State prediction method and apparatus for computer numerical control machine tool, electronic device, and medium

Abstract

A state prediction method and apparatus for a computer numerical control machine tool, an electronic device, and a medium. The state prediction method for a computer numerical control machine tool comprises: acquiring a first vibration signal of a computer numerical control machine tool in an operation process (101); on the basis of the first vibration signal, determining the life cycle stage of the operation process in the full life cycle of the computer numerical control machine tool (102); and inputting the first vibration signal into a trained state prediction model corresponding to the life cycle stage, so that the state prediction model outputs a predicted state of the computer numerical control machine tool (103). By determining a specific life cycle stage on the basis of a mean-square value feature (13) of a vibration signal, and then using a state prediction model corresponding to the life cycle stage to perform state prediction of a computer numerical control machine tool, the differences in machine tool state in different life cycle stages are fully considered, improving the accuracy of state prediction. Moreover, the state prediction model is updated based on an excitation mechanism, improving the precision of the state prediction model.

Description

State prediction method, device, electronic equipment and medium for CNC machine tools Technical Field

[0001] The present invention relates to the technical field of numerically controlled machine tools, in particular to a method, device, electronic equipment and medium for predicting the state of a numerically controlled machine tool. Background Art

[0002] CNC machine tools, short for Computer Numerical Control Machine Tools, are automated machine tools equipped with a program-controlled system. This control system logically processes programs specified by control codes or other symbolic instructions, decodes them, and transmits them to a coded digital representation via an information carrier. After computational processing, the CNC device issues various control signals, controlling the machine's movements and automatically producing parts to the shape and size specified in the drawings.

[0003] As we all know, CNC machine tools are important equipment in various fields of production, and their safe and stable operation must be guaranteed. State prediction of CNC machine tools can provide early warning before failure occurs, allowing workers to take timely action and reduce losses.

[0004] Summary of the Invention

[0005] The embodiments of the present invention provide a method, device, electronic equipment and medium for predicting the state of a CNC machine tool.

[0006] A state prediction method for a numerically controlled machine tool, comprising:

[0007] Acquiring a first vibration signal of a CNC machine tool during operation;

[0008] determining, based on the first vibration signal, a life cycle stage of the operation process in the entire life cycle of the CNC machine tool;

[0009] The first vibration signal is input into a trained state prediction model corresponding to the life cycle stage, so that the state prediction model outputs a predicted state of the CNC machine tool.

[0010] Therefore, the implementation method of the present invention first determines the life cycle stage to which the operation process belongs, and then uses the state prediction model corresponding to the life cycle stage to execute the predicted state of the CNC machine tool, fully considering the differences in the machine tool states in different life cycle stages, and improving the accuracy of state prediction.

[0011] In one embodiment, determining the life cycle stage of the operation process in the entire life cycle of the CNC machine tool based on the first vibration signal includes:

[0012] The first vibration signal is input into a trained life cycle stage classification model so that the life cycle stage classification model outputs the life cycle stage.

[0013] Therefore, the life cycle stage can be determined based on artificial intelligence, reducing the amount of calculation.

[0014] In one embodiment, the training process of the life cycle stage classification model includes:

[0015] Acquire a second vibration signal of a test CNC machine tool of the same model as the CNC machine tool during its entire life cycle;

[0016] determining a time domain characteristic of the second vibration signal;

[0017] determining a normal period, a degradation period, and a failure period in the full life cycle based on a time domain characteristic of the second vibration signal;

[0018] Adding a first label to the time domain features in the normal period, adding a second label to the time domain features in the degradation period, and adding a third label to the time domain features in the failure period;

[0019] Based on the first label, the second label and the third label, the neural network model is trained as the life cycle stage classification model.

[0020] It can be seen that the proposed training method for the life cycle stage classification model is conducive to the efficient determination of the life cycle stage.

[0021] In one embodiment, the time domain feature includes a root mean square value;

[0022] The determining of the normal period, degradation period, and failure period in the entire life cycle based on the time domain characteristics of the second vibration signal includes:

[0023] Determine a time period in which the root mean square value is less than or equal to a first predetermined threshold as the normal period;

[0024] Determining a time period in which the root mean square value is greater than the first predetermined threshold and less than or equal to a second predetermined threshold as the degradation period;

[0025] Determining a time period during which the root mean square value is greater than the second predetermined threshold as the expiration period;

[0026] The second predetermined threshold is greater than the first predetermined threshold.

[0027] It can be seen that accurate training of the life cycle stage classification model is achieved through the mean square value feature.

[0028] In one embodiment, determining the life cycle stage of the operation process in the entire life cycle of the CNC machine tool based on the first vibration signal includes:

[0029] determining a root mean square value of the first vibration signal;

[0030] Based on the root mean square value of the first vibration signal, the life cycle stage of the operation process in the entire life cycle of the CNC machine tool is determined; when the root mean square value is less than or equal to a first predetermined threshold, the life cycle stage is determined to be a normal period; when the root mean square value is greater than the first predetermined threshold and less than or equal to a predetermined second predetermined threshold, the life cycle stage is determined to be a degradation period; when the root mean square value is greater than the second predetermined threshold for a time period, the life cycle stage is determined to be a failure period; wherein the second predetermined threshold is greater than the first predetermined threshold.

[0031] Therefore, the mean square value can be calculated directly to quickly determine the life cycle stage.

[0032] In one embodiment, the training process of the state prediction model corresponding to the life cycle stage includes:

[0033] Obtaining a third vibration signal of a test CNC machine tool of the same model as the CNC machine tool at the life cycle stage and a subsequent state of the machine tool after the third vibration signal;

[0034] The third vibration signal and the subsequent state are input into a bidirectional long short-term memory model, so that the bidirectional long short-term memory model sequentially performs the following steps: determining a time domain feature of the third vibration signal; obtaining a forward hidden state through a forward time cycle layer based on the time domain feature of the third vibration signal; obtaining a reverse hidden state through a reverse time cycle layer based on the time domain feature of the third vibration signal; concatenating the forward hidden state and the reverse hidden state to obtain a hidden layer state; determining a predicted state of the CNC machine tool based on the hidden layer state; determining a loss function value based on a difference between the predicted state and the subsequent state; and adjusting model parameters of the bidirectional long short-term memory model so that the loss function value is less than a predetermined threshold value.

[0035] The adjusted bidirectional long short-term memory model is determined as the trained state prediction model corresponding to the life cycle stage.

[0036] Considering that the amount of data from long-term monitoring of machine tools is large, when the time series data is too long, the hidden layer's memory ability for historical state information with a large time span will become weaker, and the prediction effect will be weakened. The embodiment of the present invention proposes an improved bidirectional long-short-term memory model to predict state data, which can improve the prediction effect of the model under large time spans.

[0037] In one embodiment, it includes:

[0038] After the state prediction model outputs the predicted state of the CNC machine tool, recording the actual state corresponding to the predicted state;

[0039] determining a difference between the predicted state and the actual state;

[0040] Based on the difference, the trained state prediction model is updated.

[0041] It can be seen that the incentive mechanism is also used to achieve adaptive updating of the state prediction model.

[0042] A state prediction device for a numerically controlled machine tool, comprising:

[0043] An acquisition module, used for acquiring a first vibration signal of the CNC machine tool during operation;

[0044] a determination module, configured to determine, based on the first vibration signal, a life cycle stage of the operation process in the entire life cycle of the CNC machine tool;

[0045] An input module is used to input the first vibration signal into a trained state prediction model corresponding to the life cycle stage, so that the state prediction model outputs a predicted state of the CNC machine tool.

[0046] Therefore, the implementation method of the present invention first determines the life cycle stage to which the operation process belongs, and then uses the state prediction model corresponding to the life cycle stage to execute the predicted state of the CNC machine tool, fully considering the differences in the machine tool states in different life cycle stages, and improving the accuracy of state prediction.

[0047] In one embodiment, the determination module is used to input the first vibration signal into a trained life cycle stage classification model so that the life cycle stage classification model outputs the life cycle stage; wherein the training process of the life cycle stage classification model includes: obtaining a second vibration signal of a test CNC machine tool of the same model as the CNC machine tool throughout its life cycle; determining the time domain characteristics of the second vibration signal; determining the normal period, degradation period and failure period in the entire life cycle based on the time domain characteristics of the second vibration signal; adding a first label to the time domain characteristics in the normal period, a second label to the time domain characteristics in the degradation period, and a third label to the time domain characteristics in the failure period; and training the neural network model as the life cycle stage classification model based on the first label, the second label and the third label.

[0048] Therefore, the life cycle stage can be determined based on artificial intelligence, reducing the amount of calculation. A training method for the life cycle stage classification model is also proposed, which is conducive to the efficient determination of the life cycle stage.

[0049] In one embodiment, the time domain feature is a root mean square value; the determination module is used to determine a time segment in which the root mean square value is less than or equal to a first predetermined threshold as the normal period; determine a time segment in which the root mean square value is greater than the first predetermined threshold and less than or equal to a second predetermined threshold as the degradation period; and determine a time segment in which the root mean square value is greater than the second predetermined threshold as the failure period; wherein the second predetermined threshold is greater than the first predetermined threshold.

[0050] It can be seen that accurate training of the life cycle stage classification model is achieved through the mean square value feature.

[0051] In one embodiment, the invention includes: a training module for executing a training process of a state prediction model corresponding to the life cycle stage, the training process comprising: obtaining a third vibration signal of a test CNC machine tool of the same model as the CNC machine tool in the life cycle stage and a subsequent state of the machine tool after the third vibration signal; inputting the third vibration signal and the subsequent state into a bidirectional long short-term memory model, so that the bidirectional long short-term memory model sequentially performs the following: determining a time domain feature of the third vibration signal; obtaining a forward hidden state through a forward time recurrent layer based on the time domain feature of the third vibration signal; obtaining a reverse hidden state through a reverse time recurrent layer based on the time domain feature of the third vibration signal; splicing the forward hidden state and the reverse hidden state to obtain a hidden layer state; determining a predicted state of the CNC machine tool based on the hidden layer state; determining a loss function value based on a difference between the predicted state and the subsequent state; adjusting model parameters of the bidirectional long short-term memory model so that the loss function value is less than a predetermined threshold; and determining the adjusted bidirectional long short-term memory model as the trained state prediction model corresponding to the life cycle stage.

[0052] Considering that the amount of data from long-term monitoring of machine tools is large, when the time series data is too long, the hidden layer's memory ability for historical state information with a large time span will become weaker, and the prediction effect will be weakened. The embodiment of the present invention proposes an improved bidirectional long-short-term memory model to predict state data, which can improve the prediction effect of the model under large time spans.

[0053] An electronic device, comprising:

[0054] processor;

[0055] a memory for storing executable instructions of the processor;

[0056] The processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the state prediction method for a CNC machine tool as described in any one of the above items.

[0057] A computer-readable storage medium stores computer instructions, wherein the computer instructions, when executed by a processor, implement the state prediction method for a numerically controlled machine tool as described in any one of the above items.

[0058] A computer program product comprises a computer program, wherein when the computer program is executed by a processor, the method for predicting the state of a numerically controlled machine tool as described above is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, so that those skilled in the art will understand the above and other features and advantages of the present invention more clearly. In the accompanying drawings:

[0060] FIG1 is an exemplary flow chart of a method for predicting a state of a CNC machine tool according to an embodiment of the present invention.

[0061] FIG2 is an exemplary schematic diagram of determining the life cycle stage of a current running process according to an embodiment of the present invention.

[0062] FIG3 is an exemplary structural diagram of a bidirectional long short-term memory model according to an embodiment of the present invention.

[0063] FIG4 is a schematic diagram of updating a trained state prediction model according to an embodiment of the present invention.

[0064] FIG5 is a schematic diagram of the overall process of state prediction and model maintenance of a CNC machine tool according to an embodiment of the present invention.

[0065] FIG6 is an exemplary structural diagram of a state prediction device for a numerically controlled machine tool according to an embodiment of the present invention.

[0066] FIG. 7 is a structural diagram of an electronic device according to an embodiment of the present invention.

[0067] The accompanying drawings are numerals as follows: DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail with reference to the following examples.

[0069] For the sake of brevity and intuitiveness in description, the solution of the present invention is explained below by describing several representative implementations. A large number of details in the implementations are only used to help understand the solution of the present invention. However, it is obvious that the technical solution of the present invention may not be limited to these details when implemented. In order to avoid unnecessarily obscuring the solution of the present invention, some implementations are not described in detail, but only a framework is given. Hereinafter, "including" means "including but not limited to", and "according to..." means "at least according to..., but not limited to only according to...". Due to the language habits of Chinese, when the number of a component is not specifically specified below, it means that the component can be one or more, or can be understood as at least one.

[0070] In an embodiment of the present invention, a CNC machine tool health monitoring method and system based on a multi-layer perception network is proposed. This method establishes a fault evolution law model based on operating data, focusing on the evolution of machine tool failures from early, weak features to complete failure, thereby enabling monitoring of the machine tool's health status. Considering that the process from normal operation to failure of a CNC machine tool is not a linear one, but rather involves a degradation period, determining the starting point of machine tool degradation has always been a challenge. This embodiment of the present invention proposes an initial degradation point detection method based on root mean square characteristics, thereby enabling fault prediction for machine tools in the degradation period.

[0071] In addition, considering that the machine tool is monitored for a long time and the amount of data is large, when the time series data is too long, the hidden layer's ability to remember historical status information with a large time span will become weaker, and the prediction effect will be weakened. The embodiment of the present invention proposes an improved bidirectional long short-term memory (BiLSTM) network to predict status data, which can improve the prediction effect of the model under a large time span. In addition, considering the use of predicted data to guide the operation and maintenance of machine tools, how to optimize the prediction model on this basis is also an important issue. The embodiment of the present invention proposes a set of CNC machine tool fault evolution and health management systems, which combines predicted data with actual operation data to optimize the prediction model.

[0072] FIG1 is an exemplary flow chart of a method for predicting the state of a CNC machine tool according to an embodiment of the present invention. As shown in FIG1 , the method includes:

[0073] Step 101: Acquire a first vibration signal of a CNC machine tool during operation.

[0074] Here, a variety of vibration sensors can be used to measure one or more vibration signals of one or more key components of a CNC machine tool during operation (preferably during the current operation) as the first vibration signal. Accordingly, the first vibration signal can be a single time point signal or a time series signal measured from one or more key components over a period of time. For example, the key components may include: a spindle, a feed box, a slide box, a tool holder, a tailstock, a guide rod, a lead screw, and a bed, etc.

[0075] Step 102: Based on the first vibration signal, determine the life cycle stage of the operation process in the entire life cycle of the CNC machine tool.

[0076] In one embodiment, step 102 includes inputting the first vibration signal into a trained life cycle stage classification model, so that the life cycle stage classification model outputs a life cycle stage. The training process of the life cycle stage classification model includes obtaining a second vibration signal from a test CNC machine tool of the same model as the CNC machine tool throughout its life cycle; determining time domain features of the second vibration signal; determining the normal period, degradation period, and failure period of the life cycle based on the time domain features of the second vibration signal; attaching a first label to the time domain features during the normal period, a second label to the time domain features during the degradation period, and a third label to the time domain features during the failure period; and training a neural network model into a life cycle stage classification model based on the first, second, and third labels.

[0077] Therefore, the life cycle stage can be determined based on artificial intelligence, which reduces the amount of calculation. In addition, a training method for the life cycle stage classification model is proposed, which is conducive to the efficient determination of the life cycle stage.

[0078] In one embodiment, the time-domain features include root mean square (RMS) values. Determining the normal period, degradation period, and failure period within the full life cycle based on the time-domain features of the second vibration signal includes: determining a time period during which the RMS value is less than or equal to a first predetermined threshold as a normal period; determining a time period during which the RMS value is greater than the first predetermined threshold and less than or equal to a second predetermined threshold as a degradation period; and determining a time period during which the RMS value is greater than a second predetermined threshold as a failure period; wherein the second predetermined threshold is greater than the first predetermined threshold. This demonstrates that accurate training of a life cycle stage classification model is achieved through the RMS value feature.

[0079] In one embodiment, step 102 includes: determining a root mean square (RMS) value of the first vibration signal; and determining the lifecycle stage of the CNC machine tool's operation process within its entire lifecycle based on the RMS value of the first vibration signal; determining the lifecycle stage to be normal when the RMS value is less than or equal to a first predetermined threshold; determining the lifecycle stage to be degraded when the RMS value is greater than the first predetermined threshold and less than or equal to a second predetermined threshold; and determining the lifecycle stage to be failure when the RMS value is greater than the second predetermined threshold for a period of time; wherein the second predetermined threshold is greater than the first predetermined threshold. Therefore, the RMS value can be directly calculated to quickly determine the lifecycle stage.

[0080] FIG2 is an exemplary schematic diagram of determining the life cycle stage of a current running process according to an embodiment of the present invention.

[0081] First, a sensor is used to obtain a horizontal vibration signal 11 and a vertical vibration signal 12 of the spindle during machine tool operation. Then, the horizontal vibration signal 11 and the vertical vibration signal 12 are synthesized, and a root mean square value 13 (e.g., average root mean square) of the synthesized vibration signal is calculated. Next, the root mean square value 13 is compared with a first predetermined threshold and a second predetermined threshold, respectively, to determine the life cycle stage to which the machine tool operation process belongs. Specifically, when the root mean square value is less than or equal to the first predetermined threshold, the machine tool operation process is determined to be in a normal period; when the root mean square value is greater than the first predetermined threshold and less than or equal to the second predetermined threshold, the machine tool operation process is determined to be in a degradation period; and when the root mean square value is greater than the second predetermined threshold, the machine tool operation process is determined to be in an expiration period; wherein the second predetermined threshold is greater than the first predetermined threshold.

[0082] Here, the first predetermined threshold and the second predetermined threshold can be empirical values ​​set by the user, or can be mechanism values ​​calculated based on the mechanism model. Moreover, the first predetermined threshold and the second predetermined threshold are preferably adjustable.

[0083] The above exemplary description is an exemplary process for determining the life cycle stage of the current running process. Those skilled in the art will appreciate that this description is merely exemplary and is not intended to limit the scope of protection of the embodiments of the present invention.

[0084] Step 103: Input the first vibration signal into a trained state prediction model corresponding to the life cycle stage, so that the state prediction model outputs a predicted state of the CNC machine tool.

[0085] As can be seen, in the embodiments of the present invention, a trained state prediction model is associated with each lifecycle stage. Therefore, by predetermining the specific lifecycle stage to which the operation process belongs and then using the state prediction model corresponding to that lifecycle stage to predict the state of the CNC machine tool, the differences in machine tool states at different lifecycle stages are fully accounted for, thereby improving the accuracy of state prediction.

[0086] In one embodiment, the method includes a training process of the state prediction model corresponding to the life cycle stage, and the training process includes: obtaining a third vibration signal of a test CNC machine tool of the same model as the CNC machine tool in the life cycle stage and a subsequent state of the machine tool after the third vibration signal, wherein the subsequent state can be characterized by a vibration signal at a predetermined time after the third vibration signal; inputting the third vibration signal and the subsequent state into a BiLSTM model, so that the BiLSTM sequentially performs the following: determining the time domain characteristics of the third vibration signal; obtaining a forward hidden state through a forward time cycle layer based on the time domain characteristics of the third vibration signal; obtaining a reverse hidden state through a reverse time cycle layer based on the time domain characteristics of the third vibration signal; splicing the forward hidden state and the reverse hidden state to obtain a hidden layer state; determining a predicted state of the CNC machine tool based on the hidden layer state; determining a loss function value based on the difference between the predicted state and the subsequent state; adjusting model parameters of the BiLSTM model so that the loss function value is less than a predetermined threshold; and determining the adjusted BiLSTM model as the trained state prediction model corresponding to the life cycle stage.

[0087] Wherein: the time domain characteristics of the third vibration signal include at least one of the following: mean; variance; root mean square amplitude; root mean square value; maximum value; minimum value; waveform index; peak index; pulse index; edge index; skewness; kurtosis.

[0088] FIG3 is an exemplary structural diagram of a BiLSTM model according to an embodiment of the present invention.

[0089] In Figure 3, the BiLSTM model comprises an input layer 41, an output layer 42, and a hidden layer between the input and output layers 41 and 42. The BiLSTM's hidden layer is composed of two LSTMs: one processes the input sequence in the forward direction, and the other processes the sequence in the reverse direction. After processing, the outputs of the two LSTMs are concatenated. The final BiLSTM output is obtained only after all time steps have been calculated. Input layer 41 receives training data 30 and its labels 31. Training data 30 includes multiple vibration signals x1, x2, ..., xi in a time-sequential manner. Specifically, these can be the third vibration signal in a specific lifecycle phase (the third vibration signal is typically a time series). The labels x1, x2, ..., xi correspond to the machine tool states x1, x2, ..., xi at the same predetermined time (which can be represented by the vibration signal sequence after the predetermined time). After the training data 30 and its labels are input into the hidden layer, the output layer 42 outputs the predicted machine tool state after the predetermined time (which can also be represented by the predicted vibration signal sequence y1, y2, ..., yi after the predetermined time). Then, the loss function between the predicted state of the tool and the label 31 is calculated, and the BiLSTM model is reversely updated based on the loss function to complete the training, thereby obtaining a state prediction model corresponding to the specific life cycle stage.

[0090] In one embodiment, the method includes: after the state prediction model outputs the predicted state of the CNC machine tool, recording the actual state corresponding to the predicted state; determining the difference between the predicted state and the actual state; and updating the trained state prediction model based on the difference.

[0091] FIG4 is a schematic diagram of updating a trained state prediction model according to an embodiment of the present invention.

[0092] In Figure 4, it is assumed that the life cycle phase of the vibration signal between time points t1 and t3 is the normal period. In the prediction process 50, the vibration signals (x1-x3) between time points t1 and t3 are input into the trained state prediction model 40 corresponding to the normal period. The state prediction model 40 outputs the predicted state (represented by vibration signals y1-y3) after a predetermined time n (i.e., tn+1-tn+3). The actual state (represented by vibration signals a1-a3) during the arrival time period tn+1-tn+3 is then recorded. In the excitation process 60, the difference between the actual states a1-a3 and the predicted states y1-y3 is calculated to determine the excitation values ​​r1-r3. The closer the actual state is to the predicted state, the positive excitation value is used; otherwise, the farther the actual state is from the predicted state, the positive penalty value is used. The trained state prediction model 40 is then updated using the excitation values ​​r1-r3.

[0093] FIG5 is a schematic diagram of the overall process of state prediction and model maintenance of a CNC machine tool according to an embodiment of the present invention.

[0094] As shown in Figure 5, signal processing 70 is first performed to determine the RMS value of the vibration signal during the current operation of the CNC machine tool. Next, in state prediction processing 71, the vibration signal during the current operation is input into a state prediction model corresponding to the life cycle stage determined based on the RMS value to predict the machine tool state. Then, in excitation determination processing 72, an excitation value is determined based on the difference between the predicted value of the state prediction model and the recorded actual state. In model evaluation processing 73, an assessment is made as to whether the state prediction model needs to be updated. If an update is determined to be necessary, the excitation value is used to update the state prediction model in adaptive optimization processing 74.

[0095] FIG6 is an exemplary structural diagram of a state prediction device for a CNC machine tool according to an embodiment of the present invention. The state prediction device 600 for a CNC machine tool includes:

[0096] An acquisition module 601 is configured to acquire a first vibration signal of a CNC machine tool during operation;

[0097] A determination module 602 is configured to determine a life cycle stage of the operation process in the entire life cycle of the CNC machine tool based on the first vibration signal;

[0098] The input module 603 is configured to input the first vibration signal into a trained state prediction model corresponding to the life cycle stage, so that the state prediction model outputs a predicted state of the CNC machine tool.

[0099] In one embodiment, the determination module 602 is used to input the first vibration signal into a trained life cycle stage classification model so that the life cycle stage classification model outputs the life cycle stage; wherein the training process of the life cycle stage classification model includes: obtaining a second vibration signal of a test CNC machine tool of the same model as the CNC machine tool throughout its life cycle; determining the time domain characteristics of the second vibration signal; determining the normal period, degradation period and failure period in the entire life cycle based on the time domain characteristics of the second vibration signal; adding a first label to the time domain characteristics in the normal period, adding a second label to the time domain characteristics in the degradation period and adding a third label to the time domain characteristics in the failure period; and training the neural network model as a life cycle stage classification model based on the first label, the second label and the third label.

[0100] In one embodiment, the time domain feature is a root mean square value; the determination module 602 is used to determine a time segment in which the root mean square value is less than or equal to a first predetermined threshold as a normal period; determine a time segment in which the root mean square value is greater than the first predetermined threshold and less than or equal to a second predetermined threshold as a degradation period; and determine a time segment in which the root mean square value is greater than the second predetermined threshold as a failure period; wherein the second predetermined threshold is greater than the first predetermined threshold.

[0101] In one embodiment, it includes: a training module 604, which is used to execute a training process of a state prediction model corresponding to a life cycle stage, the training process including: obtaining a third vibration signal of a test CNC machine tool of the same model as the CNC machine tool in the life cycle stage and a subsequent state of the machine tool after the third vibration signal; inputting the third vibration signal and the subsequent state into a bidirectional long short-term memory model, so that the BiLSTM model sequentially executes: determining the time domain characteristics of the third vibration signal; obtaining a forward hidden state through a forward time cycle layer based on the time domain characteristics of the third vibration signal; obtaining a reverse hidden state through a reverse time cycle layer based on the time domain characteristics of the third vibration signal; splicing the forward hidden state and the reverse hidden state to obtain a hidden layer state; determining a predicted state of the CNC machine tool based on the hidden layer state; determining a loss function value based on the difference between the predicted state and the subsequent state; adjusting the model parameters of the BiLSTM model so that the loss function value is less than a predetermined threshold; and determining the adjusted BiLSTM model as a trained state prediction model corresponding to the life cycle stage.

[0102] The embodiment of the present invention also proposes an electronic device with a processor-memory architecture. Figure 7 is a structural diagram of an electronic device according to an embodiment of the present invention. As shown in Figure 7, the electronic device 700 includes a processor 701, a memory 702, and a computer program stored on the memory 702 and runnable on the processor 701. When the computer program is executed by the processor 701, it implements any of the above-mentioned state prediction methods for CNC machine tools. Among them, the memory 702 can be specifically implemented as a variety of storage media such as an electrically erasable programmable read-only memory (EEPROM), a flash memory (Flash memory), and a programmable read-only memory (PROM). The processor 701 can be implemented to include one or more central processing units or one or more field programmable gate arrays, wherein the field programmable gate array integrates one or more central processing unit cores. Specifically, the central processing unit or the central processing unit core can be implemented as a CPU, an MCU, or a DSP, and so on.

[0103] It should be noted that not all steps and modules in the above processes and structure diagrams are required, and certain steps or modules can be omitted based on actual needs. The execution order of the steps is not fixed and can be adjusted as needed. The division of the modules is merely for the convenience of describing the functional division adopted. In actual implementation, a module can be implemented by multiple modules, and the functions of multiple modules can be implemented by the same module. These modules can be located in the same device or in different devices.

[0104] The hardware modules in each embodiment can be implemented mechanically or electronically. For example, a hardware module may include a specially designed permanent circuit or logic device (such as a dedicated processor, such as an FPGA or ASIC) for performing a specific operation. The hardware module may also include a programmable logic device or circuit (such as a general-purpose processor or other programmable processor) temporarily configured by software to perform a specific operation. As for whether to implement the hardware module mechanically, or using a dedicated permanent circuit, or using a temporarily configured circuit (such as configured by software), it can be decided based on cost and time considerations.

[0105] The above are only preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting the state of a numerically controlled machine tool, characterized in that: include: Acquiring (101) a first vibration signal of a numerical control machine tool during operation; Based on the first vibration signal, determining (102) a life cycle stage of the operation process in the entire life cycle of the numerical control machine tool; The first vibration signal is input (103) into a trained state prediction model corresponding to the life cycle stage, so that the state prediction model outputs a predicted state of the numerical control machine tool.

2. The method according to claim 1, characterized in that: The determining (102) of the life cycle stage of the operation process in the entire life cycle of the CNC machine tool based on the first vibration signal includes: The first vibration signal is input into a trained life cycle stage classification model so that the life cycle stage classification model outputs the life cycle stage.

3. The method according to claim 2, characterized in that The training process of the life cycle stage classification model includes: Acquire a second vibration signal of a test CNC machine tool of the same model as the CNC machine tool during its entire life cycle; determining a time domain feature of the second vibration signal; Determining a normal period, a degradation period, and a failure period in the whole life cycle based on the time domain characteristics of the second vibration signal; Adding a first label to the time domain features in the normal period, a second label to the time domain features in the degradation period, and a third label to the time domain features in the failure period; Based on the first label, the second label and the third label, the neural network model is trained as the life cycle stage classification model.

4. The method according to claim 3, characterized in that The time domain characteristics include a root mean square value; The determining of the normal period, degradation period, and failure period in the whole life cycle based on the time domain characteristics of the second vibration signal includes: Determine a time period in which the root mean square value is less than or equal to a first predetermined threshold as the normal period; Determine a time period in which the root mean square value is greater than the first predetermined threshold and less than or equal to the second predetermined threshold as the degradation period; Determine the time period during which the root mean square value is greater than the second predetermined threshold as the expiration period; The second predetermined threshold is greater than the first predetermined threshold.

5. The method according to claim 1, characterized in that: The determining (102) of the life cycle stage of the operation process in the entire life cycle of the CNC machine tool based on the first vibration signal includes: determining a root mean square value of the first vibration signal; Based on the root mean square value of the first vibration signal, the life cycle stage of the operation process in the entire life cycle of the CNC machine tool is determined; when the root mean square value is less than or equal to a first predetermined threshold, the life cycle stage is determined to be a normal period; when the root mean square value is greater than the first predetermined threshold and less than or equal to a predetermined second predetermined threshold, the life cycle stage is determined to be a degradation period; when the root mean square value is greater than the second predetermined threshold for a period of time, the life cycle stage is determined to be a failure period; wherein the second predetermined threshold is greater than the first predetermined threshold.

6. The method according to claim 1, characterized in that The training process of the state prediction model corresponding to the life cycle stage includes: Acquire a third vibration signal of a test CNC machine tool of the same model as the CNC machine tool in the life cycle stage and a subsequent state of the machine tool after the third vibration signal; The third vibration signal and the subsequent state are input into a bidirectional long short-term memory model, so that the bidirectional long short-term memory model sequentially performs: determining the time domain characteristics of the third vibration signal; obtaining a forward hidden state through a forward time cycle layer based on the time domain characteristics of the third vibration signal; obtaining a reverse hidden state through a reverse time cycle layer based on the time domain characteristics of the third vibration signal; splicing the forward hidden state and the reverse hidden state to obtain a hidden layer state; determining a predicted state of the CNC machine tool based on the hidden layer state; determining a loss function value based on a difference between the predicted state and the subsequent state; adjusting model parameters of the bidirectional long short-term memory model so that the loss function value is less than a predetermined threshold; The adjusted bidirectional long short-term memory model is determined as a trained state prediction model corresponding to the life cycle stage.

7. The method according to claim 6, characterized in that include: After the state prediction model outputs the predicted state of the numerically controlled machine tool, recording the actual state corresponding to the predicted state; determining a difference between the predicted state and the actual state; Based on the difference, the trained state prediction model is updated.

8. A state prediction device for a numerically controlled machine tool, characterized in that: include: An acquisition module (601) is used to acquire a first vibration signal of a numerical control machine tool during operation; A determination module (602) is used to determine, based on the first vibration signal, whether the operation process is in the entirety of the CNC machine tool. life cycle phases in the life cycle; An input module (603) is used to input the first vibration signal into a trained state prediction model corresponding to the life cycle stage, so that the state prediction model outputs the predicted state of the CNC machine tool.

9. The device according to claim 8, characterized in that The determination module (602) is used to input the first vibration signal into a trained life cycle stage classification model so that the life cycle stage classification model outputs the life cycle stage; wherein the training process of the life cycle stage classification model includes: obtaining a second vibration signal of a test CNC machine tool of the same model as the CNC machine tool in its entire life cycle; determining the time domain characteristics of the second vibration signal; determining the normal period, degradation period and failure period in the entire life cycle based on the time domain characteristics of the second vibration signal; adding a first label to the time domain characteristics in the normal period, adding a second label to the time domain characteristics in the degradation period and adding a third label to the time domain characteristics in the failure period; and training the neural network model as the life cycle stage classification model based on the first label, the second label and the third label.

10. The device according to claim 9, characterized in that The time domain feature is a root mean square value; The determination module (602) is used to determine the time period in which the root mean square value is less than or equal to the first predetermined threshold as the normal period; A time period in which the root mean square value is greater than the first predetermined threshold and less than or equal to the second predetermined threshold is determined as the degradation period; a time period in which the root mean square value is greater than the second predetermined threshold is determined as the failure period; wherein the second predetermined threshold is greater than the first predetermined threshold.

11. The device according to claim 8, characterized in that include: A training module (604) is used to execute a training process of a state prediction model corresponding to the life cycle stage, the training process comprising: obtaining a third vibration signal of a test CNC machine tool of the same model as the CNC machine tool in the life cycle stage and a subsequent state of the machine tool after the third vibration signal; inputting the third vibration signal and the subsequent state into a bidirectional long short-term memory model, so that the bidirectional long short-term memory model sequentially executes: determining a time domain feature of the third vibration signal; obtaining a forward hidden state through a forward time cycle layer based on the time domain feature of the third vibration signal; obtaining a reverse hidden state through a reverse time cycle layer based on the time domain feature of the third vibration signal; concatenating the forward hidden state and the reverse hidden state to obtain a hidden layer state; determining a predicted state of the CNC machine tool based on the hidden layer state; determining a loss function value based on a difference between the predicted state and the subsequent state; adjusting model parameters of the bidirectional long short-term memory model so that the loss function value is less than a predetermined threshold; and determining the adjusted bidirectional long short-term memory model as a trained state prediction model corresponding to the life cycle stage.

12. An electronic device, characterized in that: include: Processor (701); A memory (702), configured to store executable instructions of the processor (701); The processor (701) is used to read the executable instructions from the memory (702) and execute the executable instructions to implement the state prediction method of a numerically controlled machine tool according to any one of claims 1 to 7.

13. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the state prediction method of a numerically controlled machine tool according to any one of claims 1 to 7 is implemented.

14. A computer program product, characterized in that The method comprises a computer program, which, when executed by a processor, implements the state prediction method of a numerically controlled machine tool according to any one of claims 1 to 7.

Images