubrication system in real time, and using isolated models and logical models for abnormal detection and fault diagnosis, the problem of inability to effectively capture complex abnormal behavior in the existing technology is solved, and the system is efficient Fault diagnosis and timely alarm improve the stability and production continuity of the lubrication system.
Patent Information
- Application Number
- CN202411566648.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-05
- Publication Date
- 2024-12-03
AI Technical Summary
The existing tool magazine lubrication system detection technology cannot fully reflect the overall operating status of the lubrication system. It is difficult to capture complex abnormal behaviors based on simple thresholds. It lacks effective diagnostic methods, resulting in equipment failures and production interruptions.
A detection system is designed, including detection module, data processing module, feature extraction module, abnormality detection module and fault diagnosis module. By monitoring pressure, temperature and flow parameters in real time, abnormality detection and fault diagnosis are performed using isolated models and logical models. And promptly remind the operator through the alarm display module.
Real-time monitoring and fault diagnosis of tool magazine lubrication system are realized, detection efficiency is improved, faults caused by poor lubrication are reduced, and system stability and production continuity are ensured.
Smart Images

Figure CN119057558A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of tool magazine lubrication detection, and in particular to a detection system and method for a tool magazine lubrication system. Background Art
[0002] The tool magazine lubrication system is an important component of CNC machine tools. It is mainly used to provide necessary lubrication for the tools and mechanical parts in the tool magazine to reduce wear, reduce friction, and maintain efficient operation of the system. Good lubrication can not only extend the service life of the tool, but also improve processing accuracy and surface quality. With the continuous improvement of the level of automation in the manufacturing industry, enterprises have put forward higher requirements on the stability and reliability of the tool magazine lubrication system. Therefore, it is very important to detect the lubrication system to avoid equipment failure and production interruption caused by poor lubrication, thereby ensuring production continuity and product quality.
[0003] However, the detection technology currently used in the market for tool magazine lubrication systems still has some shortcomings. Most of the existing tool magazine lubrication systems have single monitoring and cannot fully reflect the overall operating status of the lubrication system; the existing abnormality detection methods usually rely on simple threshold comparisons, which are difficult to accurately capture complex abnormal behaviors; there is a lack of effective diagnostic methods for possible faults in the lubrication system, and it is often necessary to rely on manual experience for troubleshooting, which is time-consuming and inefficient; even if an abnormality or fault is detected, there is a lack of effective feedback mechanism, and the operator cannot be reminded to take measures in time, which may cause equipment damage or even shutdown.
[0004] Therefore, a system and method capable of real-time monitoring and timely feedback is needed to detect the lubrication system for the tool magazine. Summary of the invention
[0005] In view of at least one of the above technical problems, the present invention provides a detection system and method for a tool magazine lubrication system, which can effectively improve the real-time monitoring of the tool magazine lubrication system and can provide timely feedback on system fault diagnosis.
[0006] According to a first aspect of the present invention, there is provided a detection system for a tool magazine lubrication system, comprising:
[0007] A detection module is installed on the lubrication system and is used to collect observation values. The detection module includes a pressure sensor, a temperature sensor and a flow sensor. The pressure sensor monitors the pressure of the lubrication system. The temperature sensor monitors the temperature of the lubrication system. The flow sensor monitors the flow of the lubrication system. The pressure, temperature and flow together constitute the observation value set X.
[0008] The data processing module is used to clean and standardize the observation set X and generate the standardized array Z;
[0009] The feature extraction module is used to extract the sliding window mean μ and standard deviation σ of the standardized array Z according to the observation parameters and form the feature data Znew;
[0010] an anomaly detection module, including an isolation model, for anomaly detection of the lubrication system;
[0011] The fault diagnosis module includes a logic model and is used for performing fault diagnosis on the lubrication system.
[0012] Furthermore, the pressure sensor is installed at the outlet of the lubrication system; the temperature sensor is installed inside the lubrication system or at a key node of the lubricating fluid circulation path; and the flow sensor is installed at the inlet or outlet of the lubrication system.
[0013] Furthermore, it also includes an alarm display module, which is used to issue an alarm when an abnormality is detected and display the cause of the fault when a fault is detected.
[0014] According to a second aspect of the present invention, there is also provided a detection method for a tool magazine lubrication system, comprising the following steps:
[0015] S10: real-time monitoring of the pressure, temperature and flow rate of the tool magazine lubrication system, and forming an observation value set X with the values of these three observation parameters; the observation value set X is a two-dimensional numerical array, each column represents an observation parameter, the observation parameters include pressure, temperature, flow rate, and each row represents a group of observation values;
[0016] S20: Perform data cleaning and data standardization on the observation value set X to obtain a standardized array Z; each column of Z represents a standardized observation parameter, and each row represents a set of standard values;
[0017] S30: for the standardized array Z, extract the sliding window mean μ and standard deviation σ according to the observed parameters and form feature data Znew; Znew is a two-dimensional numerical array, each row represents a group of values to be measured, and each column represents an observed parameter;
[0018] S40: Use the isolated model to perform anomaly detection on the feature data Znew, send an abnormal signal and record the abnormal value when an abnormality is detected, and after the detection, arrange all abnormal values in the order of detection time to form abnormal data Zw, where Zw is an n×1 array;
[0019] S50: Use the logic model to perform fault diagnosis on the abnormal data Zw, and send a fault signal when it is determined that the tool magazine lubrication system has a fault;
[0020] S60: issuing an alarm according to the abnormal signal and displaying the cause of the fault according to the fault signal;
[0021] S70: Staff perform maintenance based on the alarm.
[0022] Furthermore, in step S20, the specific method of performing data standardization on the observation set X is as follows:
[0023] Generate a matrix Z, where the standard value of the nth row and mth column in Z is denoted by Z nm ;
[0024] Z nm =(X nm -μ m ) / σ m ;
[0025] Among them, X nm is the observation value in the nth row and mth column of the observation set X, μ m is the mean of all observations in the mth column of the observation set X, σ m is the standard deviation of all observations in the mth column of the observation set X.
[0026] Furthermore, the method further includes step S35: performing feature selection on the feature data Znew, generating a feature subset Zs, setting Znew=Zs, and updating the value in the feature data Znew for subsequent calculations.
[0027] Furthermore, in step S30, the sliding window mean μ and standard deviation σ are extracted as follows:
[0028] μ t =(Z t-w+1 +Z t-w+2 +...+Z t ) / w;
[0029] ;
[0030] Among them, Z t-w+1 To Z t is the value of Z in the window at time t; μ t is the mean value in the window at time t; w is the set window size; σ t is the standard deviation in the window at time t;
[0031] The feature selection in step S35 is specifically as follows:
[0032] Calculate the mean α of the observed parameters in Znew j =(1 / p)·(Z 11 +Z 12 +...+Z ij );
[0033] Where p is the number of rows in Znew; Z ij is the value of the jth column in the i-th row of Znew;
[0034] Generate the generating matrix \(Z_c\), and denote the value in the \(j\)-th column of the \(i\)-th row in \(Z_c\) as \(Z_c\). ij ; Calculate \(Z_c\). ij \(= Z\). ij \(-\alpha\);
[0035] Generate the covariance matrix \(A = Z_c\). T \(\cdot Z_c\cdot1 / (p - 1)\);
[0036] Solve for the eigenvectors \(v\) and the corresponding eigenvalues \(\lambda\) of \(A\);
[0037] Select the eigenvectors corresponding to the top \(k\) largest eigenvalues that are set, and form a matrix \(M\), where \(k < d\);
[0038] Generate the feature subset \(Z_s = Z_c\cdot M\), where each row represents an observation sample and each column represents an observation parameter.
[0039] Furthermore, in step S40, the specific method of using the isolation model to perform anomaly detection on the feature data \(Z_{new}\) is as follows:
[0040] Calculate \(E[c(T, Z_{new})]_{ij}=2H(n)-2(n - 1) / n\);
[0041] s ij \(= 2\). -E[c(T,Znew)]ij / c ;
[0042] Among them, \(E[c(T, Z_{new})]\) is the average path length of \(Z_{new}\) in the tree \(T\), and \(E[c(T, Z_{new})]_{ij}\) is the average path length of the value in the \(i\)-th row and \(j\)-th column in \(new\); \(T\) represents the decision tree; \(n\) is the total number of samples in the decision tree; \(C\) is the set number of trees; \(c\) is the maximum height of the tree; s ij is the anomaly score of the value in the \(i\)-th row and \(j\)-th column in \(Z_{new}\); \(H(n)\) is the harmonic series;
[0043] When \(0.5 < s < 1\), determine that the value in the \(i\)-th row and \(j\)-th column of the corresponding \(Z_{new}\) is an abnormal sample and record it; when \(0 < s\leq0.5\), determine that the value in the \(i\)-th row and \(j\)-th column of the corresponding \(Z_{new}\) is a normal sample.
[0044] Furthermore, in step S50, the specific method of using the logic model to perform fault diagnosis on the abnormal data \(Z_w\) is as follows:
[0045] Calculate \(\theta_a = Z_{w,i}\cdot F\). T ;
[0046] Among them, \(F\) is the set weight vector; \(Z_{w,i}\) is the \(i\)-th row in \(Z_w\), representing the \(i\)-th abnormal value;
[0047] Calculate \(h(\theta_a)=1 / (1 + e\). -θa-b );
[0048] Among them, b is the set bias term;
[0049] When h(θa)>0.5, it is judged that a fault occurs; when h(θa)≤0.5, it is judged that no fault occurs.
[0050] Furthermore, step S50 also includes optimizing the logic model using a loss function.
[0051] The technical solution of the present invention can achieve the following technical effects:
[0052] The present invention provides a detection system and method for a tool magazine lubrication system. By real-time monitoring of the pressure, temperature and flow parameters of the lubrication system, various parameters are combined into the same observation value set. Instead of first distinguishing and diagnosing whether the data is faulty, the abnormal values are first picked out and then the fault diagnosis is performed. This can save a lot of repeated calculation processes, improve detection efficiency, and provide real-time feedback on system faults. The staff can perform maintenance in a timely manner, thereby improving the stability of the lubrication system, making the tool magazine run more smoothly, and effectively reducing tool magazine faults caused by poor lubrication. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0054] Figure 1 It is a schematic diagram of the structure of a detection system for a tool magazine lubrication system in an embodiment of the present invention;
[0055] Figure 2 Schematic diagram of the steps of a detection method for a tool magazine lubrication system in an embodiment of the present invention. DETAILED DESCRIPTION
[0056] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.
[0057] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only and do not represent the only implementation method.
[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.
[0059] like Figure 1 A detection system for a tool magazine lubrication system is shown, comprising:
[0060] The detection module is installed on the lubrication system to collect observation values. The detection module includes a pressure sensor, a temperature sensor and a flow sensor. The pressure sensor is installed at the outlet of the lubrication system. The temperature sensor is installed inside the lubrication system or at key nodes in the lubricating fluid circulation path, such as near the oil pump, near the outlet, and other nodes that are prone to causing the lubricating fluid to heat up. The flow sensor is installed at the inlet or outlet of the lubrication system. Installing each sensor at a key location in the operation of the lubrication system according to its function is helpful for real-time monitoring of various parameters and diagnosis.
[0061] The pressure sensor monitors the pressure of the lubrication system; the temperature sensor monitors the temperature of the lubrication system; the flow sensor monitors the flow of the lubrication system; the pressure, temperature and flow together constitute the observation value set X;
[0062] The data processing module is used to clean and standardize the observation set X and generate the standardized array Z;
[0063] The feature extraction module is used to extract the sliding window mean μ and standard deviation σ of the standardized array Z according to the observation parameters and form the feature data Znew;
[0064] an anomaly detection module, including an isolation model, for anomaly detection of the lubrication system;
[0065] The fault diagnosis module includes a logic model and is used for performing fault diagnosis on the lubrication system.
[0066] It also includes an alarm display module, which is used to sound an alarm when an abnormality is detected and display the cause of the fault when a fault is detected, so as to promptly remind the staff to maintain the tool magazine lubrication system, thereby reducing the loss of time and products and improving the reliability of the system.
[0067] According to a second aspect of the present invention, Figure 2 As shown, a detection method for a tool magazine lubrication system is also provided, as shown in the figure, comprising the following steps:
[0068] S10: Real-time monitoring of the pressure, temperature and flow of the tool magazine lubrication system and recording the values, and then the values of these three observation parameters are combined into an observation value set X; the observation value set X is a two-dimensional numerical array, each column represents an observation parameter, the observation parameters include pressure, temperature, flow, and each row represents a group of observation values;
[0069] S20: Clean and standardize the observation set X to obtain a standardized array Z; each column of Z represents a standardized observation parameter, and each row represents a set of standard values; standardizing data is to eliminate the dimensional differences between different observation parameters;
[0070] S30: for the standardized array Z, extract the sliding window mean μ and standard deviation σ according to the observed parameters and form feature data Znew; Znew is a two-dimensional numerical array, each row represents a group of values to be measured, and each column represents an observed parameter;
[0071] S40: Use the isolated model to perform anomaly detection on the feature data Znew, send an abnormal signal and record the abnormal value when an abnormality is detected, and after the detection, arrange all abnormal values in the order of detection time to form abnormal data Zw, where Zw is an n×1 array;
[0072] S50: using the logic model to perform fault diagnosis on the abnormal array Zw, and sending a fault signal when it is determined that the tool magazine lubrication system has a fault;
[0073] S60: issuing an alarm according to the abnormal signal and displaying the cause of the fault according to the fault signal;
[0074] S70: Staff perform maintenance based on the alarm.
[0075] In step S20, the preferred method for performing data standardization on the observation set X is:
[0076] Generate a matrix Z, where the standard value of the nth row and mth column in Z is denoted by Z nm ;
[0077] Z nm =(X nm -μ m ) / σ m ;
[0078] Among them, X nm is the observation value in the nth row and mth column of the observation set X, μ m is the mean of all observations in the mth column of the observation set X, σ m is the standard deviation of all observations in the mth column of the observation set X;
[0079] Observations of different observed parameters often have different dimensions and measurement ranges. For example, pressure may be measured in bar or Pascal (Pa), while temperature may be measured in degrees Celsius (°C) or Kelvin (K). Pressure and temperature cannot be directly compared. Standardization eliminates this effect by converting all observations to dimensionless values, allowing observations of different types to be compared fairly.
[0080] Please refer to Figure 2 , and also includes step S35: performing feature selection on the feature array Znew, generating a feature subset Zs, and setting Znew=Zs, updating the value in the feature data Znew for subsequent calculations; through feature selection, highly repetitive parts in the data can be removed, redundant information can be reduced, and the efficiency of subsequent processing steps can be improved.
[0081] In step S30, the sliding window mean μ and standard deviation σ are extracted, specifically:
[0082] μ t =(Z t-w+1 +Z t-w+2 +...+Z t ) / w;
[0083] ;
[0084] Among them, Z t-w+1 To Z t is the value of Z in the window at time t; μ t is the mean value in the window at time t; w is the set window size; σ t is the standard deviation in the window at time t;
[0085] The sliding window mean reflects the average operating status of the lubrication system over a period of time. By observing the changes in the mean, we can analyze whether the system has a trend of gradual deterioration. The sliding window standard deviation reflects the fluctuation of the lubrication system over a period of time. The larger the standard deviation, the worse the stability of the system during this period. This can help us capture the changing trends and fluctuations of the lubrication system over time and identify abnormal situations.
[0086] The feature selection in step S35 is preferably:
[0087] Calculate the mean of each column in Znew, that is, the mean of each observed parameter; the formula for calculating the mean of the jth column is α j =(1 / p)·(Z 1j +Z 2j +...+Z ij ); α is the mean value of the observed parameters in Znew; the observed parameters include temperature, pressure and flow rate;
[0088] Where p is the number of rows in Znew, which is the maximum value of i; Z ij is the value of the jth column in the i-th row of Znew;
[0089] Generate a matrix Zc, where the value of the jth column of the i-th row in Zc is denoted by Zc ij ; Calculate Zc ij =Z ij -α; then subtract the mean of each column from its own column to obtain Zc ij ;
[0090] Centralizing the data is done by subtracting the mean of the observed parameter from each data point; centralizing the data helps when scaling the data so that all data have the same starting point, i.e., a mean of 0, to maintain consistency and better focus on changes in the data relative to its mean;
[0091] Then calculate the covariance matrix A=Zc T ·Zc·1 / (p-1);
[0092] Solve for the eigenvector v and the corresponding eigenvalue λ of A;
[0093] A is the covariance matrix. By calculating the covariance matrix, the matrix with unequal numbers of rows and columns is converted into a square matrix, paving the way for the subsequent calculation of eigenvalues and eigenvectors. The calculation of the covariance matrix can help us quantify the degree of correlation between different observation parameters.
[0094] Select the eigenvectors corresponding to the first k largest eigenvalues of a set number to form the matrix M, and k <d;
[0095] Generate feature subset Zs=Zc·M;
[0096] Zs is the data matrix after dimensionality reduction, where each row represents an observation sample and each column represents an observation parameter;
[0097] In this application, only three observation parameters are shown for reference, and observation parameters can be added according to usage requirements. Once the number of observation parameters is large, the subsequent calculation complexity will increase. This step can extract a few main parameters that have the greatest impact on the overall system from many observation parameters through feature selection, eliminating a large number of calculation processes, significantly reducing unnecessary calculation burdens, and improving analysis efficiency. It reduces the original multiple observation parameters to two main observation parameters, helping us to select the most relevant data subsets from the original data and identify the main direction of data changes, thereby reducing the dimension of the data; this reduces redundant information in the data, reduces calculation complexity and storage requirements, thereby better fitting the data, improving the accuracy of anomaly detection and robustness to changes in data distribution, and improving the efficiency of the algorithm.
[0098] Preferably, in step S40, the specific process of using the isolation model to perform anomaly detection on the feature data Znew is as follows:
[0099] Solve the average path length of each value of Znew on the decision tree. For the value in the i-th row and j-th column, we have:
[0100] Calculate E[c(T, Znew)]ij = 2H(n) - 2(n - 1) / n;
[0101] When solving each value, the isolation forest algorithm is adopted, which involves the construction of a set of decision trees with randomly generated structures. For each sample, traverse down the path of the tree until reaching the leaf node, and calculate its average path length on each decision tree. The average path length of this process can be used as a measure of anomaly. Among them, E[c(T, Znew)] is the average path length of Znew in tree T, and E[c(T, Znew)]ij is the average path length of the value in the i-th row and j-th column of new; T represents the decision tree; n is the total number of samples in the decision tree; C is the set number of trees; c is the maximum height of the tree; H(n) is the harmonic series. The more specific principle of the isolation forest algorithm is prior art and will not be elaborated here;
[0102] After obtaining the value of E[c(T, Znew)]ij, the isolation forest algorithm applies s ij = 2 -E[c(T,Znew)]ij / c to calculate the anomaly score s of the value in the i-th row and j-th column of Znew ij ;
[0103] When 0.5 < s < 1, the value in the i-th row and j-th column of the corresponding Znew is an abnormal sample and is recorded; when 0 < s ≤ 0.5, the value in the i-th row and j-th column of the corresponding Znew is a normal sample;
[0104] Applying the isolation forest algorithm in the detection system of the tool magazine lubrication system can not only effectively monitor the state of the lubrication system, but also improve the reliability and efficiency of the system through automated anomaly detection; this algorithm can work without labeled data and is very suitable for real-time monitoring tasks in industrial environments; its characteristics of high efficiency and strong adaptability enable it to quickly identify behaviors such as abnormal lubricating oil level, sudden temperature change, and pressure fluctuation that deviate from the normal operation range, and trigger an alarm to notify the maintenance personnel for inspection; compared with the supervised learning method that requires a large amount of labeled data, the unsupervised nature of the isolation forest algorithm reduces the implementation difficulty; more importantly, based on the assumption that abnormal samples are usually a small number and significantly different from other normal samples, by setting a reasonable threshold to screen samples with lower anomaly scores, it can effectively reduce the situation that samples that only slightly deviate from the normal range are misjudged as abnormal, thus effectively reducing the false alarm rate; this is crucial for avoiding unnecessary downtime and maintenance costs.
[0105] In step S50, the fault diagnosis of abnormal data Zw using the logical model is preferably:
[0106] Calculate θa = Zwi·F T , where F is the set weight vector; Zwi is the i-th row in Zw, representing the i-th outlier;
[0107] θa is a logistic regression algorithm, which describes the relationship between input features and output observation parameters by establishing a linear model, and uses the sigmoid function to convert the result of the linear combination into a probability form; the more specific principle of the logistic regression algorithm is the existing technology and will not be repeated here;
[0108] Then calculate h(θa)=1 / (1+e -θa-b );
[0109] Among them, b is the set bias term;
[0110] h(θa) is the probability of whether a given observation sample Zwi is faulty; the reason why this output value is defined as probability is that it can provide a continuous, non-negative value that does not exceed 1, which just conforms to the characteristics of probability and better explains the results of fault diagnosis;
[0111] When h(θa)>0.5, it is judged that a fault occurs; when h(θa)≤0.5, it is judged that no fault occurs.
[0112] In the detection system of the tool magazine lubrication system, the logistic regression algorithm can effectively utilize the data processed in the previous steps to more accurately identify the relationship between pressure, temperature and flow, and then accurately diagnose the potential faults of the lubrication system; the logistic regression algorithm has high computational efficiency and can complete the diagnosis of the system in a short time, which is especially important for application scenarios that require real-time monitoring of the lubrication system status; in addition, the output of the logistic regression algorithm is a continuous probability value, and the threshold can be adjusted according to actual conditions to balance the risks of false alarms and missed alarms, ensuring that the alarm is triggered and the cause of the fault is displayed at the appropriate time, thereby helping maintenance personnel to take quick action; the logistic regression algorithm not only improves the accuracy of fault diagnosis of the tool magazine lubrication system, but also ensures the stable operation of the system and reduces maintenance costs.
[0113] In step S50, it is preferred to also include optimizing the logic model using a loss function; the logistic regression algorithm can further improve performance by optimizing the loss function, and the loss function measures the gap between the model diagnosis results and the actual faults that occur in the system. When the loss function is minimum, the model is more accurate in detecting the tool magazine lubrication system; by minimizing this loss function to adjust the weight vector F and the bias term b in the model, it is ensured that as more data is accumulated, the model can be continuously optimized to adapt to new fault modes and maintain the accuracy of diagnosis.
[0114] Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. A detection system for tool magazine lubrication system, It is characterized in that include: A detection module is installed on the lubrication system and is used to collect observation values; the detection module includes a pressure sensor, a temperature sensor and a flow sensor, the pressure sensor is installed at the outlet of the lubrication system; the temperature sensor is installed inside the lubrication system or at a key node of the lubricating fluid circulation path; the flow sensor is installed at the inlet or outlet of the lubrication system; the pressure sensor monitors the pressure of the lubrication system; the temperature sensor monitors the temperature of the lubrication system; the flow sensor monitors the flow of the lubrication system; The pressure, the temperature and the flow together constitute an observation value set X; A data processing module is used to clean and standardize the observation set X and generate a standardized array Z; The feature extraction module is used to extract the sliding window mean μ and standard deviation σ of the standardized array Z according to the observation parameters and form the feature data Znew; an abnormality detection module, including an isolated model, for performing abnormality detection on the lubrication system; The fault diagnosis module includes a logic model and is used to perform fault diagnosis on the lubrication system.
2. The detection system for tool magazine lubrication system according to claim 1, It is characterized in that It also includes an alarm display module for sounding an alarm when an abnormality is detected and displaying the cause of a fault when a fault is detected.
3. A detection method for tool magazine lubrication system, It is characterized in that The following steps are involved: S10: real-time monitoring of the pressure, temperature and flow rate of the tool magazine lubrication system, and forming an observation value set X with the values of these three observation parameters; The observation set X is a two-dimensional numerical array, each column represents an observation parameter, including pressure, temperature, and flow, and each row represents a set of observation values; S20: Perform data cleaning and data standardization on the observation value set X to obtain a standardized array Z; each column of Z represents a standardized observation parameter, and each row represents a set of standard values; S30: for the standardized array Z, extract the sliding window mean μ and standard deviation σ according to the observed parameters and form feature data Znew; Znew is a two-dimensional numerical array, each row represents a group of values to be measured, and each column represents an observed parameter; S40: Use the isolated model to perform anomaly detection on the feature data Znew, send an abnormal signal and record the abnormal value when an abnormality is detected, and after the detection, arrange all abnormal values in the order of detection time to form abnormal data Zw, where Zw is an n×1 array; S50: Use the logic model to perform fault diagnosis on the abnormal data Zw, and send a fault signal when it is determined that the tool magazine lubrication system has a fault; S60: issuing an alarm according to the abnormal signal and displaying the cause of the fault according to the fault signal; S70: Staff perform maintenance based on the alarm.
4. The detection method for a tool magazine lubrication system according to claim 3, It is characterized in that In step S20, the specific method of performing data standardization on the observation set X is as follows: Generate a matrix Z, where the standard value of the nth row and mth column in Z is denoted by Z nm ; Z nm =(X nm -m m ) / s m ; Among them, X nm is the observation value in the nth row and mth column of the observation set X, μ m is the mean of all observations in the mth column of the observation set X, σ m is the standard deviation of all observations in the mth column of the observation set X.
5. The detection method for tool magazine lubrication system according to claim 3, It is characterized in that It further includes step S35: performing feature selection on the feature data Znew to generate a feature subset Zs, and setting Znew = Zs, and updating the values in the feature data Znew for subsequent calculations.
6. The detection method for the tool magazine lubrication system according to claim 5, wherein, the specific extraction of the sliding window mean μ and standard deviation σ in step S30 is as follows: μ t =(Z t-w+1 +Z t-w+2 +...+From t ) / In; ; Among them, Z t-w+1 To Z t is the value of Z in the window at time t; μ t is the mean value in the window at time t; w is the set window size; σ t is the standard deviation in the window at time t; the specific feature selection in step S35 is as follows: Calculate the mean α of the observed parameters in Znew j =(1 / p)·(Z 11 +Z 12 +...+Z ij ); Where p is the number of rows in Znew; Z ij is the value of the jth column in the i-th row of Znew; Generate a matrix Zc, where the value of the jth column of the i-th row in Zc is denoted by Zc ij ; Calculate Zc ij =Z ij -α; Generate the covariance matrix A=Zc T ·Zc·1 / (p-1); solving the eigenvector v and corresponding eigenvalue λ of A; selecting the eigenvectors corresponding to the top k largest eigenvalues set as the matrix M, and k < d; generating the feature subset Zs = Zc·M, where each row represents an observation sample and each column represents an observation parameter.
7. The detection method for the tool magazine lubrication system according to claim 3, wherein, in step S40, the specific abnormal detection of the feature data Znew using the isolation model is as follows: calculating E[c(T,Znew)]ij = 2H(n) - 2(n - 1) / n; s ij =2 -E[c(T,Znew)]ij / c ; Where E[c(T,Znew)] is the average path length of Znew in tree T, E[c(T,Znew)]ij is the average path length of the value of row i and column j in new; T represents the decision tree; n is the total number of samples in the decision tree; C is the number of trees set; c is the maximum height of the tree; s ij is the anomaly score of the value in row i and column j in Znew; H(n) is the harmonic series; when 0.5 < s < 1, determining that the value of the i-th row and j-th column of the corresponding Znew is an abnormal sample and recording it; when 0 < s ≤ 0.5, determining that the value of the i-th row and j-th column of the corresponding Znew is a normal sample.
8. The detection method for the tool magazine lubrication system according to claim 7, wherein, in step S50, the specific fault diagnosis of the abnormal data Zw using the logic model is as follows: Calculate θa=Zw,i·F T ; wherein, F is the set weight vector; Zw,i is the i-th row in Zw, representing the i-th abnormal value; Calculate h(θa)=1 / (1+e -θa-b ); wherein, b is the set bias term; when h(θa) > 0.5, it is determined that a fault has occurred; when h(θa) ≤ 0.5, it is determined that no fault has occurred.
9. The detection method for the tool magazine lubrication system according to claim 8, wherein, in step S50, it further includes optimizing the logic model using a loss function.
Citation Information
Patent Citations
Tool magazine performance and reliability detecting system
CN110389559A
Method for monitoring operation of lubricating system and preventing faults
CN114754280A
Lubricating oil monitoring and fault diagnosis device for diesel engine
CN115436226A
Lubrication Monitoring Arrangement, Hand-Held Cutting Tool and Method
US20200276661A1
Dynamic protection method for mechanical part of machine tool, and computer numerical control machine tool device
WO2023151166A1
