Method and device for predicting residual life of film evaporator for lyocell fiber production

Abstract

The invention discloses a residual life prediction method and device for a film evaporator for lyocell fiber production, and relates to the technical field of mechanical fault diagnosis, and the method comprises the steps: a vibration sensor in a preset area on the surface of a barrel of the film evaporator collects vibration data, carries out the preprocessing, and calculates the feature frequency of a key part; adding the vibration data vectors of the single vibration sensors, and constructing a one-dimensional convolutional neural network model to obtain a characteristic curve; fitting by adopting a multi-term high-order exponential function, and establishing a residual life characteristic curve; the data set is obtained, a one-dimensional convolutional neural network model for residual life prediction is established, the data set is input into the one-dimensional convolutional neural network model, and a residual life prediction model is obtained; vibration data collected in real time and the characteristic frequency of the key component are processed and then are used as the input of the prediction model; according to the method, the complexity of a feature extraction network model can be reduced, the real-time performance of model operation is improved, data labels are provided more accurately, and the accuracy of residual life prediction can be effectively improved.

Description

Technical Field

[0001] The invention relates to the technical field of mechanical fault diagnosis, and in particular to a method and a device for predicting the remaining life of a thin film evaporator used in the production of lyocell fibers. Background Art

[0002] As a highly efficient evaporation device, thin film evaporator is widely used in the field of spinning, mainly for the concentration of spinning solution, recovery of solvent and treatment of heat-sensitive materials. Its working principle is to form a uniform liquid film on the heating surface through the rotating scraper, and evaporate efficiently under vacuum conditions.

[0003] However, various faults are inevitable in the long-term operation of thin film evaporators. Among them, the main shaft of large thin film evaporators has a large axial dimension and is sensitive to vibration. The main shaft is connected to a complex structure, and the system structural mode is complex. The vibration of the main shaft can easily excite the system structural mode, affecting the normal operation of the equipment or even damaging the equipment. At the same time, spindle locking / stuck is a common fault in the operation of large thin film evaporators, and excessive vibration displacement of the main shaft is an important cause of this fault. The causes of abnormal vibration of the main shaft include imperfect equipment design, easy resonance during operation, damage to the main shaft bearing, unbalanced main shaft wear, and main shaft bending failure. These faults will lead to a decrease in the evaporation efficiency of the thin film evaporator, unstable product quality, and even equipment shutdown, seriously affecting the continuity and economy of spinning production.

[0004] Traditional fault diagnosis methods mainly rely on manual inspections and experience-based judgments. This method has poor real-time performance and cannot detect potential fault hazards in a timely manner. It also has low accuracy and relies on personal experience, which is prone to misjudgment and omissions. It also lacks predictive power and cannot predict the development trend of faults, making it difficult to perform preventive maintenance.

[0005] On May 15, 2024, a Chinese patent disclosed a fault diagnosis method for mechanical equipment, with the announcement number CN118194136B. It collects various types of information such as vibration time domain information, acoustic time domain information and image information of the mechanical equipment to be diagnosed in the current operating state; and outputs vibration frequency domain waveform diagram, acoustic frequency domain waveform diagram and image information; performs noise reduction processing; determines the network model and its transfer function; trains the network model based on historical data and completes network model re-learning; inputs the vibration frequency domain waveform diagram, acoustic frequency domain waveform diagram and image information after noise reduction processing into the optimized network model; integrates the vibration time domain information fault type, acoustic time domain information fault type, and image information fault type, calculates the correlation between each fault type and the fault classification results of historical data, obtains the fault type of the mechanical equipment according to the correlation, and uses the measures of observation, smell, questioning and palpation to integrate and judge the fault type of the mechanical equipment, thereby greatly improving the accuracy of fault diagnosis.

[0006] This method of using fault type as diagnosis cannot make real-time predictions of equipment failures. In addition, the solution uses interactive fusion of different data types and time data information for fault diagnosis. In scenarios where equipment failures are predicted by predicting the remaining life of the equipment, it is difficult to respond quickly, especially for remaining life prediction scenarios with high real-time requirements. At the same time, when the solution extracts features, the connection between the features and the life labels is not strong, making it difficult to accurately construct a life curve. In addition, the key structural characteristics of the equipment are not fully extracted and integrated, which affects the prediction accuracy of important equipment.

[0007] Therefore, it is urgent to conduct real-time monitoring of the life cycle of the thin film evaporator during use. By real-time collection and analysis of the operating data of the thin film evaporator, a remaining life diagnosis model can be established to achieve early warning of faults and predictive maintenance, thereby ensuring the effective operation of the thin film evaporator and improving the efficiency of spinning production and product quality. Summary of the invention

[0008] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method and device for predicting the remaining life of a thin film evaporator for lyocell fiber production, which can monitor the thin film evaporator in real time and predict the remaining life, accurately and timely perform preventive maintenance, and ensure the effective operation of the thin film evaporator.

[0009] In order to solve the above technical problems, the present invention adopts the following technical solutions: The present invention provides a method for predicting the remaining life of a thin film evaporator for producing lyocell fibers, the method comprising the following steps: S1. Obtaining vibration data and spindle speed data of the entire process from the normal operation of the thin film evaporator to the failure of the spindle and / or the spindle bearing causing the thin film evaporator to fail to operate normally; The vibration data includes acceleration signals in three directions collected by vibration sensors at the upper and lower ends and the middle area of ​​the thin film evaporator cylinder; wherein the vibration sensors at the upper and lower ends monitor the fault conditions of the upper and lower end bearings of the main shaft respectively, and the vibration sensor in the middle area of ​​the cylinder monitors the fault condition of the main shaft; S2. Perform time domain analysis and frequency domain analysis on the multiple data segments obtained after segmenting the acceleration signal at set time intervals; wherein the time domain analysis is used to extract the root mean square and kurtosis of each data segment, and the frequency domain analysis is used to convert each data segment from the time domain to the frequency domain and extract the frequency center of gravity; obtain the fault characteristic frequency of the spindle according to the spindle speed data, and obtain the fault characteristic frequency of the spindle bearing according to the spindle speed data and the spindle bearing size parameters; S3. A one-dimensional convolutional neural network model including convolution, pooling, activation function and full connection layer is constructed for each of the multiple vibration vector data obtained by vector addition of the acceleration signals in three directions of each vibration sensor; the response trend of the vibration response of each vibration vector data changing with time is extracted through the one-dimensional convolutional neural network model of each vibration vector data, and the vibration vector data characteristic curves of the upper and lower end bearings and the main shaft are obtained; S4, using multiple high-order exponential functions to fit the vibration vector data characteristic curves of the upper and lower end bearings and the main shaft obtained in step S3 to form a remaining life characteristic curve of the upper end bearing of the main shaft, a remaining life characteristic curve of the lower end bearing of the main shaft and a remaining life characteristic curve of the main shaft, and then express them in percentage form; S5, taking the time interval set in step S2 as the time resolution, the root mean square, kurtosis, frequency centroid, fault characteristic frequency of the main shaft bearing and fault characteristic frequency of the main shaft obtained in step S2, and the data on the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft and the remaining life characteristic curve of the main shaft obtained in step S4 together constitute a data set, wherein the data on the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft and the remaining life characteristic curve of the main shaft are label data; S6, dividing the data set obtained in step S5 into a training set, a validation set, and a test set; after training, validation, and testing, a remaining life prediction model with accuracy meeting the requirements is obtained; S7. Acquire real-time vibration data and spindle speed of the target thin film evaporator; use the processed root mean square, kurtosis, frequency center of gravity, fault characteristic frequency of the spindle, and fault characteristic frequency of the spindle bearing as inputs of the remaining life prediction model to predict the remaining life of the spindle of the target thin film evaporator and / or the remaining life of the spindle bearing.

[0010] Further, in S2, the root mean square and kurtosis of each data segment are extracted by performing time domain analysis on each data segment; Specifically, the root mean square of each data segment is extracted through time domain analysis, and its calculation formula is: ; Where: is the RMS of the data segment; is the number of samples in the data segment; is the 𝑖th sampling value in the data segment; The kurtosis of each data segment is extracted through time domain analysis, and the calculation formula is: ; Where: is the kurtosis of the data segment; is the average value of this data segment; is the 𝑖th sample value in the data segment.

[0011] Furthermore, in S2, each data segment is converted from the time domain to the frequency domain using frequency domain analysis, and then the frequency center of gravity is extracted; Specifically, each data segment is converted from the time domain to the frequency domain, and the calculation formula is: ; Where: is the transformed frequency domain value of the data segment; is the number of discrete sampling points of the data segment; is the sequence number of the time domain discrete value of the data segment; is the nth data sample of the data segment; is an imaginary unit; The calculation formula for frequency centroid extraction is: ; Where: is the frequency center of gravity of the data segment; is the frequency value of the kth spectral line; is the total number of spectral lines.

[0012] Furthermore, the set time interval is 0.5s.

[0013] Furthermore, in S2, the fault characteristic frequency of the bearing includes: Inner race fault characteristic frequency: ; Outer race fault characteristic frequency: ; Rolling element failure characteristic frequency: ; Cage failure characteristic frequency: ; Where: is the number of rolling elements; is the diameter of the rolling element; is the bearing pitch diameter; is the contact angle; The rotation frequency of the spindle is obtained according to the spindle speed.

[0014] Furthermore, the fault characteristic frequency of the main shaft is: ; Where: is the frequency multiple; is the rotation frequency of the spindle, obtained according to the spindle speed; is the main shaft rotation disturbance frequency.

[0015] Furthermore, in S2, the root mean square, kurtosis and frequency centroid of each acceleration signal, as well as the fault characteristic frequency of the bearing and the fault characteristic frequency of the main shaft are normalized and then output.

[0016] Further, in S3, the one-dimensional convolutional neural network model includes: A one-dimensional convolutional neural network model including convolution, pooling, activation function and fully connected layer is built for each vibration vector data: Input sequence: ; Where: is the input value of the tth position in the sequence; Convolution operation: ; Where: is the output value of the tth position after the convolution operation; is the size of the convolution kernel; is the weight parameter of the i-th position in the convolution kernel; The corresponding elements in the input sequence that participate in the convolution calculation; is the bias term of the convolution operation; Using RELU activation function: ; Where: is the output value of the t-th position after activation; Using the maximum pooling operation: ; Where: is the output value of the tth pooling window; is the size of the pooling window; is the value of the i-th position in the activation function output; Fully connected layer: ; Where: is the value of the jth output node; is the weight parameter connecting the i-th input node to the j-th output node; is the output value of the i-th node in the hidden layer; is the bias term of the jth output node; Using the cross entropy loss function: ; Where: is the cross entropy loss value; is the total number of categories; Category in the real label The value of The model predicts The probability value of each category.

[0017] Furthermore, a remaining life prediction model is constructed based on the one-dimensional convolutional neural network model, and the real-time vibration data and spindle speed obtained in step S7 are processed in the manner of step S2 to obtain the root mean square, kurtosis, frequency center of gravity, fault characteristic frequency of the spindle, and fault characteristic frequency of the spindle bearing, which are input into the remaining life prediction model in real time.

[0018] Furthermore, in S3, the vibration vector data characteristic curve of the main shaft is obtained by taking the average value of the vibration vector data characteristic curves corresponding to the vibration sensors in the middle area of ​​the cylinder.

[0019] Furthermore, in S4, the formula expressed in percentage form by combining the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft and the remaining life characteristic curve expression of the main shaft is as follows: ; Where: , , , , is an adjustable parameter; is the system time of vibration data; The time limit for the equipment to go from fully normal to abnormal is defined; is the remaining life percentage; Is the natural base.

[0020] The present invention also provides a device for predicting the remaining life of a thin film evaporator for producing lyocell fibers, the device comprising: A data acquisition module is used to obtain real-time vibration data collected by vibration sensors at the upper and lower ends and the middle area of ​​the cylinder of the target thin film evaporator, and at the same time obtain the spindle speed from the central control platform of the target thin film evaporator; The data preprocessing module is used to segment the real-time vibration data acquired by the data acquisition module according to the set time interval to obtain multiple vibration data segments of the set time interval; perform time domain analysis and frequency domain analysis on each data segment to obtain the root mean square, kurtosis and frequency center of gravity of each vibration data segment; obtain the fault characteristic frequency of the spindle according to the obtained spindle speed data, and then obtain the fault characteristic frequency of the spindle bearing according to the spindle speed data and the size parameters of the spindle bearing; the data output by the data preprocessing module are the root mean square, kurtosis, frequency center of gravity of the acceleration signal of each vibration sensor in three directions, the fault characteristic frequency of the spindle bearing and the fault characteristic frequency of the spindle; the root mean square, kurtosis and frequency center of gravity of the acceleration signal of each vibration sensor in three directions, as well as the fault characteristic frequency of the bearing and the fault characteristic frequency of the spindle are normalized and then output; A remaining life prediction module is encapsulated with a remaining life prediction model in the remaining life prediction method of a thin film evaporator for producing lyocell fibers, and the data output by the data preprocessing module is used as input data of the remaining life prediction model to predict the remaining life of the target thin film evaporator. The specifically predicted remaining life includes the remaining life of the main shaft and / or the remaining life of the main shaft bearing.

[0021] In summary, the present invention has the following beneficial effects: (1) This scheme uses multi-component characteristic frequency data and big data feature extraction data to form a data matrix, and combines the vibration data collected from the upper and lower end bearings and the main shaft in the middle area of ​​the thin film evaporator cylinder with the characteristic frequencies of the key components of the main shaft and bearings. In practical applications, it can more accurately capture equipment degradation information, make the correlation between characteristic data and life labels higher, and more effectively construct the relationship model between fault characteristics and life labels, thereby forming a more accurate remaining life prediction model.

[0022] (2) This application collects vibration data from the upper and lower ends of the thin film evaporator cylinder and the middle area of ​​the cylinder within a single life cycle from normal to abnormal, and designs a remaining life characteristic curve by coupling and driving multiple high-order exponential function empirical mathematical models with big data feature extraction, thereby reducing the complexity of the feature extraction network model and improving the real-time performance of the model operation. While meeting the real-time life prediction needs, it also provides more accurate data labels, which can effectively improve the accuracy of the remaining life prediction.

[0023] (3) The average value of the four characteristic curves about the main shaft in the middle area of ​​the thin film evaporator is taken to obtain the characteristic curve of the remaining life of the main shaft. The characteristic curves of the upper bearing, lower bearing and main shaft of the thin film evaporator that independently reflect the health status are constructed to simplify the calculation process.

[0024] (4) Integrate the multi-component feature frequency data in the historical data with the big data feature extraction data and the remaining life characteristic curve data to form a three-dimensional training data set, train the remaining life prediction model based on deep learning model design, train the remaining life prediction model based on the data set, establish the association mapping between the feature data and the life label, improve the prediction accuracy, and thus obtain a remaining life prediction model with accuracy that meets the requirements.

[0025] (5) Lyocell thin film evaporator is the main equipment for Lyocell production. Large thin film evaporators are huge in size and sensitive to vibration during operation. If a failure occurs, it will have a huge impact on production. In addition, it is very inconvenient to repair large thin film evaporators, especially the immediate repair after a failure. Therefore, it is very important to conduct real-time monitoring and remaining life prediction of large thin film evaporators. Therefore, a remaining life prediction model is established based on the thin film evaporator. The combination of multi-component characteristic frequency and big data feature extraction is used to monitor the vulnerable parts of the thin film evaporator in real time and output the life prediction, which is conducive to improving the fault tolerance rate and facilitating effective production management. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is the layout diagram of the vibration sensor measurement points on the thin film evaporator cylinder.

[0027] Figure 2 Figure 2 is a flow chart of the data preprocessing module.

[0028] Figure 3 This is a flow chart of the remaining life prediction module.

[0029] Figure 4 Schematic diagram of the design process for remaining life prediction and remaining life label data.

[0030] Figure 5 Schematic diagram of the process flow for designing a remaining life prediction model.

[0031] Figure 6 It is a vibration vector data diagram.

[0032] Figure 7 It is the vibration vector data characteristic curve. DETAILED DESCRIPTION

[0033] The present invention is further described in detail below in conjunction with embodiments.

[0034] Embodiment 1: The remaining life prediction method of the thin film evaporator used in the production of lyocell fiber specifically comprises the following steps: S1. Obtain vibration data of the entire process from normal operation of the thin film evaporator to failure of the spindle and / or spindle bearing causing the thin film evaporator to fail to operate normally, and obtain spindle speed data from the central control platform; The vibration data includes acceleration signals in three directions collected by vibration sensors at the upper and lower ends and the middle area of ​​the thin film evaporator cylinder; wherein the vibration sensors at the upper and lower ends monitor the fault conditions of the upper and lower end bearings of the main shaft respectively, and the vibration sensor in the middle area of ​​the cylinder monitors the fault condition of the main shaft; S2. Perform time domain analysis and frequency domain analysis on the multiple data segments obtained after segmenting the acceleration signal at set time intervals; wherein the time domain analysis is used to extract the root mean square and kurtosis of each data segment, and the frequency domain analysis is used to convert each data segment from the time domain to the frequency domain and extract the frequency center of gravity; obtain the fault characteristic frequency of the spindle according to the spindle speed data, and obtain the fault characteristic frequency of the spindle bearing according to the spindle speed data and the spindle bearing size parameters; During implementation, the time interval is set to 0.5s.

[0035] In S2, the root mean square and kurtosis of each data segment are extracted by performing time domain analysis on each data segment; specifically, The root mean square of each data segment is extracted through time domain analysis, and its calculation formula is: ; Where: is the RMS of the data segment; is the number of samples in the data segment; is the 𝑖th sample value in the data segment.

[0036] The kurtosis of each data segment is extracted through time domain analysis, and the calculation formula is: ; Where: is the kurtosis of the data segment; is the average value of this data segment; is the 𝑖th sample value in the data segment.

[0037] In S2, frequency domain analysis is used to convert each data segment from the time domain to the frequency domain, and then the frequency center of gravity is extracted; Specifically, each data segment is converted from the time domain to the frequency domain, and the calculation formula is: ; Where: is the transformed frequency domain value of the data segment; is the number of discrete sampling points of the data segment; is the sequence number of the time domain discrete value of the data segment; is the nth data sample of the data segment; Is an imaginary unit.

[0038] The calculation formula for frequency centroid extraction is: ; Where: is the frequency center of gravity of the data segment; is the frequency value of the kth spectral line; is the total number of spectral lines.

[0039] In S2, the fault characteristic frequencies of the bearing include: Inner race fault characteristic frequency: ; Outer race fault characteristic frequency: ; Rolling element failure characteristic frequency: ; Cage failure characteristic frequency: ; Where: is the number of rolling elements; is the diameter of the rolling element; is the bearing pitch diameter; is the contact angle; The rotation frequency of the spindle is obtained according to the spindle speed.

[0040] The fault characteristic frequency of the spindle is: ; Where: is the frequency multiple; is the rotation frequency of the spindle, obtained according to the spindle speed; is the main shaft rotation disturbance frequency.

[0041] In S2, the root mean square, kurtosis and frequency centroid of each acceleration signal, as well as the fault characteristic frequency of the bearing and the fault characteristic frequency of the main shaft are normalized and then output.

[0042] The vibration data normalization process of this embodiment adopts the most conventional method, and the calculation method is: the difference between the size of the current data and the minimum value of the data is divided by the difference between the maximum value and the minimum value of the data, to obtain a number between 0 and 1.

[0043] The obtained root mean square, kurtosis, frequency centroid, fault characteristic frequency of the main shaft and the fault characteristic frequency of the main shaft bearing are normalized and used as inputs to the remaining life prediction model to predict the remaining life of the target thin film evaporator. The specific predicted remaining life includes the remaining life of the main shaft and / or the remaining life of the main shaft bearing.

[0044] S3. A one-dimensional convolutional neural network model including convolution, pooling, activation function and full connection layer is constructed for each of the multiple vibration vector data obtained by vector addition of the acceleration signals in the three directions of each vibration sensor; the response trend of the vibration response of each vibration vector data changing with time is extracted through the one-dimensional convolutional neural network model of each vibration vector data, and the vibration vector data characteristic curves of the upper and lower end bearings and the main shaft are obtained; the model is used to capture the changing trend of the vibration response of the vibration vector data over time. This trend reflects the characteristic data of the equipment life degradation, which is presented as a time-extended trend. Figure 7 A curve is shown. Figure 7 In the figure, the horizontal axis is time, and the data on the curve reflects the relationship between vibration data and life at a certain point in time.

[0045] In S3, the one-dimensional convolutional neural network model includes: For each vibration vector data, a one-dimensional convolutional neural network model including convolution, pooling, activation function and fully connected layer is built: Input sequence: ; The input layer receives vibration vector data, where: is the input value of the tth position in the sequence;

[0046] Convolution operation: ; Where: is the output value of the tth position after the convolution operation; is the size of the convolution kernel; is the weight parameter of the i-th position in the convolution kernel; The corresponding elements in the input sequence that participate in the convolution calculation; is the bias term for the convolution operation.

[0047] Using RELU activation function: ; Where: is the output value of the t-th position after activation.

[0048] Using the maximum pooling operation: ; Where: is the output value of the tth pooling window; is the size of the pooling window; is the value of the i-th position in the activation function output.

[0049] Fully connected layer: ; Where: is the value of the jth output node; is the weight parameter connecting the i-th input node to the j-th output node; is the output value of the i-th node in the hidden layer; is the bias term of the jth output node.

[0050] Using the cross entropy loss function: ; Where: is the cross entropy loss value; is the total number of categories; Category in the real label The value of The model predicts The probability value of each category.

[0051] The output of the one-dimensional convolutional neural network model changes over time. Figure 7 The vibration vector data shown shows the changing trend of vibration response over time.

[0052] S4, using multiple high-order exponential functions to fit the vibration vector data characteristic curves of the upper and lower end bearings and the main shaft obtained in step S3 to form a remaining life characteristic curve of the upper end bearing of the main shaft, a remaining life characteristic curve of the lower end bearing of the main shaft and a remaining life characteristic curve of the main shaft, and then express them in percentage form; In S4, the vibration vector data characteristic curve of the main shaft is obtained by taking the average value of the vibration vector data characteristic curves corresponding to the vibration sensors in the middle area of ​​the cylinder.

[0053] In S4, the formula expressed in percentage form by combining the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft and the remaining life characteristic curve expression of the main shaft is as follows: ; Where: , , , , It is an adjustable parameter. The specific debugging process is as follows: first set the initial value to 1, further observe the fit between the vibration vector data characteristic curve and the remaining life characteristic curve obtained after fitting, and then adjust , , , , until the vibration vector data characteristic curve is well matched with the remaining life characteristic curve obtained after fitting; wherein, is the system time of vibration data; The time limit for the equipment to go from fully normal to abnormal is defined; is the remaining life percentage; Is the natural base.

[0054] S5, taking the time interval set in step S2 as the time resolution, the root mean square, kurtosis, frequency centroid, fault characteristic frequency of the spindle bearing and the fault characteristic frequency of the spindle obtained in step S2, and the data on the remaining life characteristic curve of the upper end bearing of the spindle, the remaining life characteristic curve of the lower end bearing of the spindle and the remaining life characteristic curve of the spindle obtained in step S4, these data together constitute a data set, wherein the data on the remaining life characteristic curve of the upper end bearing of the spindle, the remaining life characteristic curve of the lower end bearing of the spindle and the remaining life characteristic curve of the spindle are label data; In practical applications, for ease of understanding, the root mean square, kurtosis, frequency centroid, fault characteristic frequency of the spindle bearing and the fault characteristic frequency of the spindle obtained in step S2, as well as the remaining life characteristic curve of the upper end bearing of the spindle, the remaining life characteristic curve of the lower end bearing of the spindle and the remaining life characteristic curve of the spindle obtained in step S4 are plotted on the same two-dimensional coordinate. Each moment on the two-dimensional coordinate corresponds to a series of data, which together constitute the data set.

[0055] S6. Establish a remaining life prediction model, divide the data set obtained in step S5 into a training set, a validation set, and a test set; according to the deep learning model training process, use the divided data set to train, verify, and test the remaining life prediction model. Take the ratio of 8:1:1 as an example to divide the training set, validation set, and test set, and finally obtain a remaining life prediction model with accuracy that meets the requirements.

[0056] In S6, a remaining life prediction model is constructed based on the one-dimensional convolutional neural network model, and the root mean square, kurtosis, frequency center of gravity, fault characteristic frequency of the spindle and fault characteristic frequency of the spindle bearing obtained by processing the real-time vibration data and spindle speed obtained in step S7 in the manner of step S2 are input into the remaining life prediction model in real time.

[0057] In the one-dimensional convolutional neural network model, the input sequence is: ; ; in, represents the remaining life corresponding to the t-th data segment calculated according to the remaining life characteristic curve; , , , , , , , They correspond to the normalized root mean square, kurtosis, frequency centroid, inner ring fault characteristic frequency, outer ring fault characteristic frequency, rolling element fault characteristic frequency, cage fault characteristic frequency, and spindle fault characteristic frequency corresponding to the t-th data segment.

[0058] S7. Acquire real-time vibration data and spindle speed of the target thin film evaporator; use the processed root mean square, kurtosis, frequency center of gravity, fault characteristic frequency of the spindle, and fault characteristic frequency of the spindle bearing as inputs of the remaining life prediction model to predict the remaining life of the spindle of the target thin film evaporator and / or the remaining life of the spindle bearing.

[0059] Since the mechanical system will vibrate during operation, the vibration will change when a component fails, and the vibration response will be different for different fault levels. Therefore, the time of component failure or even failure can be predicted by analyzing the characteristic changes of the vibration response, that is, life prediction can be achieved. The root mean square, kurtosis and frequency center of gravity, the fault characteristic frequency of the main shaft and the fault characteristic frequency of the main shaft bearing reflect the characteristics of the vibration response, and these characteristics can be used to describe the vibration situation. The remaining life prediction model is established through a deep learning model. The establishment process uses the characteristic data of the vibration response of the equipment parts throughout the life cycle to train the deep learning model. The deep learning model learns the characteristics of the component degradation process through training with a large amount of data, thereby realizing the prediction of component life.

[0060] Embodiment 2: This embodiment provides a method for establishing a remaining life prediction model for a large thin film evaporator. When implemented, the real-time vibration data includes 1 vibration sensor respectively set at the upper and lower ends of the thin film evaporator cylinder and 4 vibration sensors in four areas of the cylinder, a total of 18 acceleration signals in three directions collected by 6 vibration sensors, where the three directions refer to the x-axis, y-axis and z-axis directions; wherein the vibration sensor at the upper end of the cylinder is used to monitor the fault condition of the upper end bearing of the main shaft, the vibration sensor at the lower end of the cylinder is used to monitor the fault condition of the lower end bearing of the main shaft, and the 4 vibration sensors in four areas of the cylinder are used to monitor the fault condition of the main shaft. The 18 acquired acceleration signals are segmented according to the set time intervals to obtain multiple data segments of the set time intervals; each data segment is analyzed in the time domain and the frequency domain respectively to obtain the root mean square, kurtosis and frequency center of gravity of each vibration data segment; the fault characteristic frequency of the main shaft is obtained according to the acquired spindle speed data, and then the fault characteristic frequency of the main shaft bearing is obtained according to the spindle speed data and the size parameters of the spindle bearing.

[0061] The method for establishing the remaining life prediction model includes: S1. Obtain vibration data and spindle speed data of the entire process from the normal operation of the thin film evaporator to the failure of the spindle and the spindle bearing causing the thin film evaporator to fail to operate normally.

[0062] The vibration data includes one vibration sensor respectively arranged at the upper and lower ends of the thin film evaporator cylinder and four vibration sensors in four areas of the cylinder, a total of 18 acceleration signals in three directions collected by six vibration sensors; among them, the vibration sensor at the upper end of the cylinder is used to monitor the fault condition of the upper end bearing of the main shaft, the vibration sensor at the lower end of the cylinder is used to monitor the fault condition of the lower end bearing of the main shaft, and the four vibration sensors in four areas of the cylinder are used to monitor the fault condition of the main shaft.

[0063] S2. Segment the 18 acceleration signals obtained in step S1 according to set time intervals to obtain multiple data segments with set time intervals; perform time domain analysis and frequency domain analysis on each data segment to obtain the root mean square, kurtosis and frequency center of gravity of each vibration data segment; obtain the fault characteristic frequency of the spindle according to the spindle speed data obtained in step S1, and then obtain the fault characteristic frequency of the spindle bearing according to the spindle speed data and the size parameters of the spindle bearing.

[0064] S3. Perform vector addition on the acceleration signals in three directions of each vibration sensor of the vibration data obtained in step S1 to obtain 6 vibration vector data, and build a one-dimensional convolutional neural network model including convolution, pooling, activation function and fully connected layer for each vibration vector data; extract the response trend of the vibration response of each vibration vector data over time through the one-dimensional convolutional neural network model of each vibration vector data.

[0065] S4. Use multiple high-order exponential functions to fit the vibration vector data characteristic curves of the upper and lower end bearings and the main shaft obtained in step S3 to form the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft and the remaining life characteristic curve of the main shaft. After fitting, the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft and the remaining life characteristic curve of the main shaft are expressed in percentage form.

[0066] Taking the time interval set in step S2 as the time resolution and the response time of the vibration data as the horizontal coordinate, the root mean square, kurtosis, frequency centroid, fault characteristic frequency of the spindle bearing and the fault characteristic frequency of the spindle obtained in step S2, and the data on the remaining life characteristic curve of the upper end bearing of the spindle, the remaining life characteristic curve of the lower end bearing of the spindle and the remaining life characteristic curve of the spindle obtained in step S4, these data together constitute the data set, among which the data on the remaining life characteristic curve of the upper end bearing of the spindle, the remaining life characteristic curve of the lower end bearing of the spindle and the remaining life characteristic curve of the spindle are label data.

[0067] For ease of understanding, in step S5, the time interval set in step S2 is used as the time resolution, the response time of the vibration data is used as the horizontal coordinate, and the root mean square, kurtosis, frequency center of gravity, fault characteristic frequency of the main shaft bearing and fault characteristic frequency of the main shaft obtained in step S2, as well as the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft and the remaining life characteristic curve of the main shaft obtained in step S4 are plotted on the same two-dimensional coordinate. Each moment on the two-dimensional coordinate corresponds to a series of data, which together constitute a data set, wherein the data on the remaining characteristic curve is label data.

[0068] S6. Establish a deep learning model for remaining life prediction, and divide the data set obtained in step S5 into training set, validation set and test set in a ratio of 8:1:1; according to the deep learning model training process, use the divided data set to train, verify and test the remaining life prediction deep learning model, and finally obtain a remaining life prediction model with accuracy that meets the requirements.

[0069] After the remaining life prediction model is established, the vibration data collected by 6 vibration sensors at the upper and lower ends and four areas of the target thin film evaporator cylinder are obtained, and the spindle speed is obtained from the central control platform of the target thin film evaporator. The target thin film evaporator is the thin film evaporator for which the remaining life prediction is required.

[0070] The acquired vibration data is segmented according to the set time interval to obtain multiple vibration data segments with set time intervals; time domain analysis and frequency domain analysis are performed on each data segment to obtain the root mean square, kurtosis and frequency centroid of each vibration data segment; the fault characteristic frequency of the spindle is obtained according to the spindle speed data, and the fault characteristic frequency of the spindle bearing is obtained according to the spindle speed data and the size parameters of the spindle bearing; the root mean square, kurtosis and frequency centroid of 18 acceleration signals in three directions of 6 vibration sensors, as well as the fault characteristic frequency of the bearing and the fault characteristic frequency of the shaft are normalized and output.

[0071] The root mean square, kurtosis and frequency center of gravity of 18 acceleration signals of 6 vibration sensors in three directions of the target thin film evaporator, as well as the fault characteristic frequency of the main shaft and the fault characteristic frequency of the main shaft bearing of the target thin film evaporator are taken as input and input into the remaining life prediction model established by the method for establishing the remaining life prediction model of the thin film evaporator in steps S1 to S6 of this embodiment, and the remaining life of the target thin film evaporator is predicted. The specific predicted remaining life includes the remaining life of the main shaft and / or the remaining life of the main shaft bearing.

[0072] like Figure 1 As shown, multiple vibration sensors distributed in preset areas on the surface of the target thin film evaporator cylinder collect vibration data of the upper and lower bearings of the main shaft and the vibration data of the main shaft in real time.

[0073] Figure 1 The six sensors are respectively vibration sensor 1 for collecting vibration data of the upper bearing in the upper end area of ​​the thin film evaporator, vibration sensor 6 for collecting vibration data of the lower bearing in the lower end area, and vibration sensors 2, 3, 4 and 5 for collecting vibration data of the main shaft in the middle area. In implementation, preferably, vibration sensor 1 is installed on the cylinder in the area corresponding to the upper bearing and the upper mounting plate, and vibration sensor 2 is installed on the cylinder in the area corresponding to the lower bearing and the lower mounting plate. The vibration data includes 18 acceleration signals in three directions collected by a total of six vibration sensors at the upper and lower ends of the cylinder and in four areas of the cylinder of the thin film evaporator; among them, the vibration sensor at the upper end of the cylinder is used to monitor the fault condition of the upper end bearing of the main shaft, the vibration sensor at the lower end of the cylinder is used to monitor the fault condition of the lower end bearing of the main shaft, and the four vibration sensors in the four areas of the cylinder are used to monitor the fault condition of the main shaft.

[0074] After preprocessing the vibration data, the root mean square, kurtosis and frequency center of gravity are obtained, and the fault characteristic frequency of the bearing and the fault characteristic frequency of the spindle are calculated. After normalizing these data, they are input into the remaining life prediction model established by the method for establishing the remaining life prediction model in real time. The remaining life prediction model predicts the spindle and the spindle bearing, and outputs the remaining life of the spindle and the spindle bearing.

[0075] The vibration data normalization process of this embodiment adopts the most conventional method, and the calculation method is: the difference between the size of the current data and the minimum value of the data is divided by the difference between the maximum value and the minimum value of the data, to obtain a number between 0 and 1.

[0076] Remaining life prediction module process diagram Figure 3 As shown, the module takes the output of data preprocessing as input, and then predicts the remaining life of the target thin film evaporator. The specific predicted remaining life includes the remaining life of the main shaft and the remaining life of the main shaft bearing.

[0077] The remaining life prediction module adopts a knowledge- and data-driven remaining life prediction method, and is designed based on a deep learning network. Figure 4 As shown in the figure, the design method is as follows: First, design the label data set. First, in multiple tests, the data acquisition module is used to obtain 18 acceleration signals in three directions from six vibration sensors during the whole process from the equipment running intact to the failure of the main shaft and the main shaft bearing, which causes the equipment to fail to operate normally. Then, the vibration acceleration in three directions of each sensor is vector-added to obtain 6 vibration vector data. The schematic diagram of one vibration vector data is shown in the figure below. Figure 6As shown in FIG. 1 , the amplitude represents the magnitude of the vibration acceleration collected by the vibration sensor, and the sample point represents the number of sampling points of the vibration acceleration collected by the vibration sensor.

[0078] Then, a one-dimensional convolutional neural network model including convolution, pooling, activation function and fully connected layer is built for each vibration vector data: A one-dimensional convolutional neural network model including convolution, pooling, activation function and fully connected layer is built for each vibration vector data: Input sequence: ; Where: is the input value of the tth position in the sequence; Convolution operation: ; Where: is the output value of the tth position after the convolution operation; is the size of the convolution kernel; is the weight parameter of the i-th position in the convolution kernel; The corresponding elements in the input sequence that participate in the convolution calculation; is the bias term of the convolution operation; Using RELU activation function: ; Where: is the output value of the t-th position after activation; Using the maximum pooling operation: ; Where: is the output value of the tth pooling window; is the size of the pooling window; is the value of the i-th position in the activation function output; Fully connected layer: ; Where: is the value of the jth output node; is the weight parameter connecting the i-th input node to the j-th output node; is the output value of the i-th node in the hidden layer; is the bias term of the jth output node; Using the cross entropy loss function: ; Where: is the cross entropy loss value; is the total number of categories; Category in the real label The value of The model predicts The probability value of each category.

[0079] The one-dimensional convolutional neural network model is used to extract the response trend of the vibration response over time, and finally the vibration vector data characteristic curve formed by the response trend of the vibration response of the upper and lower end bearings and the main shaft of the six main shafts over time is obtained. Figure 7 This is a schematic diagram of one of the vibration vector data characteristic curves. Among them, the amplitude represents the magnitude of the vibration acceleration collected by the vibration sensor, and the sample point represents the number of sampling points of the vibration acceleration collected by the vibration sensor. The obtained vibration vector data characteristic curve has obvious characteristics. As time goes by, the damage of the spindle and spindle bearings gradually accumulates, and quickly collapses when it accumulates to a certain level. This phenomenon is the same as the damage phenomenon of mechanical parts in real situations. Therefore, this curve can be used as a reference for estimating the remaining life of the spindle and spindle bearings. And the remaining life of the spindle bearings.

[0080] Among the six vibration vector data characteristic curves, two vibration vector data characteristic curves are derived from the data of the upper and lower sensors of the thin film evaporator, which respectively reflect the health status of the upper and lower main shaft bearings. Another vibration vector data characteristic curve is derived from the data of the four vibration sensors on the cylinder of the thin film evaporator, which reflects the health status of the main shaft. For the convenience of calculation, the data of these four vibration vector data characteristic curves are averaged to obtain a vibration vector data characteristic curve of the main shaft. Finally, a total of three vibration vector data characteristic curves reflecting the health status of the main shaft and main shaft bearings of the thin film evaporator will be obtained, namely, one vibration vector data characteristic curve of the upper end bearing of the main shaft, one vibration vector data characteristic curve of the lower end bearing of the main shaft, and one vibration vector data characteristic curve of the main shaft.

[0081] However, the obtained vibration vector data characteristic curve fluctuates significantly, which is inconsistent with the destruction caused by the accumulation of component damage. In reality, component damage gradually accumulates over time and presents a changing trend, while the current curve has obvious fluctuations, which does not match the gradual process of physical laws. This is due to the presence of many interferences in the measured vibration signal. In view of this, the remaining life characteristic curve is designed by combining empirical mathematical models with big data feature extraction. Multiple high-order exponential functions are used to fit the curve. At the same time, the mathematical model expressed in percentage form by combining the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft, and the remaining life characteristic curve expression of the main shaft is as follows: ; Where: , , , , is an adjustable parameter; is the horizontal coordinate of the vibration data, i.e., the system time; The time limit for the equipment to go from fully normal to abnormal is defined; is the remaining life percentage; is the natural base. So far, with a time resolution of 0.5s and the response time of the vibration data as the horizontal coordinate, the root mean square, kurtosis, frequency center of gravity, bearing fault characteristic frequency, spindle fault characteristic frequency and remaining life characteristic curve data of the vibration data are plotted on the same two-dimensional coordinate. Each moment in the two-dimensional coordinate corresponds to a series of data, which together constitute the data set, among which the data on the remaining life characteristic curve is the label data.

[0082] We further designed a remaining life prediction model based on a deep learning model. The design process is as follows: Figure 5 . 80% of the data set is used as the training set, 10% as the validation set, and the remaining 10% as the test set. Then, a convolutional deep learning network model for remaining life prediction is established. According to the conventional deep learning model development process, the remaining life prediction deep learning model is trained, verified, and tested using the divided data set, and finally a remaining life prediction model with accuracy that meets the requirements is obtained.

[0083] According to the above method, the remaining life prediction module design proposed in this application can be realized. In practical applications, the vibration data of the preset area on the target thin film evaporator cylinder is collected in real time by the vibration sensor, the vibration data collected by the vibration sensor is preprocessed, and the characteristic frequency of the key components of the thin film evaporator is calculated.

[0084] The preprocessing includes data segmentation of the vibration data, wherein the data segmentation adopts the principle of time segmentation, and the continuously collected vibration data is divided into multiple data segments according to the time interval. The data segmentation divides the continuously collected vibration data into multiple data segments according to the time interval of 0.5 seconds.

[0085] The preprocessing also includes time domain analysis; The time domain analysis completes the extraction of the root mean square of the vibration data of each data segment, and the calculation formula is: ; Where: is the RMS of the data segment; is the number of samples in the data segment; is the 𝑖th sampling value in the data segment; The time domain analysis completes the extraction of the kurtosis of the vibration data of each data segment, and the calculation formula is: ; Where: is the kurtosis of the data segment; is the average value of this data segment; is the 𝑖th sample value in the data segment.

[0086] Preprocessing also includes frequency domain analysis; Frequency domain analysis completes the conversion of the vibration data of each data segment from the time domain to the frequency domain. The calculation formula is: ; Where: is the transformed frequency domain value of the data segment; is the number of discrete sampling points of the data segment; is the sequence number of the time domain discrete value of the data segment; is the nth data sample of the data segment; is an imaginary unit; Frequency domain analysis also includes frequency centroid extraction, which is calculated as follows: ; Where: is the frequency center of gravity of the data segment; is the frequency value of the kth spectral line; is the total number of spectral lines.

[0087] Finally, the root mean square, kurtosis, and frequency centroid obtained from each vibration data are arranged in order and output.

[0088] The key characteristic frequency calculation includes the fault characteristic frequency of the bearing, and the fault characteristic frequency of the bearing includes: Inner race fault characteristic frequency: ; Outer race fault characteristic frequency: ; Rolling element failure characteristic frequency: ; Cage failure characteristic frequency: ; Where: is the number of rolling elements; is the diameter of the rolling element; is the bearing pitch diameter; is the contact angle; is the rotation frequency of the shaft, obtained according to the spindle speed.

[0089] The calculation of the characteristic frequency of key components also includes the fault characteristic frequency of the main shaft, and the fault characteristic frequency model of the main shaft is: ; Where: is the frequency multiple; The rotation frequency of the main shaft is obtained according to the main shaft speed. is the main shaft rotation disturbance frequency.

[0090] The remaining life prediction module will input the root mean square, kurtosis, frequency center of gravity, bearing fault characteristic frequency, and main shaft fault characteristic frequency of the normalized vibration response data output from the data preprocessing module in real time, and output the remaining life of the key vulnerable parts of the target thin film evaporator, thereby evaluating the remaining life of the entire equipment.

[0091] Embodiment 3: This embodiment provides a device for predicting the remaining life of a thin film evaporator for producing lyocell fibers, the device comprising: The data acquisition module is used to obtain real-time vibration data collected by 6 vibration sensors, including 1 measuring point at the upper end, 1 measuring point at the lower end and 4 measuring points in the middle of the target thin film evaporator cylinder, and simultaneously obtain the spindle speed from the central control platform of the target thin film evaporator.

[0092] The data preprocessing module is used to segment the vibration data acquired by the data acquisition module according to the set time interval to obtain multiple vibration data segments with set time intervals; perform time domain analysis and frequency domain analysis on each data segment to obtain the root mean square, kurtosis and frequency center of gravity of each vibration data segment; obtain the fault characteristic frequency of the spindle according to the obtained spindle speed data, and obtain the fault characteristic frequency of the spindle bearing according to the spindle speed data and the size parameters of the spindle bearing; the data output by the final data preprocessing module are the root mean square, kurtosis, frequency center of gravity of 18 acceleration signals in three directions of 6 vibration sensors, the fault characteristic frequency of the spindle bearing and the fault characteristic frequency of the spindle; the root mean square, kurtosis and frequency center of gravity of the 18 acceleration signals in three directions of the 6 vibration sensors, as well as the fault characteristic frequency of the bearing and the fault characteristic frequency of the spindle are normalized and then output.

[0093] A remaining life prediction module encapsulates the remaining life prediction model in Example 1 or Example 2, uses the data output by the data preprocessing module as input data of the remaining life prediction model, and predicts the remaining life of the target thin film evaporator. The specifically predicted remaining life includes the remaining life of the main shaft and / or the remaining life of the main shaft bearing.

[0094] This embodiment is also provided with a status display and alarm module, which is configured as follows: Visualize real-time remaining life prediction value; If the remaining life of the spindle or bearing is lower than the preset threshold, an audible and visual alarm is triggered and an alarm message is pushed. For example, in implementation, an alarm is issued when the remaining life of the target thin film evaporator and its wearing parts is lower than the set remaining life threshold of 5%.

[0095] The vibration data consists of 18 acceleration signals from three directions of six vibration sensors. The vibration data is normalized using the most conventional method, which is calculated as follows: the difference between the current data size and the minimum value of the data, divided by the difference between the maximum value and the minimum value of the data, to obtain a number between 0 and 1.

[0096] Remaining life prediction module process diagram Figure 3 As shown, the module takes the output of data preprocessing as input, and then predicts the remaining life of the target thin film evaporator. The specific predicted remaining life includes the remaining life of the main shaft and the remaining life of the main shaft bearing.

[0097] The remaining life prediction module adopts a knowledge- and data-driven remaining life prediction method, and is designed based on a deep learning network. Figure 4 As shown, the design method is as follows: Design a label data set. First, in multiple experiments, obtain 18 acceleration signals in three directions from six vibration sensors during the entire process from intact operation to failure of the spindle and spindle bearing, resulting in the inability of the equipment to operate normally. Then, the vibration accelerations in three directions of each sensor are vector-added to obtain 6 vibration vector data. Then, for each vibration vector data, a one-dimensional convolutional neural network model containing convolution, pooling, activation function and fully connected layer is built: Input sequence: ; Where: is the input value of the tth position in the sequence; Convolution operation: ; Where: is the output value of the tth position after the convolution operation; is the size of the convolution kernel; is the weight parameter of the i-th position in the convolution kernel; The corresponding elements in the input sequence that participate in the convolution calculation; is the bias term of the convolution operation; Using RELU activation function: ; Where: is the output value of the t-th position after activation; Using the maximum pooling operation: ; Where: is the output value of the tth pooling window; is the size of the pooling window; is the value of the i-th position in the activation function output; Fully connected layer: ; Where: is the value of the jth output node; is the weight parameter connecting the i-th input node to the j-th output node; is the output value of the i-th node in the hidden layer; is the bias term of the jth output node; Using the cross entropy loss function: ; Where: is the cross entropy loss value; is the total number of categories; Category in the real label The value of The model predicts The probability value of each category.

[0098] The model is used to extract the response trend of vibration response over time, and eventually six vibration vector data characteristic curves will be obtained. Figure 7 The figure is a schematic diagram of a vibration vector data characteristic curve. The obtained vibration vector data characteristic curve has obvious characteristics. As time goes by, the damage of the spindle and the spindle bearing gradually accumulates, and when it accumulates to a certain level, it collapses rapidly. This phenomenon is the same as the damage phenomenon of mechanical parts in reality. Therefore, the vibration vector data characteristic curve can be used as a reference for estimating the remaining life of the spindle and the spindle bearing.

[0099] Two of the six vibration vector data characteristic curves are derived from the data of the upper and lower sensors of the thin film evaporator, which respectively reflect the health status of the upper and lower main shaft bearings. The remaining four vibration vector data characteristic curves are derived from the data of the four vibration sensors in the middle area of ​​the thin film evaporator cylinder, which reflect the health status of the main shaft. For the convenience of calculation, the data of these four vibration vector data characteristic curves are averaged to obtain a vibration vector data characteristic curve of the main shaft. Finally, a total of three vibration vector data characteristic curves reflecting the health status of the main shaft and main shaft bearings of the thin film evaporator will be obtained, namely, one vibration vector data characteristic curve of the upper end bearing of the main shaft, one vibration vector data characteristic curve of the lower end bearing of the main shaft, and one vibration vector data characteristic curve of the main shaft. Vibration vector data characteristic curve of the main shaft.

[0100] However, the obtained vibration vector data characteristic curve still has defects, such as obvious fluctuations, which are inconsistent with the damage caused by the accumulation of component damage. This is due to the presence of many interferences in the measured vibration signal. In view of this, the remaining life characteristic curve is designed by combining empirical mathematical models with big data feature extraction. Multiple high-order exponential functions are used to fit the curve. At the same time, the mathematical model of the remaining life characteristic curve expressed in percentage form in combination with the vibration vector data characteristic curve expression is as follows: ; Where: , , , , is an adjustable parameter; is the system time of vibration data; The time limit for the equipment to go from fully normal to abnormal is defined; is the remaining life percentage; is the natural base. So far, with a time resolution of 0.5s and the response time of the vibration data as the horizontal coordinate, the root mean square, kurtosis, frequency center of gravity, bearing fault characteristic frequency, spindle fault characteristic frequency and remaining life characteristic curve data of the vibration data are plotted on the same two-dimensional coordinate. Each moment in the two-dimensional coordinate corresponds to a series of data, which together constitute the data set, among which the data on the remaining life characteristic curve is the label data.

[0101] We further designed a remaining life prediction model based on a deep learning model. The design process is as follows: Figure 5 . 80% of the data set is used as the training set, 10% as the validation set, and the remaining 10% as the test set. Then, a one-dimensional convolutional deep learning network model for remaining life prediction is established. According to the conventional deep learning model development process, the remaining life prediction deep learning model is trained, verified, and tested using the divided data set, and finally a remaining life prediction model with accuracy that meets the requirements is obtained.

[0102] According to the above method, the remaining life prediction module design proposed in this application can be realized. In practical applications, the vibration data of the preset area on the target thin film evaporator cylinder is collected in real time by the vibration sensor, the vibration data collected by the vibration sensor is preprocessed, and the characteristic frequency of the key components of the thin film evaporator is calculated.

[0103] The preprocessing includes data segmentation of the vibration data, wherein the data segmentation adopts the principle of time segmentation, and the continuously collected vibration data is divided into multiple data segments according to the time interval. The data segmentation divides the continuously collected vibration data into multiple data segments according to the time interval of 0.5 seconds.

[0104] The preprocessing also includes time domain analysis; The time domain analysis completes the extraction of the root mean square of the vibration data of each data segment, and the calculation formula is: ;

[0105] Where: is the RMS of the data segment; is the number of samples in the data segment; is the 𝑖th sampling value in the data segment; The time domain analysis completes the extraction of the kurtosis of the vibration data of each data segment, and the calculation formula is: ; Where: is the kurtosis of the data segment; is the average value of this data segment; is the 𝑖th sample value in the data segment.

[0106] Preprocessing also includes frequency domain analysis; Frequency domain analysis completes the conversion of the vibration data of each data segment from the time domain to the frequency domain. The calculation formula is: ; Where: is the transformed frequency domain value of the data segment; is the number of discrete sampling points of the data segment; is the sequence number of the time domain discrete value of the data segment; is the nth data sample of the data segment; is an imaginary unit; Frequency domain analysis also includes frequency centroid extraction, which is calculated as follows: ; Where: is the frequency center of gravity of the data segment; is the frequency value of the kth spectral line; is the total number of spectral lines.

[0107] Finally, the root mean square, kurtosis, and frequency centroid obtained from each vibration data are arranged in order and output.

[0108] The key characteristic frequency calculation includes the fault characteristic frequency of the bearing, and the fault characteristic frequency of the bearing includes: Inner race fault characteristic frequency: ; Outer race fault characteristic frequency: ; Rolling element failure characteristic frequency: ; Cage failure characteristic frequency: ; Where: is the number of rolling elements; is the diameter of the rolling element; is the bearing pitch diameter; is the contact angle; is the rotation frequency of the shaft, obtained according to the spindle speed.

[0109] The calculation of the characteristic frequency of key components also includes the fault characteristic frequency of the main shaft, and the fault characteristic frequency model of the main shaft is: ; Where: is the frequency multiple; is the rotation frequency of the spindle, obtained according to the spindle speed; is the disturbance frequency.

[0110] The root mean square, kurtosis and frequency centroid of each acceleration signal, as well as the fault characteristic frequency of the bearing and the fault characteristic frequency of the main shaft are integrated into a data matrix, and the root mean square, kurtosis and frequency centroid of each acceleration signal, the fault characteristic frequency of the bearing and the fault characteristic frequency of the main shaft in the data matrix are normalized and then output.

[0111] The remaining life prediction module will input the root mean square, kurtosis, frequency center of gravity, bearing fault characteristic frequency, and main shaft fault characteristic frequency of each normalized acceleration signal output from the data preprocessing module in real time, and output the remaining life of the key vulnerable parts of the target thin film evaporator, thereby evaluating the remaining life of the entire equipment.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting the remaining life of a thin film evaporator for producing lyocell fibers, characterized in that: The following steps are involved: S1. Obtaining vibration data and spindle speed data of the entire process from the normal operation of the thin film evaporator to the failure of the spindle and / or the spindle bearing causing the thin film evaporator to fail to operate normally; The vibration data includes acceleration signals in three directions collected by vibration sensors at the upper and lower ends and the middle area of ​​the thin film evaporator cylinder; wherein the vibration sensors at the upper and lower ends monitor the fault conditions of the upper and lower end bearings of the main shaft respectively, and the vibration sensor in the middle area of ​​the cylinder monitors the fault condition of the main shaft; S2. Perform time domain analysis and frequency domain analysis on the multiple data segments obtained after segmenting the acceleration signal at set time intervals; wherein the time domain analysis is used to extract the root mean square and kurtosis of each data segment, and the frequency domain analysis is used to convert each data segment from the time domain to the frequency domain and extract the frequency center of gravity; obtain the fault characteristic frequency of the spindle according to the spindle speed data, and obtain the fault characteristic frequency of the spindle bearing according to the spindle speed data and the spindle bearing size parameters; S3. A one-dimensional convolutional neural network model including convolution, pooling, activation function and full connection layer is constructed for each of the multiple vibration vector data obtained by vector addition of the acceleration signals in three directions of each vibration sensor; the response trend of the vibration response of each vibration vector data changing with time is extracted through the one-dimensional convolutional neural network model of each vibration vector data, and the vibration vector data characteristic curves of the upper and lower end bearings and the main shaft are obtained; S4, using multiple high-order exponential functions to fit the vibration vector data characteristic curves of the upper and lower end bearings and the main shaft obtained in step S3 to form a remaining life characteristic curve of the upper end bearing of the main shaft, a remaining life characteristic curve of the lower end bearing of the main shaft and a remaining life characteristic curve of the main shaft, and then express them in percentage form; S5, taking the time interval set in step S2 as the time resolution, the root mean square, kurtosis, frequency centroid, fault characteristic frequency of the main shaft bearing and fault characteristic frequency of the main shaft obtained in step S2, and the data on the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft and the remaining life characteristic curve of the main shaft obtained in step S4 together constitute a data set, wherein the data on the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft and the remaining life characteristic curve of the main shaft are label data; S6, dividing the data set obtained in step S5 into a training set, a validation set, and a test set; after training, validation, and testing, a remaining life prediction model with accuracy meeting the requirements is obtained; S7. Acquire real-time vibration data and spindle speed of the target thin film evaporator; use the processed root mean square, kurtosis, frequency center of gravity, fault characteristic frequency of the spindle, and fault characteristic frequency of the spindle bearing as inputs of the remaining life prediction model to predict the remaining life of the thin film evaporator spindle and / or the remaining life of the spindle bearing.

2. The method for predicting the remaining life of a thin film evaporator for producing lyocell fibers according to claim 1, characterized in that: In S2, the root mean square and kurtosis of each data segment are extracted by performing time domain analysis on each data segment; Specifically, the root mean square of each data segment is extracted through time domain analysis, and its calculation formula is: ； Where: is the RMS of the data segment; is the number of samples in the data segment; is the 𝑖th sampling value in the data segment; The kurtosis of each data segment is extracted through time domain analysis, and the calculation formula is: ； Where: is the kurtosis of the data segment; is the average value of this data segment; is the 𝑖th sample value in the data segment.

3. The method for predicting the remaining life of a thin film evaporator for producing lyocell fibers according to claim 2, characterized in that: In S2, frequency domain analysis is used to convert each data segment from the time domain to the frequency domain, and then the frequency center of gravity is extracted; Specifically, each data segment is converted from the time domain to the frequency domain, and the calculation formula is: ； Where: is the transformed frequency domain value of the data segment; is the number of discrete sampling points of the data segment; is the sequence number of the time domain discrete value of the data segment; is the nth data sample of the data segment; is an imaginary unit; The calculation formula for frequency centroid extraction is: ; Where: is the frequency center of gravity of the data segment; is the frequency value of the kth spectral line; is the total number of spectral lines.

4. The method for predicting the remaining life of a thin film evaporator for lyocell fiber production according to any one of claims 1 to 3, characterized in that: The set time interval is 0.5s.

5. The method for predicting the remaining life of a thin film evaporator for producing lyocell fibers according to claim 3, characterized in that: In S2, the fault characteristic frequencies of the bearing include: Inner race fault characteristic frequency: ； Outer race fault characteristic frequency: ； Rolling element failure characteristic frequency: ； Cage failure characteristic frequency: ； Where: is the number of rolling elements; is the diameter of the rolling element; is the bearing pitch diameter; is the contact angle; The rotation frequency of the spindle is obtained according to the spindle speed.

6. The method for predicting the remaining life of a thin film evaporator for producing lyocell fibers according to claim 5, characterized in that: The fault characteristic frequency of the spindle is: ； Where: is the frequency multiple; is the rotation frequency of the spindle, obtained according to the spindle speed; is the main shaft rotation disturbance frequency.

7. The method for predicting the remaining life of a thin film evaporator for producing lyocell fibers according to claim 6, characterized in that: In S2, the root mean square, kurtosis and frequency centroid of each acceleration signal, as well as the fault characteristic frequency of the bearing and the fault characteristic frequency of the main shaft are normalized and then output.

8. The method for predicting the remaining life of a thin film evaporator for producing lyocell fibers according to claim 1 or 7, characterized in that: In S3, the one-dimensional convolutional neural network model includes: A one-dimensional convolutional neural network model including convolution, pooling, activation function and fully connected layer is built for each vibration vector data: Input sequence: ; Where: is the input value at the tth position in the sequence; Convolution operation: ; Where: is the output value of the tth position after the convolution operation; is the size of the convolution kernel; is the weight parameter of the i-th position in the convolution kernel; The corresponding elements in the input sequence that participate in the convolution calculation; is the bias term of the convolution operation; Using RELU activation function: ; Where: is the output value of the t-th position after activation; Using the maximum pooling operation: ; Where: is the output value of the tth pooling window; is the size of the pooling window; is the value of the i-th position in the activation function output; Fully connected layer: ; Where: is the value of the jth output node; is the weight parameter connecting the i-th input node to the j-th output node; is the output value of the i-th node in the hidden layer; is the bias term of the jth output node; Using the cross entropy loss function: ； Where: is the cross entropy loss value; is the total number of categories; Category in the real label The value of The model predicts The probability value of each category.

9. The method for predicting the remaining life of a thin film evaporator for producing lyocell fibers according to claim 8, characterized in that: A remaining life prediction model is constructed based on the one-dimensional convolutional neural network model, and the real-time vibration data and spindle speed obtained in step S7 are processed in the manner of step S2 to obtain the root mean square, kurtosis, frequency center of gravity, fault characteristic frequency of the spindle, and fault characteristic frequency of the spindle bearing, which are input into the remaining life prediction model in real time.

10. The method for predicting the remaining life of a thin film evaporator for producing lyocell fibers according to claim 8, characterized in that: In S3, the vibration vector data characteristic curve of the main shaft is obtained by taking the average value of the vibration vector data characteristic curves corresponding to the vibration sensors in the middle area of ​​the cylinder.

11. The method for predicting the remaining life of a thin film evaporator for producing lyocell fibers according to claim 10, characterized in that: In S4, the formula expressed in percentage form by combining the remaining life characteristic curve of the upper end bearing of the main shaft, the remaining life characteristic curve of the lower end bearing of the main shaft and the remaining life characteristic curve expression of the main shaft is as follows: ； Where: , , , , is an adjustable parameter; is the system time of vibration data; The time limit for the equipment to go from fully normal to abnormal is defined; is the remaining life percentage; Is the natural base.

12. A device for predicting the remaining life of a thin film evaporator for producing lyocell fibers, characterized in that: The device includes: A data acquisition module is used to obtain real-time vibration data collected by vibration sensors at the upper and lower ends and the middle area of ​​the cylinder of the target thin film evaporator, and at the same time obtain the spindle speed from the central control platform of the target thin film evaporator; The data preprocessing module is used to segment the real-time vibration data acquired by the data acquisition module according to the set time interval to obtain multiple vibration data segments of the set time interval; perform time domain analysis and frequency domain analysis on each data segment to obtain the root mean square, kurtosis and frequency center of gravity of each vibration data segment; obtain the fault characteristic frequency of the spindle according to the obtained spindle speed data, and then obtain the fault characteristic frequency of the spindle bearing according to the spindle speed data and the size parameters of the spindle bearing; the data output by the data preprocessing module are the root mean square, kurtosis, frequency center of gravity of the acceleration signal of each vibration sensor in three directions, the fault characteristic frequency of the spindle bearing and the fault characteristic frequency of the spindle; the root mean square, kurtosis and frequency center of gravity of the acceleration signal of each vibration sensor in three directions, as well as the fault characteristic frequency of the bearing and the fault characteristic frequency of the spindle are normalized and then output; A remaining life prediction module, wherein the remaining life prediction module encapsulates a remaining life prediction model in the remaining life prediction method for a thin film evaporator for producing lyocell fibers according to any one of claims 1 to 11, and uses the data output by the data preprocessing module as input data of the remaining life prediction model to predict the remaining life of the target thin film evaporator, wherein the specifically predicted remaining life includes the remaining life of the main shaft and / or the remaining life of the main shaft bearing.

Images