Cutter state monitoring and control method, device and equipment and storage medium

Abstract

The invention belongs to the technical field of numerical control machining, and discloses a tool state monitoring and control method, device and equipment and a storage medium. The method comprises the steps that a main shaft load torque signal and a servo motor current signal of a CNC machining center in the cutting process are collected; the collected torque signals and current signals are preprocessed through a programmable logic controller, and torque waveform data and current waveform data which are synchronous in time are generated; carrying out conjoint analysis on the preprocessed torque waveform data and current waveform data by adopting a signal processing algorithm, and extracting a feature set representing the dynamic load state of the cutter from a time domain and a frequency domain; matching the extracted feature set with a pre-established cutter state feature library, and identifying the current state of the cutter according to a matching result; and generating a corresponding control instruction based on the current state of the tool, and executing the control instruction through the programmable logic controller to control the machining process of the tool. According to the invention, the problems of high cost and difficulty in intelligent monitoring of an external detection device are solved.

Description

Technical Field

[0001] This invention belongs to the field of CNC machining technology, specifically relating to a tool status monitoring and control method, device, equipment, and storage medium. Background Technology

[0002] In the manufacturing process of CNC machining centers, real-time monitoring of tool status is crucial to ensuring machining quality and production efficiency.

[0003] Traditional tool condition monitoring methods primarily rely on external tool breakage detection devices, which are limited to detecting tool breakage. They struggle to effectively identify and warn of progressive or localized damage such as tool wear and chipping. The lack of dynamic monitoring of the entire tool condition (including normal cutting, wear, chipping, and breakage) easily leads to machining defects caused by undetected tool abnormalities. Examples include dimensional deviations and increased surface roughness due to tool wear, workpiece structural deformation due to chipping, and even machining interruptions and workpiece scrap due to sudden tool breakage. Furthermore, traditional monitoring methods typically require individual detection devices for each machine tool, resulting in high costs and hindering centralized, intelligent multi-machine collaborative monitoring and feedback control. Summary of the Invention

[0004] This invention provides a tool status monitoring and control method, device, equipment, and storage medium to solve the problems of high cost, lack of real-time closed-loop control capability, and difficulty in achieving centralized intelligent monitoring of multiple machines caused by reliance on external detection devices in the prior art.

[0005] To address the aforementioned technical problems, in a first aspect, the present invention provides a tool status monitoring and control method, comprising: Acquire spindle load torque signal and servo motor current signal during the cutting process of the CNC machining center; The collected torque and current signals are preprocessed by a programmable logic controller to generate time-synchronized torque waveform data and current waveform data. Signal processing algorithms are used to jointly analyze the preprocessed torque waveform data and current waveform data, and feature sets characterizing the dynamic load state of the tool are extracted from the time domain and frequency domain. The extracted feature set is matched with a pre-established tool state feature library, and the current state of the tool is identified based on the matching result. The current state of the tool includes at least normal cutting, abnormal wear, chipping, and breakage. Based on the current state of the cutting tool, corresponding control instructions are generated and executed by the programmable logic controller to control the machining process of the cutting tool.

[0006] Optionally, the step of using signal processing algorithms to jointly analyze the preprocessed torque waveform data and current waveform data to extract a feature set characterizing the dynamic load state of the tool from the time and frequency domains includes: The wavelet packet transform algorithm is used to perform multi-resolution analysis on the torque waveform data and current waveform data; Extract energy distribution features from different frequency bands as a subset of the feature set; The subsets are combined to form the feature set characterizing the dynamic load state of the tool.

[0007] Optionally, the step of employing signal processing algorithms for joint analysis to extract a feature set characterizing the dynamic load state of the tool from the time and frequency domains further includes: The empirical mode decomposition algorithm is used to decompose the preprocessed torque waveform data and current waveform data; Extract the instantaneous frequency features of several intrinsic mode function components as a subset of the feature set; The subsets are combined to form the feature set characterizing the dynamic load state of the tool.

[0008] Optionally, before jointly analyzing the preprocessed torque waveform data and current waveform data using signal processing algorithms to extract a feature set characterizing the dynamic load state of the tool from the time and frequency domains, the method further includes: The torque waveform data and the current waveform data are input into a pre-trained deep learning detection model; The features used to identify tool torque patterns are extracted by the deep learning detection model and used as the pre-established tool state feature library.

[0009] Optionally, before performing joint analysis of the preprocessed torque waveform data and current waveform data using signal processing algorithms to extract a feature set characterizing the dynamic load state of the tool from the time and frequency domains, the method further includes: Standardize the time-synchronized torque waveform data and current waveform data; Calculate the cross-correlation function between the standardized torque waveform data and the current waveform data, and obtain the maximum cross-correlation value of the cross-correlation function; The relative time delay is determined based on the maximum cross-correlation value; Phase compensation is performed on the torque waveform data and current waveform data based on the relative time delay to align them.

[0010] Optionally, the step of matching the extracted feature set with a pre-established tool state feature library and identifying the current state of the tool based on the matching result includes: Set dynamic thresholds and multi-level early warning mechanisms corresponding to different tool states; The feature set is matched with standard patterns in the feature library, and different levels of early warning or alarm are triggered based on the comparison between the matching result and the dynamic threshold.

[0011] Optionally, the step of executing the control instructions through the programmable logic controller to intervene in the processing includes: The control commands are received through a human-machine interface and sent to the programmable logic controller based on a signal slot mechanism. The load torque setpoint of the spindle can be dynamically adjusted by modifying the output parameters of the programmable logic controller; or, The programmable logic controller is triggered to execute emergency stop or tool change logic.

[0012] Secondly, the present invention provides a tool condition monitoring and control device, comprising: The signal acquisition module is used to acquire the spindle load torque signal and servo motor current signal of the CNC machining center during the cutting process; The generation module is used to preprocess the acquired torque signal and current signal through a programmable logic controller to generate time-synchronized torque waveform data and current waveform data. The feature extraction module is used to perform joint analysis on the preprocessed torque waveform data and current waveform data using signal processing algorithms, so as to extract a feature set characterizing the dynamic load state of the tool from the time domain and frequency domain. The status recognition module is used to match the extracted feature set with a pre-established tool status feature library, and identify the current status of the tool based on the matching result. The status includes at least normal cutting, abnormal wear, chipping, and breakage. The execution module is used to generate corresponding control commands based on the current state of the tool, and execute the control commands through the programmable logic controller to intervene in the machining process.

[0013] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described tool status monitoring and control method.

[0014] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described tool status monitoring and control method.

[0015] The aforementioned tool condition monitoring and control method, device, equipment, and storage medium achieves this solution by real-time acquisition of spindle load torque and servo motor current signals from a CNC machining center, and synchronous preprocessing using a built-in PLC to generate waveform data. Further, signal processing algorithms are employed to perform time-domain and frequency-domain joint analysis of the torque and current waveforms, extracting a feature set characterizing the tool's dynamic load state. This feature set is then matched with a pre-built tool condition feature library to accurately identify the real-time tool condition, including normal cutting, wear, chipping, and breakage. Finally, control commands are generated based on the identification results and executed by the PLC, achieving real-time intervention and closed-loop control of the machining process. This solution forms a complete technical closed loop from data perception and intelligent diagnosis to feedback control, systematically solving the problems of traditional monitoring methods, such as limited functionality, inability to identify wear and chipping, high cost due to reliance on externally purchased detection devices, and lack of real-time control capabilities. By directly utilizing the existing electrical control signals and PLC architecture of the machine tool, low-cost and high-precision full-state monitoring of the tool is achieved without modifying the mechanical structure or installing special sensors. Through algorithm model and feature library matching, the accuracy and timeliness of identifying progressive tool wear and sudden chipping are significantly improved. Finally, through a closed-loop control mechanism, process parameters can be adjusted in real time or protective actions can be triggered, thereby effectively reducing workpiece scrap, protecting the machine tool spindle, optimizing tool life, and supporting centralized monitoring of multiple machines and intelligent production management.

[0016] In summary, this solution can solve the problems of high cost due to reliance on external detection devices, lack of real-time closed-loop control capability, and difficulty in achieving centralized intelligent monitoring of multiple machines in existing technologies. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a tool status monitoring and control method in one embodiment of the present invention.

[0019] Figure 2 This is another flowchart illustrating the tool status monitoring and control method in one embodiment of the present invention.

[0020] Figure 3 This is another flowchart illustrating the tool status monitoring and control method in one embodiment of the present invention.

[0021] Figure 4 yes Figure 1A flowchart of step S140.

[0022] Figure 5 yes Figure 1 Another flowchart of step S140.

[0023] Figure 6 yes Figure 1 A flowchart of step S150.

[0024] Figure 7 yes Figure 1 Another flowchart of step S160.

[0025] Figure 8 This is a schematic diagram of a tool status monitoring and control device in one embodiment of the present invention.

[0026] Figure 9 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention.

[0027] Figure 10 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Please see Figure 1 As shown in the flowchart, an embodiment of the present invention provides a tool status monitoring and control method, which includes the following steps.

[0030] Step S110: Collect the spindle load torque signal and servo motor current signal of the CNC machining center during the cutting process.

[0031] It should be noted that the signal acquisition mechanism in step S110 is a non-intrusive, low-cost real-time sensing technology based on the existing electrical control architecture of the machine tool. Its core lies in utilizing the spindle servo drive system built into the CNC machining center itself as a "sensor," directly reading the raw electrical signals reflecting the cutting force state of the tool through standard industrial interfaces and data protocols, and constructing a basic time-series dataset for intelligent tool condition diagnosis. This mechanism is suitable for the real-time monitoring and early warning needs of tool wear, chipping, and breakage in various machining scenarios such as milling and turning.

[0032] At the data type level, the acquired signals include: spindle load torque signal: This signal directly characterizes the resistance torque experienced by the tool during cutting. It is typically acquired from the analog output port of the spindle servo drive in the form of an analog voltage signal, and this voltage value is linearly proportional to the real-time output torque of the drive; servo motor current signal: This signal is a direct electrical quantity that generates electromagnetic torque, and can quickly and sensitively reflect instantaneous changes in load. Its changing trend is strongly correlated with the torque signal and can be used for cross-validation and feature supplementation. These signals together constitute the electronic fingerprint of the dynamic load during the cutting process. Their waveform characteristics, such as amplitude, frequency components, ripple, and abrupt change points, are closely related to the tool's health status, providing the most basic input for subsequent feature extraction and state recognition algorithms.

[0033] Regarding the acquisition equipment and methods, it mainly relies on the programmable logic controller (PLC) and its analog input module within the CNC system. Specifically: Physical connection: The analog output terminal on the spindle servo drive, dedicated to outputting actual torque or motor current values, is connected to the PLC's analog input channel via a shielded cable; PLC program configuration: In the machine tool's PLC ladder diagram program, corresponding function blocks are added or modified to configure the input channel and set the range conversion coefficient; Data reading: The PLC reads and caches the converted digital torque and current values ​​in real time using a cyclic scanning method.

[0034] The advantage of this solution is that it eliminates the need for additional force sensors or current transformers on the machine tool's mechanical structure, maximizing the use of existing hardware resources and significantly reducing implementation costs and complexity. The sampling frequency is set based on the dynamic characteristics of the machining process and the requirements of the algorithm analysis. For conventional milling, to effectively capture potential chipping impacts or chatter frequency components, the sampling rate typically needs to be set above 1kHz. The PLC synchronously acquires two signals at this fixed frequency to ensure data timing alignment. The acquired raw data stream is first buffered in a ring within the memory of the PLC or a connected edge computing device, and then uploaded in real-time as data packets via industrial Ethernet (such as Profinet, Ethernet / IP) or TCP / IP protocols to the local monitoring computing unit or cloud server for use by subsequent preprocessing and intelligent analysis modules.

[0035] Step S120: The collected torque signal and current signal are preprocessed by the programmable logic controller to generate time-synchronized torque waveform data and current waveform data.

[0036] It should be noted that the preprocessing mechanism in step S120 is based on PLC real-time signal synchronization and quality enhancement technology. Its core lies in utilizing the determinism and high reliability of the PLC to digitize, align, filter, and format the two original analog signals, generating high-fidelity, strictly synchronized time-series waveform data. This provides high-quality, directly computable foundational data for subsequent precise algorithm analysis. This processing is a crucial bridge connecting raw sensing and intelligent diagnosis, directly determining the accuracy and reliability of feature extraction and state recognition.

[0037] At the level of preprocessing content and steps, the main components include: Timestamp alignment and synchronization: A unified, high-precision timestamp (usually based on the PLC's internal clock or external synchronization signal) is applied to the torque and current sample values ​​read by the PLC in each scan cycle to ensure complete alignment of the two signals in the time domain, eliminating minor delays that may be introduced by the acquisition channel or scanning sequence; Digital filtering and noise reduction: Software filters (such as moving average filtering and low-pass filtering) are applied to the PLC or edge computing unit to suppress interference from high-frequency electrical noise and driver switching ripple on signal quality, while retaining the effective frequency band components reflecting the cutting state. Filtering parameters can be adaptively set according to the spindle's rated speed and expected fault frequency; Engineering unit conversion and normalization: The raw digital quantities from the analog input channel (e.g., 0-27648 corresponding to -10V to +10V) are converted into engineering unit values ​​with clear physical meaning (e.g., torque in Nm, current in A) through a preset scaling factor. Simultaneously, the signals can be normalized according to the tool, workpiece material, and cutting parameters to eliminate baseline differences under different working conditions, making the data more comparable. Data packaging and format generation: The processed, timestamp-aligned torque and current values ​​are organized into structured data packets according to fixed time windows. Each data packet contains a time series array, forming torque waveform data and current waveform data that can be directly used for waveform analysis.

[0038] In terms of specific implementation, it is mainly accomplished through the collaborative operation of PLC programs and edge computing unit software modules: PLC internal processing: Within the PLC's loop program, dedicated function blocks are used to read analog-to-digital conversion values, append timestamps, perform basic range conversion, and perform initial-order filtering. The deterministic and real-time nature of the PLC ensures the timeliness of data acquisition and preprocessing. Edge computing unit in-depth processing: When preprocessing tasks are complex (such as high-order filtering and dynamic normalization), the PLC can stream the pre-processed data to a connected industrial computer (IPC) via industrial communication protocols (such as OPC UA and TCP Socket). The background service program running on the IPC utilizes its enhanced computing power to perform more refined digital filtering, synchronization checks, data packaging, and format conversion. Synchronization guarantee mechanism: To ensure strict synchronization, hardware interrupts based on the PLC clock can be used to trigger sampling, or a distributed I / O system with synchronization capabilities can be used. At the software level, the two data streams undergo timestamp-based interpolation alignment. The processed data is organized in a standardized format (such as JSON, CSV, or a custom binary format) and transmitted in real-time via a network interface to the computing node (local server or cloud) where the monitoring algorithm is deployed. The time-synchronized torque and current waveform data produced in this step are direct inputs for subsequent advanced signal analysis and state recognition, such as wavelet packet transform, empirical mode decomposition, and deep learning model inference. Their quality and synchronization accuracy are crucial foundations for achieving high-accuracy tool condition monitoring in the entire solution.

[0039] In some embodiments of the present invention, such as Figure 2 As shown, the following steps are included after step S120: Step S1311: Input the torque waveform data and the current waveform data into the pre-trained deep learning detection model; Step S1312: Extract features for identifying tool torque patterns using the deep learning detection model as the pre-established tool state feature library.

[0040] Specifically, in step S1311, the pre-trained deep learning detection model is an end-to-end feature learning and pattern recognition network designed specifically for time-series industrial signal analysis. Its input is the strictly time-synchronized torque and current waveform data pairs generated in step S120 (e.g., a two-dimensional time series with 5000 sampling points and a duration of 1 second for each sample). The core architecture of this model typically employs a one-dimensional convolutional neural network (1D-CNN) or a CNN-LSTM hybrid network incorporating an attention mechanism. The pre-training process is crucial for building its feature extraction capabilities: firstly, the network is supervisedly trained using massive amounts of historical machining data (covering various state samples such as normal cutting, wear, chipping, and breakage under different tools, materials, and process parameters). The goal of the training is to enable the model to automatically learn and abstract deep feature representations that most effectively distinguish different tool states directly from the raw, high-dimensional waveform data, rather than relying on manually preset signal processing rules. In the field of industrial monitoring, this data-driven feature learning method is gradually replacing traditional feature engineering based on fixed formulas and threshold rules, because it can capture the implicit, non-linear correlation between equipment status and complex signal patterns more flexibly and comprehensively.

[0041] Specifically, in step S1312, the feature extraction process occurs during the forward propagation of the pre-trained deep learning model. When new torque / current waveform data is input into the model, the data flows through each layer of the network and undergoes nonlinear transformation. Instead of directly using the model's final classification output, we use the output activation value of an intermediate layer in the network, typically before the last convolutional or fully connected layer, as the extracted feature vector. This feature vector is a highly abstract, low-dimensional, dense numerical array that condenses the core information most relevant to the tool state in the original waveform, and can be regarded as a deep fingerprint of the tool's current load mode. Subsequently, these deep feature vectors extracted from a large number of known state samples (normal, various levels of wear, typical chipping, breakage) are stored and organized together with their corresponding real state labels, thus forming the pre-established tool state feature library. This feature library is essentially a map in a high-dimensional feature space, where different categories of tool state samples form their own clustering regions. In subsequent online monitoring, the deep feature vectors extracted from new data can be quickly and accurately matched and identified by calculating their distance or similarity (such as cosine similarity or Mahalanobis distance) to the cluster centers of each category of features in the feature library. In the field of industrial predictive maintenance, building such a reference library based on deep learning features is fundamental to achieving intelligent and adaptive diagnostics. It allows the system to incrementally update the feature library when faced with new tool types or new machining materials, without having to redesign the entire signal processing flow.

[0042] Understandably, by deeply integrating the deep learning model into the feature extraction and feature library construction stages through steps S1311 and S1312, a technological paradigm upgrade has been achieved, moving from manually defined features and shallow model matching to data-driven learning of deep features and high-dimensional space matching. Traditional feature extraction methods heavily rely on expert experience and struggle to fully capture the complex, nonlinear, and multi-scale coupled hidden patterns that tool wear, chipping, and other anomalies may exhibit in torque and current signals. Pre-trained deep learning models, however, can automatically learn these optimal discriminative features end-to-end from massive amounts of data. This directly supports the monitoring system's ability to identify abnormal tool wear and chipping earlier and more accurately, effectively addressing the pain point of traditional detection devices' limited functionality. This deep learning-based feature learning and library construction framework demonstrates strong adaptability to varying processing conditions. Whether machining aluminum alloys or high-strength steel, as long as the pre-training data covers a sufficiently wide range of working conditions, the feature representations learned by the model possess strong generalization capabilities. Faced with new processing scenarios, the system can expand its monitoring capabilities by collecting new data from these scenarios, fine-tuning the model, and updating the feature library, without having to redevelop the algorithm from scratch. This technological framework can also be applied in the broader fields of Industrial Internet of Things (IIoT) and smart manufacturing: in fault prediction for rotating machinery (such as fans and pumps), deep learning can be used to extract deep features from vibration and temperature signals to build a fault spectrum library, enabling more refined fault type diagnosis; in semiconductor etching process monitoring, deep features can be learned from plasma emission spectra and a library built to identify process drift in real time. This effectively meets the core needs of modern industry for deeper state perception, intelligent diagnosis, and system adaptability.

[0043] In some embodiments of the present invention, such as Figure 3 As shown, the following steps are included after step S120: Step S1321: Standardize the time-synchronized torque waveform data and current waveform data; Step S1322: Calculate the cross-correlation function between the standardized torque waveform data and the current waveform data, and obtain the maximum cross-correlation value of the cross-correlation function; Step S1323: Determine the relative time delay based on the maximum cross-correlation value; Step S1324: Perform phase compensation on the torque waveform data and current waveform data according to the relative time delay to align the two.

[0044] Specifically, in step S1321, the standardization process aims to eliminate the incomparability between torque and current signals due to differences in dimensions, magnitudes, and baselines, transforming them into a unified numerical range to lay the foundation for subsequent accurate correlation analysis. In this scheme, the Z-score standardization method is typically used. This involves subtracting the mean (μ) from the entire analysis window data of each signal (e.g., 5000 points per second) and dividing by its standard deviation (σ), resulting in a new data sequence with a mean of 0 and a standard deviation of 1. This step is crucial because even though the PLC has performed preliminary engineering unit conversion, the absolute values ​​of the loads under different tools and cutting parameters vary greatly. Directly calculating the cross-correlation would result in the results being dominated by absolute amplitudes, failing to accurately reflect the consistency of the waveform shape changes between the two signals. In the field of industrial signal processing, standardization is a standard preprocessing operation before multi-source signal fusion and comparison. Similar standardization concepts are also prevalent in financial data analysis. For example, when constructing a multi-factor stock selection model, it is necessary to standardize factors of different dimensions (such as price-to-earnings ratio, market capitalization, and volatility) using Z-score so that they can be weighted and combined on the same scale, thus avoiding the model being dominated by factors with large numerical ranges (such as market capitalization, which is usually very large).

[0045] Specifically, in step S1322, the cross-correlation function is used to quantify the similarity between two signals under different relative time delays. For discrete-time series normalized torque and normalized current, the peak value of the calculated cross-correlation function directly reflects the waveform similarity of the two signals at optimal alignment; the closer its absolute value is to 1, the more synchronized and consistent the torque and current change patterns are. In an ideal physical system, the spindle torque and drive current should change highly synchronously. However, in actual systems, due to factors such as signal transmission paths, filter phase shifts, and mechanical response lag, there may be a slight inherent phase difference or time delay between the two. One of the purposes of this step is to quantitatively evaluate this synchronicity. In communication and radar signal processing, cross-correlation is a fundamental technique for time delay estimation and signal detection.

[0046] Specifically, in step S1323, the relative delay is determined by the delay parameter corresponding to the maximum value of the cross-correlation function calculated in step S1322. If the relative delay is greater than 0, it usually indicates that the change in the current signal leads the torque signal; if the relative delay is less than 0, it indicates that the torque signal leads. Under healthy processing conditions, this relative delay should be a relatively stable small value. Monitoring abnormal fluctuations in this delay can itself serve as a potential feature, suggesting possible system anomalies (such as loose transmission components causing response lag). In industrial fault diagnosis, the source of the fault can be located by analyzing the time delay between vibration signals at different measuring points. In high-frequency trading in finance, accurately determining the propagation delay of price information between different trading markets is crucial for judging arbitrage opportunity windows.

[0047] Specifically, in step S1324, phase compensation involves shifting (time-shifting) one of the signals based on the calculated relative time delay to eliminate the inherent time delay of the system and achieve precise alignment of the two signals on the time axis. The shifting operation typically employs linear interpolation or spline interpolation methods from digital signal processing to ensure the continuity of data points. After phase compensation, the torque and current signals should theoretically achieve the optimal instantaneous correspondence, meaning that the current and torque under load change are perfectly matched at any given time. This strict alignment is a prerequisite for subsequent multi-physical quantity collaborative analysis. Only with aligned signals can their phase relationship be accurately calculated and feature-level fusion performed, thereby more sensitively detecting subtle anomalies such as torque and current response mismatches that may occur in the early stages of tool wear.

[0048] Understandably, the signal coordination preprocessing chain consisting of steps S1321 to S1324 systematically solves the time misalignment problem caused by differences in physical transmission and system response characteristics among multiple sensor signals in actual industrial environments. This technical step is particularly critical in tool condition monitoring scenarios, as many subsequent advanced analysis algorithms are based on the assumption of strict synchronization between torque and current changes. Ignoring this inherent microsecond delay and directly performing joint analysis on the collected data may lead to distortion of extracted features, such as misjudging peaks that should be in phase as having a phase difference, thereby reducing the accuracy of the condition recognition model. This solution, through online, adaptive cross-correlation analysis and phase compensation, ensures that regardless of the machine tool model or driver configuration, it can automatically correct and output time-accurately matched data pairs, providing accurate input for upper-level algorithms. This significantly improves the robustness and accuracy of the entire monitoring system across different hardware platforms. This signal alignment framework has broad applicability. Similar precise time alignment techniques are needed as the foundation for advanced analysis in fields such as multi-channel vibration monitoring of rotating machinery and multi-node synchronous phasor measurement of power systems. The system automates time delay estimation and compensation through standardized processes, avoiding tedious manual debugging and meeting the requirements of intelligent manufacturing systems for adaptability and plug-and-play functionality.

[0049] Step S140: The collected torque signal and current signal are preprocessed by the programmable logic controller to generate time-synchronized torque waveform data and current waveform data.

[0050] It should be noted that this step is the key to converting raw signals into analyzable waveform data. Its core lies in using the real-time processing capability of the programmable logic controller to synchronously acquire, convert, and preliminarily shape multiple signals, providing a high-quality and highly consistent input data foundation for subsequent feature extraction and intelligent analysis.

[0051] In some embodiments of the present invention, such as Figure 4 As shown, step S140 includes the following steps: Step S1411: The wavelet packet transform algorithm is used to perform multi-resolution analysis on the torque waveform data and current waveform data; Step S1412: Extract energy distribution features of different frequency bands as a subset of the feature set; Step S1413: Combine the subsets to form the feature set characterizing the dynamic load state of the tool.

[0052] Specifically, in step S1411, the core of multi-resolution analysis lies in the rational selection of wavelet basis functions, determination of the number of decomposition levels, and execution of the decomposition algorithm to ensure coverage of all characteristic frequency bands that may be involved in tool state changes. Daubechies wavelets (such as db4 or db8) are preferred as basis functions because they have good localization characteristics, effectively balance time and frequency resolution, and are suitable for simultaneously capturing transient impacts (such as chipping) and gradual trends (such as wear) in the signal. The number of decomposition levels is adaptively determined based on the signal sampling frequency and the highest effective fault characteristic frequency to be analyzed, typically 4 to 6 levels, to obtain sufficiently fine frequency band division. This is achieved through a filter bank iterative algorithm. Using the low-pass and high-pass filters corresponding to the selected wavelet, the synchronized torque and current waveform data are filtered and downsampled layer by layer, recursively generating two complete wavelet packet coefficient trees. This process decomposes each original time-domain signal into a series of sub-bands with different center frequencies and bandwidths, realizing multi-scale observation of the signal from coarse to fine.

[0053] Specifically, in step S1412, the core of feature extraction is to quantify and standardize the energy of each sub-band, thereby forming a stable description of the signal's frequency domain energy structure. For each final sub-band obtained after wavelet packet decomposition, the sum of squares of all sample points in its coefficient sequence is calculated to obtain the energy of that sub-band. Subsequently, to eliminate the influence of the absolute amplitude of the signal, the energy of each sub-band is normalized, i.e., its energy is calculated as a proportion of the total energy of all sub-bands. This operation makes the features more robust to changes in cutting parameters; generating feature subsets: the above process is performed independently on the torque waveform and the current waveform respectively, resulting in two normalized energy proportion feature vectors. Each vector describes the energy concentration distribution of the corresponding signal in different fine frequency ranges. Changes in tool condition (such as wear, chipping) will cause changes in the characteristics of cutting force fluctuations, which in turn lead to predictable shifts in these energy distribution features.

[0054] Specifically, in step S1413, the core of feature set construction lies in effectively fusing frequency domain information from both torque and current channels, and further constructing enhanced synergistic features. Basic feature concatenation: The energy distribution feature vector extracted from the torque signal and the energy distribution feature vector extracted from the current signal are directly concatenated to form a joint feature vector. This vector simultaneously contains the frequency domain characteristics of the two key physical quantities. To more sensitively capture electromechanical response mismatch, the energy ratio difference or ratio of the two signals in the same sub-frequency band can be calculated to generate a set of synergistic features. These features reflect the correspondence between torque generation and current excitation in different frequency bands. The final output feature set is a fixed-dimensional numerical vector that comprehensively characterizes the frequency domain energy distribution pattern of the tool's dynamic load and the frequency domain synergistic relationship between torque and current signals within the current analysis window. This feature set will be directly used as input to the subsequent state recognition module.

[0055] It is understandable that the feature extraction and fusion process based on wavelet packet energy distribution, consisting of steps S1411 to S1413, provides a stable, precise, and physically meaningful description method for the dynamic load state of cutting tools. This technology is particularly crucial in tool condition monitoring scenarios. It overcomes the shortcomings of traditional time-domain statistical features or standard spectral analysis in handling non-stationary signals and transient anomalies. The multi-resolution characteristics of wavelet packet transform enable it to adaptively focus on different scales of the signal, accurately capturing features closely related to fault modes such as wear and chipping. Through normalization and dual-channel information fusion, the constructed feature set is robust to fluctuations in machining conditions and highly sensitive to fundamental changes in tool condition. Furthermore, this feature engineering framework has strong versatility; its idea of ​​transforming complex raw waveforms into stable frequency domain patterns is also widely used in fields such as industrial equipment vibration diagnosis and speech signal processing, serving as an effective bridge connecting low-level sensor data and high-level intelligent decision-making.

[0056] In some embodiments of the present invention, such as Figure 5 As shown, step S140 includes the following steps: Step S1421: The empirical mode decomposition algorithm is used to decompose the preprocessed torque waveform data and current waveform data; Step S1422: Extract the instantaneous frequency features of several intrinsic mode function components as a subset of the feature set; Step S1423: Combine the subsets to form the feature set characterizing the dynamic load state of the tool.

[0057] Specifically, in step S1421, the Empirical Mode Decomposition (EMD) algorithm is used to decompose the preprocessed torque and current waveforms. Its core is to decompose the complex non-stationary signal into a series of Intrinsic Mode Functions (IMFs) through an adaptive screening process.

[0058] The specific implementation is as follows: The decomposition process is performed separately for the synchronous waveform data of torque and current. For each signal, the algorithm identifies all local maxima and minima of the signal through iterative filtering steps, and fits the upper and lower envelopes using cubic spline interpolation; the mean of the upper and lower envelopes is calculated to obtain the mean envelope; this mean envelope is subtracted from the original signal to obtain a new signal component; this process is repeated until the newly generated component satisfies the two basic conditions of the IMF: the number of extreme points is equal to or differs from the number of zero crossings by at most one throughout the entire data segment; at any point, the mean of the upper and lower envelopes defined by the local maxima and minima is zero. The component that satisfies the conditions is an IMF. This IMF is separated from the original signal, and the above filtering process is repeated for the remaining signal until the remaining signal is a monotonic function or a constant (i.e., a residual term). Through this process, the original complex cutting load waveform is adaptively decomposed into several IMF components arranged from high frequency to low frequency (for example, the first three IMFs may contain high-frequency information such as tool chatter and chipping impact, while subsequent IMFs contain low-frequency information such as wear trend and spindle rotation) and a residual term. This data-driven decomposition method does not require pre-setting basis functions and is particularly suitable for processing nonlinear and non-stationary signals such as cutting processes, effectively separating the oscillation modes of different physical sources in the signal.

[0059] Specifically, in step S1422, the instantaneous frequency characteristics of several intrinsic mode functions (IMF) components are extracted. The core of this process lies in using the Hilbert transform to analyze each meaningful IMF component to obtain its time-varying frequency characteristics. The specific implementation is as follows: For each IMF component obtained from step S1421, a Hilbert transform is applied to construct an analytic signal. Based on this analytic signal, the instantaneous amplitude and instantaneous frequency of the IMF component at each sampling moment can be calculated. The instantaneous frequency reflects the instantaneous rate of change of the oscillation mode on the time axis. From these instantaneous frequency sequences, representative statistical features are extracted as a subset, which together constitute a feature subset describing the dynamic state of the tool from the perspective of the time-varying spectral structure within the signal. Compared to fixed frequency domain analysis, instantaneous frequency features can more precisely characterize the evolution of non-stationary dynamic processes.

[0060] Specifically, in step S1423, the subsets are combined to form the final feature set. The core of this process lies in the effective fusion and structuring of the multi-dimensional instantaneous frequency features extracted from the torque and current signals. The specific implementation is as follows: First, the instantaneous frequency feature subsets of the K IMFs obtained from the torque waveform decomposition and the K IMFs obtained from the current waveform decomposition are vertically concatenated to form a preliminary joint feature vector. To enhance the discriminative power of the features, cross-signal collaborative features can be further introduced: the difference or ratio between the instantaneous frequency features corresponding to the torque and current signals at the same or similar orders of IMF components is calculated to quantify the consistency of their dynamic responses in specific oscillation modes. Finally, these basic features and collaborative features are combined to form the final, high-dimensional feature set characterizing the dynamic load state of the tool. This feature set provides a deep description of the tool state from the perspective of the inherent, time-evolving oscillation modes of the signal and their interrelationships.

[0061] It is understandable that the feature extraction process based on empirical mode decomposition and instantaneous frequency analysis, consisting of steps S1421 to S1423, provides another powerful and complementary feature description method for this invention. This technology is particularly crucial in tool condition monitoring scenarios. Traditional fixed basis function transformations or time-frequency analysis based on preset scales may have limitations when dealing with highly non-stationary and nonlinear cutting signals. The complete adaptability of the EMD algorithm allows it to decompose based on the characteristics of the data itself, thereby more purely separating the signal components generated by different physical mechanisms. Furthermore, by performing Hilbert instantaneous frequency analysis on the IMF, the dynamic time-varying characteristics of these intrinsic oscillation modes in frequency can be captured, which has unique sensitivity for identifying the gradual process and transient events of tool condition. This feature set constructed from the perspective of the instantaneous dynamics of the signal's intrinsic modes effectively complements and enhances the feature set constructed from the perspective of the distribution of signal energy in a preset frequency band. By fusing features from these two perspectives, a more comprehensive and robust information input can be provided for subsequent state recognition models, thereby systematically improving the detection accuracy and reliability of complex states such as tool wear and tool chipping, fundamentally solving the detection challenges pointed out in the background technology. This methodological framework is not only applicable to tool monitoring, but its core ideas of adaptive decomposition and instantaneous feature extraction also have wide application value in industrial fields such as rotating machinery fault diagnosis and structural health monitoring, which require precise analysis of non-stationary dynamic signals.

[0062] Step S150: The extracted feature set is matched with the pre-established tool state feature library, and the current state of the tool is identified according to the matching result. The current state of the tool includes at least normal cutting, abnormal wear, chipping, and breakage.

[0063] It should be noted that this step is the core intelligent decision-making process that transforms the quantitative description obtained from front-end signal processing and feature extraction into a tool status conclusion with clear engineering semantics. Its core lies in using a pre-established feature library as a reference benchmark, and classifying and identifying real-time features through pattern matching algorithms, thereby achieving a mapping from data to state.

[0064] In some embodiments of the present invention, such as Figure 6 As shown, step S150 includes the following steps: Step S151: Set dynamic thresholds and multi-level early warning mechanisms corresponding to different tool states; Step S152: Match the feature set with the standard patterns in the feature library, and trigger different levels of early warning or alarm based on the comparison of the matching result with the dynamic threshold.

[0065] Specifically, in step S151, setting dynamic thresholds and a multi-level early warning mechanism is the core rule foundation for building an intelligent early warning system, aiming to achieve precise and graded responses to gradual and abrupt changes in tool status. The specific implementation is as follows: Dynamic threshold setting: The threshold is not a fixed value, but is dynamically calculated or selected based on the tool type, workpiece material, current cutting parameters, and historical tool service data. The system has a built-in parameter and threshold mapping table or a lightweight regression model; Multi-level early warning mechanism definition: Based on the severity and urgency of the anomaly, at least three levels of response mechanisms are established: Level 1 early warning (hint / observation level): Triggered when the deviation of the feature set from the normal cutting standard pattern first exceeds the lower limit of the dynamic threshold (e.g., deviation between 10% and 25%), or when certain sensitive features show early signs of anomalies. The system provides visual prompts on the human-machine interface and records logs, but does not force a shutdown, serving to alert operators; Level 2 alarm (early warning / intervention level): Triggered when the feature set highly matches the abnormal wear pattern, or when the deviation reaches the upper limit of the dynamic threshold (e.g., deviation between 25% and 50%), indicating that wear has intensified and may affect machining quality. The system issues audible and visual alarms and clearly displays a tool wear warning on the interface. It can also automatically recommend or execute logic to adjust cutting parameters (such as reducing the feed rate). A three-level alarm (emergency / stop level) is triggered immediately when the feature set successfully matches a chipping or breakage pattern, or when a critical feature (such as impact energy) undergoes an extreme change. The system issues the highest-level alarm and can send emergency stop or forced tool change commands to the CNC control system via the PLC to protect the workpiece and machine tool.

[0066] Specifically, in step S152, matching the feature set with the standard patterns in the feature library and triggering an early warning is a key decision-making step that transforms algorithm analysis into concrete action. The specific implementation is as follows: Feature matching method: Distance / similarity metric: Calculate the distance between the real-time extracted feature set and the standard feature vector (usually the center point of the sample features, such as the mean vector) of each state (normal, worn, chipped, broken) in the feature library. Common methods include Euclidean distance, Mahalanobis distance (considering the correlation between features), or cosine similarity. Classifier decision: Input the real-time feature set into a pre-trained multi-classification model (such as Support Vector Machine (SVM) or Random Forest), and the model directly outputs the probability of it belonging to each state category. The matching result is the category with the highest probability. Early warning triggering logic: If a distance metric is used, compare the calculated minimum distance (or the distance to the normal state) with the dynamic threshold set in step S151. For example, if the distance from the normal state exceeds the first-level threshold but does not reach the second-level threshold, a first-level warning is triggered; if it exceeds the second-level threshold and the distance from the center of the wear state is less than the distance from the normal state, a second-level alarm is triggered. If a classifier is used, the maximum probability value output by the model and its corresponding category are checked. For example, if it is classified as wear and the probability value of this category is higher than 85%, a second-level alarm is triggered; if it is classified as chipping or breakage, a third-level alarm is triggered immediately regardless of the probability. The system also performs continuity checks: anomalies in a single analysis window may be caused by interference. Therefore, the warning status is continuously monitored, and if a second-level alarm is triggered in three consecutive analysis windows, it is confirmed as a valid alarm and more aggressive intervention is implemented.

[0067] Understandably, through the dynamic threshold, hierarchical matching, and decision triggering mechanism constituted by steps S151 and S152, this invention achieves a closed loop from feature space to operation and maintenance decision-making, effectively transforming monitoring capabilities into control capabilities. This mechanism is particularly crucial in tool status monitoring scenarios. It avoids the problem of frequent false alarms or missed alarms caused by changes in working conditions in traditional fixed threshold methods, and improves the system's generalization ability and practicality through dynamically adaptive thresholds. The multi-level early warning mechanism realizes a gradient response from status prompts to proactive protection, providing operators with buffer judgment time, optimizing tool usage efficiency, and enabling decisive measures to be taken before or when major faults occur, minimizing losses. This mechanism, combined with a cloud architecture, allows threshold rules and early warning logic to be uniformly managed and optimized on a central server and synchronized to all networked devices, ensuring the consistency and iterability of monitoring strategies. This design concept is not only applicable to tool monitoring, but also has universal applicability to any industrial equipment (such as bearings, gearboxes, pumps and valves) that requires condition assessment and predictive maintenance. By defining dynamic thresholds and hierarchical response strategies corresponding to specific failure modes, a corresponding intelligent monitoring and protection system can be constructed, thereby comprehensively improving the reliability and intelligent management level of production equipment.

[0068] Step S160: Based on the current state of the tool, generate corresponding control instructions and execute the control instructions through the programmable logic controller to control the machining process of the tool.

[0069] It should be noted that this step is the key execution link to realize the closed loop from state perception to active control. Its core lies in transforming the state conclusions identified by the intelligent algorithm into specific, executable control actions, and applying them to the physical machining process through the machine tool's core control unit and programmable logic controller (PLC), thereby truly realizing the intervention and control functions of the monitoring system.

[0070] In some embodiments of the present invention, such as Figure 7 As shown, step S160 includes the following steps: Step S161: Receive the control command through the human-machine interface and send the command to the programmable logic controller based on the signal slot mechanism; Step S162: By modifying the output parameters of the programmable logic controller, the load torque setting value of the spindle is dynamically adjusted; or, the programmable logic controller is triggered to execute emergency stop or tool change logic.

[0071] Specifically, in step S161, receiving control commands through the human-machine interface and transmitting them to the programmable logic controller (PLC) is the pivotal link connecting intelligent decision-making and underlying execution. The specific implementation is as follows: Command Source and Interface Interaction: Control commands mainly come from two sources. First, automatic system generation: When a level 2 or 3 alarm is triggered, the status recognition module (step S150) automatically generates corresponding control commands (such as "load reduction command" or "emergency stop command"). Second, manual input: Operators or maintenance engineers can manually initiate control commands through the device's touchscreen human-machine interface based on alarm information and personal experience. The interface provides intuitive buttons or menus, such as adjusting parameters, performing tool changes, and confirming emergency stops. Signal Slot Mechanism Transmission: The system adopts a signal and slot software design pattern to achieve efficient and decoupled command transmission. When a control command is generated or confirmed at the application layer (monitoring software), it is issued as a specific signal. The slot function pre-bound to the PLC communication service immediately captures this signal. Subsequently, the communication service encapsulates the instruction into specific industrial communication protocol data frames (such as Profinet communication frames, Modbus TCP messages, etc.) based on the instruction content, and sends them to the programmable logic controller of the CNC system in real time and reliably via the industrial Ethernet network. This mechanism ensures low latency and high reliability in instruction response.

[0072] Specifically, in step S162, the execution of specific control actions by the programmable logic controller (PLC) is the final step in directly intervening in the machining process. The specific implementation is as follows: Dynamically adjusting the spindle load torque setting (corresponding to gradual abnormalities such as wear): Logic implementation: When the instruction is an adjustment parameter (such as reducing the load), the PLC receives a new torque setting value (which may be a percentage or an absolute value). The PLC program dynamically adjusts the upper limit or target value of the drive's torque by modifying the analog output channel value sent to the spindle servo drive or the torque limit parameter in the communication message. Execution effect: For example, when the system determines that the tool is in a moderate wear state, it automatically sends an instruction to reduce the maximum allowable torque of the spindle from 100% to 85%. The drive then limits the output current, thereby physically reducing the cutting force. This can alleviate the rate of further tool wear, provide a buffer time for scheduled tool changes, and prevent chipping or overcutting of the workpiece due to excessive cutting force. Triggering emergency stop or tool change logic (corresponding to sudden severe abnormalities such as chipping or breakage): Emergency stop logic: When the instruction is an emergency stop, the PLC will immediately set an emergency stop output signal. This signal is directly hardwired to the emergency stop safety circuit and the spindle driver enable circuit of the CNC system. Within milliseconds, it cuts off the spindle driver enable and feed axis enable, causing all motion axes to stop at the preset emergency deceleration rate. Tool changing logic: When the instruction is tool changing, the PLC calls and executes the standard automatic tool changing subroutine. First, the PLC controls the machine tool to complete the safe tool retraction action at the current machining point; then, it starts the tool magazine selection and tool changing robot process, replacing the damaged tool and loading a spare tool; finally, the tool offset can be reset as needed, and the operator is prompted for confirmation before machining resumes. The entire process is automated, minimizing downtime.

[0073] Understandably, through the closed-loop control link of instruction reception and logic execution constituted by steps S161 and S162, this invention ultimately achieves complete intelligent control from state perception to decision generation and then to physical execution. This stage is the direct manifestation and ultimate focus of this invention in addressing the pain point of traditional solutions lacking real-time control capabilities. It endows the monitoring system with true intervention capabilities, rather than merely alarm functions. Dynamic parameter adjustment enables adaptive machining, optimizing efficiency and extending tool life while ensuring safety. Emergency stop and automatic tool change provide the fastest and most reliable protection in the event of a fault, preventing costly workpiece scrapping and machine tool damage. By deeply integrating the control logic into the machine tool's original PLC, this solution ensures the determinism, high reliability, and safety of control instruction execution, fully complying with industrial control standards. This closed-loop control architecture upgrades this tool status monitoring device from an auxiliary monitoring tool to an intelligent control unit capable of actively participating in and optimizing the manufacturing process, providing a key technological guarantee for achieving unmanned and intelligent machining.

[0074] In summary, the solution implemented in the embodiments of the present invention can solve the problems of high cost, lack of real-time closed-loop control capability, and difficulty in achieving centralized intelligent monitoring of multiple machines caused by reliance on external detection devices in the prior art.

[0075] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. Software tools or components not belonging to this company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.

[0076] In one embodiment, a tool condition monitoring and control device is provided, which corresponds one-to-one with the tool condition monitoring and control methods described in the above embodiments. For example... Figure 8 As shown, the tool status monitoring and control device includes a signal acquisition module 810, a generation module 820, a feature extraction module 830, a status recognition module 840, and an execution module 850. Detailed descriptions of each functional module are as follows: The signal acquisition module 810 is used to acquire the spindle load torque signal and servo motor current signal of the CNC machining center during the cutting process. The generation module 820 is used to preprocess the acquired torque signal and current signal through a programmable logic controller to generate time-synchronized torque waveform data and current waveform data. The feature extraction module 830 is used to perform joint analysis on the preprocessed torque waveform data and current waveform data using signal processing algorithms, so as to extract a feature set characterizing the dynamic load state of the tool from the time domain and frequency domain. The state recognition module 840 is used to match the extracted feature set with a pre-established tool state feature library, and to identify the current state of the tool based on the matching result. The state includes at least normal cutting, abnormal wear, chipping, and breakage. The execution module 850 is used to generate corresponding control instructions based on the current state of the tool, and execute the control instructions through the programmable logic controller to intervene in the machining process.

[0077] In one embodiment, the feature extraction module 830 is specifically used for: The wavelet packet transform algorithm is used to perform multi-resolution analysis on the preprocessed torque waveform data and current waveform data; Extract energy distribution features from different frequency bands as a subset of the feature set; The subsets are combined to form the feature set characterizing the dynamic load state of the tool.

[0078] In one embodiment, the feature extraction module 830 is further specifically used for: The wavelet packet transform algorithm is used to perform multi-resolution analysis on the preprocessed torque waveform data and current waveform data; Extract energy distribution features from different frequency bands as a subset of the feature set; The subsets are combined to form the feature set characterizing the dynamic load state of the tool.

[0079] In one embodiment, the generation module 820 is specifically used for: The torque waveform data and the current waveform data are input into a pre-trained deep learning detection model; The features used to identify tool torque patterns are extracted by the deep learning detection model and used as the pre-established tool state feature library.

[0080] In one embodiment, the generation module 820 is further configured to: Standardize the time-synchronized torque waveform data and current waveform data; Calculate the cross-correlation function between the standardized torque waveform data and the current waveform data, and obtain the maximum cross-correlation value of the cross-correlation function; The relative time delay is determined based on the maximum cross-correlation value; Phase compensation is performed on the torque waveform data and current waveform data based on the relative time delay to align them.

[0081] In one embodiment, the state recognition module 840 is specifically used for: Set dynamic thresholds and multi-level early warning mechanisms corresponding to different tool states; The feature set is matched with standard patterns in the feature library, and different levels of early warning or alarm are triggered based on the comparison between the matching result and the dynamic threshold.

[0082] In one embodiment, the execution module 850 is specifically used for: The control commands are received through a human-machine interface and sent to the programmable logic controller based on a signal slot mechanism. The load torque setpoint of the spindle can be dynamically adjusted by modifying the output parameters of the programmable logic controller; or, The programmable logic controller is triggered to execute emergency stop or tool change logic.

[0083] This invention provides a solution for a tool condition monitoring and control device. It acquires real-time spindle load torque and servo motor current signals from a CNC machining center and uses a built-in PLC for synchronous preprocessing to generate waveform data. Further, signal processing algorithms are employed to perform time-domain and frequency-domain joint analysis of the torque and current waveforms, extracting a feature set characterizing the tool's dynamic load state. This feature set is then matched with a pre-built tool condition feature library to accurately identify the real-time tool condition, including normal cutting, wear, chipping, and breakage. Finally, control commands are generated based on the identification results and executed by the PLC, achieving real-time intervention and closed-loop control of the machining process. This solution forms a complete technical closed loop from data perception and intelligent diagnosis to feedback control, systematically solving the problems of traditional monitoring methods, such as limited functionality, inability to identify wear and chipping, high cost due to reliance on externally purchased detection devices, and lack of real-time control capabilities. By directly utilizing the existing electrical control signals and PLC architecture of the machine tool, low-cost and high-precision full-state monitoring of the tool is achieved without modifying the mechanical structure or installing special sensors. Through algorithm model and feature library matching, the accuracy and timeliness of identifying progressive tool wear and sudden chipping are significantly improved. Finally, through a closed-loop control mechanism, process parameters can be adjusted in real time or protective actions can be triggered, thereby effectively reducing workpiece scrap, protecting the machine tool spindle, optimizing tool life, and supporting centralized monitoring of multiple machines and intelligent production management.

[0084] In summary, this solution can solve the problems of high cost due to reliance on external detection devices, lack of real-time closed-loop control capability, and difficulty in achieving centralized intelligent monitoring of multiple machines in existing technologies.

[0085] Specific limitations regarding the tool condition monitoring and control device can be found in the limitations of the tool condition monitoring and control method described above, and will not be repeated here. Each module in the aforementioned tool condition monitoring and control device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0086] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements a recommended method for an optimal strategy, representing the functions or steps on the server side.

[0087] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements a recommended method for an optimal strategy, representing client-side functions or steps.

[0088] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Acquire spindle load torque signal and servo motor current signal during the cutting process of the CNC machining center; The collected torque and current signals are preprocessed by a programmable logic controller to generate time-synchronized torque waveform data and current waveform data. Signal processing algorithms are used to jointly analyze the preprocessed torque waveform data and current waveform data, and feature sets characterizing the dynamic load state of the tool are extracted from the time domain and frequency domain. The extracted feature set is matched with a pre-established tool state feature library, and the current state of the tool is identified based on the matching result. The current state of the tool includes at least normal cutting, abnormal wear, chipping, and breakage. Based on the current state of the cutting tool, corresponding control instructions are generated and executed by the programmable logic controller to control the machining process of the cutting tool.

[0089] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Acquire spindle load torque signal and servo motor current signal during the cutting process of the CNC machining center; The collected torque and current signals are preprocessed by a programmable logic controller to generate time-synchronized torque waveform data and current waveform data. Signal processing algorithms are used to jointly analyze the preprocessed torque waveform data and current waveform data, and feature sets characterizing the dynamic load state of the tool are extracted from the time domain and frequency domain. The extracted feature set is matched with a pre-established tool state feature library, and the current state of the tool is identified based on the matching result. The current state of the tool includes at least normal cutting, abnormal wear, chipping, and breakage. Based on the current state of the cutting tool, corresponding control instructions are generated and executed by the programmable logic controller to control the machining process of the cutting tool.

[0090] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0091] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0092] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0093] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for monitoring and controlling the condition of a cutting tool, characterized in that, include: Acquire spindle load torque signal and servo motor current signal during the cutting process of the CNC machining center; The collected torque and current signals are preprocessed by a programmable logic controller to generate time-synchronized torque waveform data and current waveform data. Signal processing algorithms are used to jointly analyze the preprocessed torque waveform data and current waveform data, and feature sets characterizing the dynamic load state of the tool are extracted from the time domain and frequency domain. The extracted feature set is matched with a pre-established tool state feature library, and the current state of the tool is identified based on the matching result. The current state of the tool includes at least normal cutting, abnormal wear, chipping, and breakage. Based on the current state of the cutting tool, corresponding control instructions are generated and executed by the programmable logic controller to control the machining process of the cutting tool.

2. The tool status monitoring and control method according to claim 1, characterized in that, The method employs signal processing algorithms to jointly analyze the preprocessed torque waveform data and current waveform data, extracting a feature set characterizing the dynamic load state of the tool from the time and frequency domains, including: The wavelet packet transform algorithm is used to perform multi-resolution analysis on the preprocessed torque waveform data and current waveform data; Extract energy distribution features from different frequency bands as a subset of the feature set; The subsets are combined to form the feature set characterizing the dynamic load state of the tool.

3. The tool status monitoring and control method according to claim 1, characterized in that, The method of employing signal processing algorithms for joint analysis to extract a feature set characterizing the dynamic load state of the tool from the time and frequency domains also includes: The empirical mode decomposition algorithm is used to decompose the preprocessed torque waveform data and current waveform data; Extract the instantaneous frequency features of several intrinsic mode function components as a subset of the feature set; The subsets are combined to form the feature set characterizing the dynamic load state of the tool.

4. The tool status monitoring and control method according to claim 1, characterized in that, Before the step of using signal processing algorithms to jointly analyze the preprocessed torque waveform data and current waveform data to extract feature sets characterizing the dynamic load state of the tool from the time and frequency domains, the following steps are also included: The torque waveform data and the current waveform data are input into a pre-trained deep learning detection model; The features used to identify tool torque patterns are extracted by the deep learning detection model and used as the pre-established tool state feature library.

5. The tool status monitoring and control method according to claim 1, characterized in that, Before employing signal processing algorithms to jointly analyze the preprocessed torque waveform data and current waveform data to extract a feature set characterizing the dynamic load state of the tool from the time and frequency domains, the method further includes: Standardize the time-synchronized torque waveform data and current waveform data; Calculate the cross-correlation function between the standardized torque waveform data and the current waveform data, and obtain the maximum cross-correlation value of the cross-correlation function; The relative time delay is determined based on the maximum cross-correlation value; Phase compensation is performed on the torque waveform data and current waveform data based on the relative time delay to align them.

6. The tool status monitoring and control method according to claim 1, characterized in that, The step of matching the extracted feature set with a pre-established tool state feature library and identifying the current state of the tool based on the matching result includes: Set dynamic thresholds and multi-level early warning mechanisms corresponding to different tool states; The feature set is matched with standard patterns in the feature library, and different levels of early warning or alarm are triggered based on the comparison between the matching result and the dynamic threshold.

7. The tool status monitoring and control method according to claim 1, characterized in that, The step of executing the control instructions through the programmable logic controller to intervene in the processing includes: The control commands are received through a human-machine interface and sent to the programmable logic controller based on a signal slot mechanism. The load torque setpoint of the spindle can be dynamically adjusted by modifying the output parameters of the programmable logic controller; or, The programmable logic controller is triggered to execute emergency stop or tool change logic.

8. A tool status monitoring and control device, characterized in that, include: The signal acquisition module is used to acquire the spindle load torque signal and servo motor current signal of the CNC machining center during the cutting process; The generation module is used to preprocess the acquired torque signal and current signal through a programmable logic controller to generate time-synchronized torque waveform data and current waveform data. The feature extraction module is used to perform joint analysis on the preprocessed torque waveform data and current waveform data using signal processing algorithms, so as to extract a feature set characterizing the dynamic load state of the tool from the time domain and frequency domain. The status recognition module is used to match the extracted feature set with a pre-established tool status feature library, and identify the current status of the tool based on the matching result. The status includes at least normal cutting, abnormal wear, chipping, and breakage. The execution module is used to generate corresponding control commands based on the current state of the tool, and execute the control commands through the programmable logic controller to intervene in the machining process.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the tool status monitoring and control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the tool status monitoring and control method as described in any one of claims 1 to 7.

Images