An intelligent particle screening device and a filter damage monitoring method thereof

Abstract

The present application provides an intelligent particle screening device and a filter damage monitoring method thereof, which belongs to the technical field of using filter screens to screen solid particles. The device includes: two layers of filter screens, two tension and pressure sensors, a power module, a data acquisition and transmission module, a control module and an outer frame. The two layers of filter screens are respectively connected to the two tension and pressure sensors. The power module is used to drive the screening, the control module is used to control the screening process, and the data acquisition and transmission module is used to collect and transmit sensor data. The filter damage monitoring method uses a machine learning model to train the model by collecting the tension data of the two layers of filter screens under normal operation and different degrees of damage, so that it can capture the characteristics of the tension change relationship between the two layers of filter screens. In the actual screening process, the tension data of the two layers of filter screens are monitored in real time, and it is judged whether the filter screen is damaged and the degree of damage. This technical solution can sensitively and real-time monitor the status of the filter screen, reduce the losses caused by filter screen damage, and improve screening efficiency.

Description

Technical Field

[0001] The present application relates to the technical field of using filter screens to screen solid particles, and in particular to an intelligent particle screening device and a filter screen damage monitoring method thereof. Background Art

[0002] During the particle screening process, real-time monitoring of the filter status is crucial to ensure product quality.

[0003] In the prior art, the intelligence of particle screening devices is mainly focused on the screening effects of particles of different shapes, and there are few improvements in the intelligence of filter status monitoring.

[0004] In terms of filter status monitoring methods, the existing technologies provide the following methods: monitoring methods based on marker particles, weighing methods, and monitoring methods based on electromagnetic signals. However, the above methods have many shortcomings: the monitoring method based on marker particles has hysteresis and cannot quantify the degree of damage because it relies on particles passing through the damage point, especially for early minor damage. The weighing method indirectly measures the weight change of the sieved material to judge the filter status. This a posteriori method has low sensitivity and is easily interfered by other factors. Moreover, the monitoring is performed when the screening quality has already decreased. The monitoring method based on electromagnetic signals requires the screening operation to be stopped, which not only affects production efficiency, but also increases costs, and even poses a safety hazard in flammable and explosive environments.

[0005] In summary, the prior art lacks a screening device that supports real-time and accurate monitoring of the filter status, and also lacks an effective method that can monitor the filter status in real-time and accurately. Summary of the invention

[0006] The purpose of the present application is to provide an intelligent particle screening device and a filter damage monitoring method thereof to solve or alleviate the problems existing in the above-mentioned prior art.

[0007] In order to achieve the above objectives, this application provides the following technical solutions:

[0008] The present application provides an intelligent particle screening device, comprising:

[0009] Two layers of filters, two tension and pressure sensors, a power module, a data acquisition and transmission module, a control module and an outer frame;

[0010] The mesh size of the two layers of filter screens is the same and is used for screening particles;

[0011] The two layers of the filter screen are fixedly connected to the two tension and pressure sensors respectively; the other end of the tension and pressure sensor is fixed to the transmission rod, and the transmission rod is driven by the power module;

[0012] The data acquisition and transmission module is used to receive the filter tension time series data collected by the tension and pressure sensor and forward it to the background server, and at the same time receive the analysis results and control instructions returned by the background server, and send the control instructions to the control module;

[0013] The control module is used to control the working state of the intelligent particle screening device according to the control instruction.

[0014] In the above scheme, the outer frame includes an outer frame bracket, an upper frame, a middle frame and a lower frame;

[0015] The upper frame and the upper filter screen, and the middle frame and the lower filter screen are connected through soft materials, and are connected to the corresponding flow guide ports of the two filter screens.

[0016] In the above solution, the intelligent particle screening device further includes: a base;

[0017] The base is a square or rectangular frame with a storage space inside for accommodating the data acquisition and transmission module and the control module;

[0018] The base is connected to the power module via a shock-absorbing spring.

[0019] The present application also provides a filter damage monitoring method, in the model building stage:

[0020] S101, receiving the filter tension time series data in the no-load state collected by the intelligent particle screening device;

[0021] S102, based on the time series data of the tension of the filter screen under no-load condition, using the autocorrelation function to determine the period of the time series data of the tension under no-load condition P ; Wherein, the autocorrelation function is defined as:

[0022] ,

[0023] In the formula, is the lag time after the start time. is the filter tension sequence in the The value of the timing point, is the mean tension, is the length of the sequence;

[0024] S103, according to the cycle P , the Fourier series is used to fit the time series data of the filter tension under the no-load state of the equipment to obtain the fitting periodic function;

[0025] S104, subtracting the value of the fitting periodic function at the same time point from the collected time series data of the filter tension in the no-load state of the equipment to obtain a fitting error;

[0026] S105, if the fitting error is less than a preset error threshold, the fitting periodic function is used to perform periodic signal fitting reduction on the tension data collected under the working state to obtain the reduced tension time series data of the filter screen; wherein the tension time series data collected under the working state includes the tension time series data when the filter screen is intact and the tension time series data of the filter screen under different degrees of damage;

[0027] S106, normalizing the reduced filter screen tension time series data, and calculating the upper and lower tension ratios of the normalized filter screen tension time series data;

[0028] S107, inputting the normalized upper and lower filter tension time series data and the upper and lower tension ratio into the long short-term memory network for training to obtain a trained model;

[0029] S108, classifying the degree of damage of the filter into multiple categories according to the actual degree of damage of the filter, and using the multiple categories as damage degree labels;

[0030] S109, adding the damage degree label to the test data set of the long short-term memory network;

[0031] S110, using the trained model, calculating the damage degree prediction value of the test data, and calculating the optimal split point according to the damage degree prediction value of the test data and its damage degree label, and constructing a decision tree;

[0032] During the model application phase:

[0033] S111, normalizing the upper and lower filter screen tension time series data collected when the equipment is in working state and calculating the upper and lower tension ratio;

[0034] S112, loading the trained model, and inputting the normalized upper and lower filter screen tension time series data and the upper and lower tension ratio into the trained model, and calculating the damage degree prediction value in real time;

[0035] S113. According to the predicted value of the damage degree, the decision tree is used to make a real-time judgment on the damage degree to obtain a damage degree category.

[0036] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned filter damage monitoring method are implemented.

[0037] An embodiment of the present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned filter damage monitoring method when executing the computer program.

[0038] The embodiment of the present application also provides a computer program product, which includes computer executable instructions, and the computer executable instructions are used to implement the steps of the filter damage monitoring method described above when executed by a computer.

[0039] Beneficial effects:

[0040] The intelligent particle screening device provided in the present application includes: two layers of filter screens, two tension and pressure sensors, a power module, a data acquisition and transmission module, a control module and an outer frame, and on this basis provides a filter screen damage monitoring method. This technical solution has the following significant advantages:

[0041] Real-time monitoring of the working status of core components: The intelligent particle screening device innovatively integrates a double-layer filter and a tension pressure sensor, which can collect the tension data of the upper and lower filter layers in real time, and transmit the data to the control module for analysis through the data acquisition and transmission module, thus realizing real-time monitoring of the working status of the core component - the filter. Compared with the existing method that requires shutdown monitoring, this has significant real-time advantages. For example, compared with the method based on the electromagnetic signal of the filter, this embodiment does not require regular shutdown monitoring and is highly efficient.

[0042] Sensitive damage warning: The intelligent particle screening device collects the tension data of the upper and lower filter layers in real time and forwards it to the backend server, so that the filter damage monitoring method can be analyzed based on the change in the force relationship between the two layers of the filter, without having to wait for the marker ball to leak out to identify the filter damage like the method based on adding marker balls. As long as the force relationship between the two layers of the filter changes, the abnormality of the filter can be identified, including slight damage or even aging, with sensitivity, thus achieving earlier warning and more timely maintenance.

[0043] Double-layer filter screen ensures product safety: The intelligent particle screening device adopts a double-layer filter screen structure. Even if a single-layer filter screen is damaged, the other layer of filter screen can continue to work, ensuring the quality of the screened fine material, avoiding material contamination, and having a high degree of safety. Compared with the method based on weighing the screened material, when an abnormality is found in this embodiment, due to the double-layer filter screen structure, the fine material screened before the abnormality is found has no effect, thus ensuring product quality.

[0044] Efficient power system: The power module mainly drives the transmission rod and the filter. The outer frame serves as a supporting structure and is connected to the filter through soft materials. The outer frame is not subjected to force. This design can ensure that the data collected by the sensor is not disturbed. On the other hand, it improves the working efficiency of the power module, extends its service life, and reduces energy consumption and maintenance costs.

[0045] In summary, in the technical solution provided by the present invention, through the combination of the intelligent particle screening device and the filter damage monitoring method, real-time collection, monitoring, prediction and control are achieved, timely warning and rapid response of filter damage are achieved, screening efficiency is significantly improved, maintenance costs are reduced, and product quality is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a schematic diagram of the structure of an intelligent particle screening device provided according to some embodiments of the present application.

[0047] Figure 2 This is a schematic diagram of the structure of an intelligent particle screening device provided according to some embodiments of the present application.

[0048] Figure 3 A schematic flow chart of a filter damage monitoring method provided according to some embodiments of the present application.

[0049] Figure 4 A schematic diagram of a test set quantified damage degree prediction result provided according to some embodiments of the present application.

[0050] Figure 5 A schematic diagram of a test set damage degree decision tree classifier provided according to some embodiments of the present application.

[0051] Description of reference numerals:

[0052] 1-upper filter, 2-lower filter, 3-upper pull pressure sensor, 4-lower pull pressure sensor, 5-transmission rod, 6-power module, 7-spring, 8-base, 9-control module, 10-data acquisition and transmission module, 11-feeding port, 12-upper frame, 13-middle frame, 14-lower frame, 15-fine discharge port, 16-coarse discharge port, 17-outer frame bracket. DETAILED DESCRIPTION

[0053] The terms "first", "second", "third" and "fourth" etc. in the specification and claims of the present application and the drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0054] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0055] The embodiments of the present application are described below in conjunction with the accompanying drawings.

[0056] Embodiment 1:

[0057] This embodiment provides an intelligent particle screening device. Figure 1 , Figure 2 FIG. 2 shows a schematic diagram of the structure of the intelligent particle screening device. Figure 1 , Figure 2 As shown, the device comprises:

[0058] Two layers of filter screens, two tension and pressure sensors, a power module 6, a data acquisition and transmission module 10, a control module 9 and an outer frame;

[0059] The mesh size of the two layers of filter screens is the same and is used for screening particles;

[0060] The two layers of the filter screen are fixedly connected to the two tension and pressure sensors respectively; the other end of the tension and pressure sensor is fixed to the transmission rod 5, and the transmission rod 5 is driven by the power module 6;

[0061] The data acquisition and transmission module 10 is used to receive the filter screen tension time series data collected by the tension and pressure sensor and forward it to the backend server, and at the same time receive the analysis results and control instructions returned by the backend server, and send the control instructions to the control module 9;

[0062] The control module 9 is used to control the working state of the intelligent particle screening device according to the control instruction.

[0063] In the technical solution of this embodiment, the particle screening device refers to a device used to separate particulate matter of different particle sizes, which can be used in the fields of industry, mining, chemical industry, pharmaceuticals, etc. The material can be classified into different levels according to the particle size through the filter system, thereby realizing the classification and grading of the material.

[0064] In this embodiment, the filter system of the intelligent particle screening device includes two layers of filter screens (upper filter screen 1 and lower filter screen 2), and the mesh numbers of the two layers of filter screens are the same. Each layer of filter screen is fixed to the tension pressure sensor through a bracket, and is used to screen particles. The particles passing through the two layers of filter screens are fine particles, and the particles that do not pass through the two layers of filter screens are coarse particles. The design of two layers of filter screens with the same mesh number is adopted. The upper filter screen 1 is first screened once, and the lower filter screen 2 is further screened. Each layer can perform the same mesh number screening operation, which can not only ensure the effective separation of coarse and fine particles, improve the accuracy and precision of screening, and ensure that the fineness of the final particles meets the requirements, but also reduce the misjudgment or uneven distribution caused by insufficient screening of one layer of filter screen during the screening process. Moreover, when a single layer of filter screen is damaged, the other layer of filter screen continues to work effectively, reducing the situation of incomplete or missed screening caused by the single layer of filter screen, and improving the reliability and consistency of the screening process. In addition, by collecting the tension time series data of the two layers of filter screens in real time, a solid data foundation is provided for the monitoring of the filter screen status.

[0065] In this embodiment, the vibration of the filter is driven by the power module 6, and the tension data of the filter is monitored in real time through two tension and pressure sensors, and these data are forwarded to the background server, so that the operating status during the screening process can be monitored in real time, and the tension changes of the filter can be captured, so as to accurately and real-time obtain the working status of the filter.

[0066] Among them, the two tension pressure sensors are upper tension pressure sensor 3 and lower tension pressure sensor 4, respectively, which are connected between the upper and lower filter screens and the transmission rod 5, that is, the upper tension pressure sensor 3 is fixedly connected to the upper filter screen 1, and the lower tension pressure sensor 4 is fixedly connected to the lower filter screen 2. The other end of the tension pressure sensor is fixed to the transmission rod 5, and the transmission rod 5 is driven by the power module 6. The two tension pressure sensors are used to collect the tension value of the filter screen during the screening process and measure the tension change of the filter screen in real time.

[0067] The power module 6 can use a vibration motor to provide vibration power for the filter screen to drive the screening process.

[0068] In this embodiment, the data acquisition and transmission module 10 is used to realize data collection, transmission and control feedback, and is one of the core components for realizing the intelligence of the device. By real-time and continuous reception of data from the tension and pressure sensor, especially the tension timing data of the filter screen, the timing and integrity of the data are guaranteed. The data acquisition and transmission module 10 communicates with the background server in a wireless or wired manner, transmits the tension timing data to the background server, receives the analysis results and control instructions of the background server, and the background server generates control instructions by analyzing the data, feeds back to the data acquisition and transmission module 10, and forwards to the control module 9 to instruct the control module 9 to adjust the working state of the device (for example, control the start or stop of the power module 6, etc.). Through real-time interaction with the background server, the powerful computing and analysis capabilities of the background can be used to optimize the decision-making process. The results of the background analysis can reflect the potential problems in the screening process, and the control instructions can accurately adjust the operation of the equipment, thereby improving the screening effect and stability. Through the coordinated work of the data acquisition and transmission module 10, the intelligent particle screening device can achieve high automation, reduce manual intervention, and the automatic adjustment and feedback mechanism can ensure that the screening process is always in the optimal state.

[0069] Furthermore, the outer frame is connected to the filter screen by a soft material, and the outer frame mainly serves the purpose of guiding particles and preventing dust. Due to the use of a soft material connection method, the outer frame has no force relationship with the internal power module 6 and does not participate in the vibration of the screen.

[0070] The soft material refers to a material that can be significantly deformed under the action of external forces and can recover to its original shape after deformation or has a large plastic deformation. Specifically, the soft material can be, for example, plastic, rubber, elastic cloth, etc.

[0071] When the screening device is working, due to the vibration of the screen, the impact and friction of the particles, certain vibrations or impacts will be generated between the outer frame and the filter. The use of soft materials for connection can effectively absorb and alleviate these mechanical vibrations and impacts, and effectively reduce the transmission of vibrations to the outer frame or other components. It not only ensures that the outer frame does not vibrate with the screening, which is beneficial to improving the quality of filter tension data collection, but also reduces equipment wear and extends the service life of the device. In addition, the soft material has good elasticity and adaptability, and can form a better sealing effect between the outer frame and the filter to prevent particulate materials from leaking through the seams, thereby improving the sealing and dustproof properties of the equipment.

[0072] In short, through improvements in real-time monitoring, data collection and analysis, automatic adjustment, remote control and adaptability of soft materials in the intelligent particle screening device, real-time and accurate monitoring of filter status data collection and analysis can be achieved. Without stopping the operation, it is possible to know in real time and accurately whether the filter is damaged, thereby improving the screening efficiency.

[0073] Optionally, in the intelligent particle screening device, the outer frame includes an outer frame bracket 17, an upper frame 12, a middle frame 13 and a lower frame 14; the upper frame 12 and the upper filter screen 1, and the middle frame 13 and the lower filter screen 2 are connected by soft materials and connected to the corresponding flow guide ports of the two filter screens.

[0074] In this embodiment, the entire weight of the outer frame is borne by the outer frame bracket 17, that is, the outer frame bracket 17 is the basic supporting part of the entire screening device, and is responsible for the overall support and stability of the device.

[0075] Specifically, Figure 2 As shown, the outer frame is composed of an upper frame 12, a middle frame 13, a lower frame 14 and an outer frame bracket 17. Among them, the upper frame 12 is located at the top of the device and is used to support the upper filter screen 1. The middle frame 13 plays a role in supporting and connecting the upper frame 12 and the lower frame 14, usually fixing the various parts together to ensure the firmness of the structure. The lower frame 14 is located at the bottom of the screening device.

[0076] The upper frame 12 and the upper filter screen 1, and the middle frame 13 and the lower filter screen 2 are connected through soft materials, and these frames are respectively connected to the flow guide ports of the filter screens.

[0077] Furthermore, the soft material can be specifically used by bonding or clamping to loosely connect the upper frame 12 and the middle frame 13 with the upper filter 1 and the lower filter 2 respectively, and then guide the particles into the diversion port, thereby effectively isolating the hard connections between different components. At the same time, since loose soft connections are adopted between the upper frame 12 and the upper filter 1, and the middle frame 13 and the lower filter 2, only the role of particle diversion can ensure that the tension data of the filter collected by the sensor is not disturbed, thereby ensuring the accuracy of tension measurement and reducing the negative effects of vibration, impact, etc. on the filter.

[0078] The main function of the diversion port is to divert the screened material according to its particle size. According to the different particle sizes, the diversion port is divided into a coarse discharge port 16 and a fine discharge port 15. The coarse discharge port 16 is a diversion port corresponding to the upper filter screen 1 and the lower filter screen 2, and is used to discharge the coarse particles that fail to pass through the screen after screening. The fine discharge port 15 is a diversion port under the layer filter screen, and is used to discharge the fine particles that pass through the screen.

[0079] It can be understood that the outer frame of the device is also provided with a feed port 11, which can be connected to the outer frame by bolts, welding or other fixing methods. The feed port 11 is an interface between the solid particulate material and the screening device, and its function is to feed the material to be screened into the screening area of ​​the screening device to ensure that the material can smoothly enter the screening system to start the screening process. Specifically, the feed port 11 can be set at the top of the screening device. Of course, the feed port 11 can also be set on one side of the screening device. The present embodiment does not limit the position of the feed port 11.

[0080] In an optional embodiment, the intelligent particle screening device further includes: a base 8; the base 8 is a square or rectangular frame with a storage space therein for accommodating the data acquisition and transmission module 10 and the control module 9; the base 8 is connected to the power module 6 via a shock-absorbing spring 7.

[0081] The base 8 is used to support the entire device and can be made of steel or other solid materials to provide sufficient strength and stability to ensure that the device does not deform or shake during operation.

[0082] The base 8 can be designed as a square or rectangular frame, and the four corners or edges can be reinforced to ensure the overall stability of the device. The accommodation space of the base 8 is used to accommodate various components of the device, mainly accommodating the data acquisition and transmission module 10 and the control module 9.

[0083] There is a shock-absorbing spring 7 between the power module 6 and the base 8. In other words, the shock-absorbing spring 7 connects the power module 6 and the base 8 to reduce the strong vibration generated when the screening device is working, reduce the interference of vibration on other components, and play a buffering role.

[0084] Embodiment 2:

[0085] Based on the intelligent particle screening device provided in any of the above embodiments, this embodiment provides a filter damage monitoring method, which is applied to an electronic device, which may have a display screen or a display function, or may be connected to a display screen; the electronic device includes a terminal and / or a backend server and / or a device with a display screen or a display function. In the model building stage, the method includes:

[0086] Step S101, receiving the filter tension time series data in the no-load state of the equipment collected by the intelligent particle screening device provided by any of the above embodiments.

[0087] The purpose of step S101 is to collect data when the device (intelligent particle screening device) is empty. Specifically, the upper tension is collected when the device is empty. and lower tension The time series data, where the upper tension refers to the tension value of the upper filter 1, and the lower tension refers to the tension value of the lower filter 2. It's time. , is the total duration of the collection.

[0088] Step S102: Based on the filter tension time series data, the period of the tension data is determined using the autocorrelation function (ACF). P . The autocorrelation function (ACF) is defined as:

[0089] (1)

[0090] Where: is the lag time after the start time. is the filter tension sequence in the The value of the timing point, is the mean tension, is the length of the sequence.

[0091] Based on formula (1), find the lag time corresponding to the first significant peak of the ACF function when the lag time is greater than 0, which is the period .

[0092] Step S103: according to the cycle P , the Fourier series is used to fit the filter tension time series data under the no-load state of the equipment to obtain the fitting periodic function.

[0093] Step S103 aims to implement Fourier series fitting of no-load data. Specifically, assuming that The fitting function is:

[0094] (2)

[0095] Where: is the fitting periodic function of the upper filter 1, is the fitting periodic function of the lower filter 2, is an integer representing the harmonic of the frequency, P It is a cycle. It's time. , , , is the Fourier coefficient, obtained by the least squares method or other fitting methods, is a constant term.

[0096] Step S104: subtract the value of the fitting periodic function at the same time point from the collected time series data of the filter tension in the no-load state of the equipment to obtain a fitting error.

[0097] Subtract the fitted periodic function from the collected tension data, and the corresponding expression is as follows:

[0098] (3)

[0099] Where: , are the fitting errors of the tension of the upper and lower filter screen 2, respectively; , are the values ​​of the fitted periodic function of the tension of the upper and lower filter screens 2, respectively. mod is the modulus operator, which indicates the t Perform remainder operation.

[0100] Step S105: If the fitting error is less than a preset error threshold, the fitting periodic function is used to perform periodic signal fitting reduction on the tension data collected under the working state to obtain the reduced filter tension time series data.

[0101] The purpose of steps S104 and S105 is to achieve periodic signal fitting reduction. Specifically, step S104 calculates and ,when and Basically stable and less than the acceptable error threshold , indicating that the idling fitting function (fitting periodic function) is valid, go to the next step, otherwise, if and Greater than the error threshold , re-optimize the fitting function.

[0102] Step S106, normalizing the reduced screen tension time series data, and calculating the upper and lower tension ratios of the normalized screen tension time series data.

[0103] The purpose of step S106 is to collect and preprocess training data. Specifically, the intelligent particle screening device collects the time series data of the filter screen tension in the working state, including: collecting the tension data when the equipment filter screen is intact, and collecting the tension time series data under working states with different damage degrees.

[0104] It should be noted that when collecting the tension time series data in the working state, steps S104 and S105 are executed to perform periodic signal fitting reduction on the collected original tension time series data to obtain the reduced filter tension time series data.

[0105] The time series data of the reduced filter tension includes the upper tension and lower tension ,in, , The total duration of data collection, It is the degree of filter damage, 0 means the filter is in good condition.

[0106] Furthermore, in order to improve the accuracy of the model, the reduced filter screen tension time series data is normalized, and the normalized result is used as the training data of the model. The normalized expression is as follows:

[0107] , (4)

[0108] Where: , are the normalized upper and lower filter tensions, and are the maximum and minimum functions respectively.

[0109] Considering that the tension relationship between the two filters is also an important data feature in the double-layer filter damage monitoring and analysis, the normalized ratio of the two-layer filter tension is added as a new feature value during the model training process:

[0110] (5)

[0111] Where: is the ratio of upper and lower tension.

[0112] By inputting the upper and lower tension ratio as a feature into the model for training, the model can learn the tension relationship between the two layers of filter and improve the accuracy of filter status judgment.

[0113] Step S107: input the normalized upper and lower filter screen tension time series data and the upper and lower tension ratio into the long short-term memory network for training to obtain a trained model.

[0114] Step S107 includes two aspects: model design and training. For model design, considering that the filter tension data itself is time series data, in order to better capture the intrinsic characteristics of these time series data, the model uses a long short-term memory network (LSTM) with a model input dimension of 3, which are , and That is, the normalized upper and lower filter tension time series data and the upper and lower tension ratio are used as the input data of the model.

[0115] In terms of model structure design, the size of the hidden layer and number of layers It is determined as a hyperparameter during training. The LSTM model outputs the predicted damage degree value after full connection and ReLU activation function.

[0116] The model training uses mean square error (MSE) as the loss function:

[0117] (6)

[0118] in, is the sample size, is the actual damage value, is the model's predicted value.

[0119] The Adam optimization algorithm is used to update the model parameters, and the learning rate is defined as The specific training process is as follows:

[0120] Divide the data samples into training set, validation set and test set (for example, 8:1:1), and only use the training set during training. Input the training data into the model in batches each time, and the number of each batch is defined as , calculate the loss function, and update the model parameters through the back propagation algorithm. The model's hyperparameters are confirmed by cross-validation. Finally, the accuracy of the test set is used to determine that the model training is complete, and the trained model is saved.

[0121] In order to further improve the generalization prediction ability of the model, the model building stage also includes:

[0122] Step S108: classify the degree of damage of the filter into multiple categories according to the actual degree of damage of the filter, and use the multiple categories as damage degree labels.

[0123] Step S109: adding the damage degree label to the test data set of the long short-term memory network.

[0124] Step S110: Utilize the trained model to calculate the damage degree prediction value of the test data, and calculate the best split point according to the damage degree prediction value of the test data and its damage degree label, and construct a decision tree.

[0125] Specifically, the data is divided into four damage degree categories: no damage, light damage (mild damage), moderate damage, and severe damage according to the actual damage degree. These labels are added to the test data set as damage degree labels. According to the predicted damage degree values ​​and damage degree labels in the test data, the optimal split point is calculated, and a decision tree is constructed to realize the determination of damage degree prediction values ​​to damage degree categories.

[0126] It should be noted that in the process of building a decision tree, the best split point refers to selecting a feature (or a certain value of a feature) to divide the data set into two parts, so that the divided subsets are purer, that is, the samples in the subsets are more similar. In this embodiment, the best split point is evaluated by the Gini index. The specific implementation steps can be performed with reference to the prior art, and this embodiment does not limit this.

[0127] In the model application phase, the method includes:

[0128] Step S111, normalizing the upper and lower filter screen tension time series data collected when the equipment is in working state and calculating the upper and lower tension ratio.

[0129] Here, data collection, normalization processing and calculation are performed to finally obtain the upper and lower filter tension time series data and the upper and lower tension ratio under the working state of the equipment. The above process can be performed with reference to steps S101 to S106 in the model building stage. The specific process is as follows:

[0130] Obtain the upper tension reduced by periodic signal fitting (fitting periodic function) under the working state of the filter and lower tension Time series data, normalize the data:

[0131] ,

[0132] ,

[0133] Calculate the tension ratio:

[0134] .

[0135] Step S112, loading the trained model, and inputting the normalized upper and lower filter screen tension time series data and the upper and lower tension ratio into the trained model, and calculating the damage degree prediction value in real time.

[0136] Specifically, when calculating the model, load the trained model and , and Input into the model to calculate the damage prediction value dr .

[0137] Step S113: Based on the predicted value of the damage degree, the decision tree is used to make a real-time judgment on the damage degree to obtain a damage degree category.

[0138] Specifically, the damage degree prediction value calculated in the previous step is dr Substitute it into the constructed decision tree classifier and output the predicted damage degree classification, that is, the damage degree category.

[0139] The present application is described in further detail below in conjunction with application examples.

[0140] Model construction and model application based on intelligent particle screening device, such as Figure 3 As shown, you can follow the steps below:

[0141] 1. Model building stage

[0142] (1) Equipment no-load data collection: The equipment idles for 10 seconds, and the data acquisition module is used to collect upper and lower tension sensor data at a frequency of 300 Hz. The sampling frequency is required to be more than 4 times the operating frequency of the vibration motor, and a total of 3000 data points are collected.

[0143] (2) Determine the period: Use formula (1) to calculate the autocorrelation function (ACF) of the no-load tension data. It is observed that the ACF function has the first significant peak at the lag of 6 data points, so the period is determined P is 6.

[0144] (3) Fourier series fitting of no-load data: Set , the no-load tension data is fitted by the 6th-order Fourier series to obtain the fitting function and The coefficient of .

[0145] (4) Periodic signal fitting and subtraction: Use the fitted periodic function to calculate the periodic tension value at each time point and subtract it from the original data to obtain and .calculate and If the standard deviation is less than the preset threshold =0.1, the fitting is considered valid.

[0146] (5) Training data collection and preprocessing: Select four types of filters, namely, intact filter, slightly damaged filter, moderately damaged filter and severely damaged filter, and install the four types of filters on the equipment respectively. Use the same weight of materials and quantify the degree of damage of the screen according to the remaining amount of coarse particle material per unit time. In this embodiment, the degree of damage of the filter can be set to 0, 0.01, 0.02, 0.04 and 0.1 respectively. Use different filters and run the equipment for 300 seconds respectively, collect tension data, and use formula (3) to perform periodic signal fitting and subtraction. Use formula (4) to normalize the data, and use formula (5) to calculate the upper and lower tension ratio. . Combine all data and add corresponding, quantitative damage degree labels.

[0147] (6) Model design and training: Use Pytorch to build an LSTM model with an input dimension of 3 ( , and ), hidden layer size h =64, number of layers lays =1, the output layer is a single neuron, using the ReLU activation function. The Adam optimizer is used (learning rate =0.0004), and the MSE loss function is used for training. The dataset is divided into a training set (80%), a validation set (10%), and a test set (10%), using a batch size of batch =64, number of iterations epoch =500 for training. The prediction results of the model on the test set after training are as follows Figure 4 As shown, the mean absolute error (MAE) of the model test set is 0.003239.

[0148] (7) Damage degree classification: Since there are only 4 leaf points, the maximum depth of the decision tree is set to 3, and a decision tree classifier is constructed based on the predicted values ​​and true labels of the test set. Figure 5 As shown, the constructed decision tree has a prediction accuracy of 99.9% for the no damage category, 99.4% for the light damage category, 99.8% for the moderate damage category, and 100% for the severe damage category.

[0149] 2. Model Application Phase

[0150] (1) Equipment no-load data collection: The equipment idles for 10 seconds, and the data acquisition module is used to collect upper and lower tension sensor data at a frequency of 300 Hz. The sampling frequency is required to be more than 4 times the operating frequency of the vibration motor, and a total of 3000 data points are collected.

[0151] (2) Determine the period: Use formula (1) to calculate the autocorrelation function (ACF) of the no-load tension data. It is observed that the ACF function has the first significant peak at the lag of 6 data points, so the period is determined P .

[0152] (3) Fourier series fitting of no-load data: Set , perform a 6th-order Fourier series fit on the no-load tension data and obtain the fitting function and The coefficient of .

[0153] (4) Periodic signal fitting and subtraction: Use the fitting function to calculate the periodic tension value at each time point and subtract it from the original data to obtain and .calculate and If the standard deviation is less than the preset threshold =0.1, the fitting is considered valid.

[0154] (5) Collecting equipment carrying tension data: The equipment starts screening the granular material, collects tension data, and uses the fitted Functions and The function performs periodic signal fitting and elimination on the tension data, and further normalizes the data to output , and calculate the tension ratio .

[0155] (6) Model calculation: , Input to the trained LSTM model and output the quantified damage value .

[0156] (7) Predict damage degree classification: The quantitative damage degree value calculated in the previous step , substituted into the constructed decision tree classifier, and the output is the predicted damage degree classification, that is, the damage degree category.

[0157] (8) The control module 9 sends an instruction: According to the predicted damage degree classification, if it is not in a non-damaged state, the control module 9 sends a stop working instruction to control the power module 6 to stop working and prompt the maintenance personnel to carry out repairs in time.

[0158] By using the filter damage monitoring method provided in this embodiment, the time from filter damage to monitoring and identification can be controlled within 5 seconds, which is timely. Because the filter damage is often a single layer damage, rather than two layers of filter damage at the same time, it can be ensured that the fine material screened before the shutdown is still usable and will not contaminate the products screened in the previous stage, which is highly efficient.

[0159] In summary, the filter damage monitoring method provided in this embodiment has the following advantages:

[0160] Strong real-time performance: This embodiment uses a tension pressure sensor to monitor the change of filter screen tension in real time without stopping for monitoring. It is combined with the LSTM model for real-time analysis, which can promptly detect filter screen damage and avoid reduced production efficiency or product quality problems caused by filter screen damage.

[0161] High accuracy: The LSTM model is used to model the filter tension time series data, which can capture the complex pattern of tension changes during filter damage, thereby improving the accuracy of damage degree determination. Compared with the traditional threshold-based judgment method, the data-driven method provided in this embodiment is more objective and accurate.

[0162] High degree of intelligence: The solution of this embodiment integrates functions such as data collection, data processing, model calculation and damage degree determination, realizing the automation and intelligence of filter damage monitoring. The evaluation of the filter status can be completed without human intervention, reducing labor costs and human errors.

[0163] Predictability: By predicting the extent of filter damage, maintenance can be scheduled in advance to avoid production interruptions caused by sudden filter damage. Traditional monitoring methods can usually only monitor damage that has already occurred and cannot predict it.

[0164] Reduced maintenance costs: Real-time monitoring and prediction can avoid unnecessary filter replacement, thereby reducing maintenance costs.

[0165] In summary, the technical solution of this embodiment provides a real-time, accurate, intelligent and predictable filter damage monitoring method, which can effectively improve screening efficiency, reduce maintenance costs, and improve product quality.

[0166] Embodiment 3:

[0167] Based on the same inventive concept, this embodiment further provides an electronic device, the electronic device comprising:

[0168] one or more processors;

[0169] The computer-readable storage medium may be configured to store one or more programs, and when one or more processors execute the one or more programs, the following steps are implemented:

[0170] S101, receiving the time series data of the filter tension in the no-load state collected by the intelligent particle screening device;

[0171] S102, based on the time series data of the tension of the filter screen under no-load condition, using the autocorrelation function to determine the period of the time series data of the tension under no-load condition P ; Wherein, the autocorrelation function is defined as:

[0172] ,

[0173] In the formula, is the lag time after the start time. is the filter tension sequence in the The value of the timing point, is the mean tension, is the length of the sequence;

[0174] S103, according to the cycle P , the Fourier series is used to fit the time series data of the filter tension under the no-load state of the equipment to obtain the fitting periodic function;

[0175] S104, subtracting the value of the fitting periodic function at the same time point from the collected time series data of the filter tension in the no-load state of the equipment to obtain a fitting error;

[0176] S105, if the fitting error is less than a preset error threshold, the fitting periodic function is used to perform periodic signal fitting reduction on the tension data collected under the working state to obtain the reduced tension time series data of the filter screen; wherein the tension time series data collected under the working state includes the tension time series data when the filter screen is intact and the tension time series data of the filter screen under different degrees of damage;

[0177] S106, normalizing the reduced filter screen tension time series data, and calculating the upper and lower tension ratios of the normalized filter screen tension time series data;

[0178] S107, inputting the normalized upper and lower filter tension time series data and the upper and lower tension ratio into the long short-term memory network for training to obtain a trained model;

[0179] S108, classifying the degree of damage of the filter into multiple categories according to the actual degree of damage of the filter, and using the multiple categories as damage degree labels;

[0180] S109, adding the damage degree label to the test data set of the long short-term memory network;

[0181] S110, using the trained model, calculating the damage degree prediction value of the test data, and calculating the optimal split point according to the damage degree prediction value of the test data and its damage degree label, and constructing a decision tree;

[0182] During the model application phase:

[0183] S111, normalizing the upper and lower filter screen tension time series data collected when the equipment is in working state and calculating the upper and lower tension ratio;

[0184] S112, loading the trained model, and inputting the normalized upper and lower filter screen tension time series data and the upper and lower tension ratio into the trained model, and calculating the damage degree prediction value in real time;

[0185] S113. According to the predicted value of the damage degree, the decision tree is used to make a real-time judgment on the damage degree to obtain a damage degree category.

[0186] This embodiment further provides an electronic device, the hardware structure of which may include: a processor, a communication interface, a computer-readable storage medium (also referred to as a memory), and a communication bus.

[0187] The processor, the communication interface, and the computer-readable storage medium communicate with each other via a communication bus.

[0188] A computer-readable storage medium may be configured to store one or more programs.

[0189] Optionally, the communication interface may be an interface of a communication module, such as an interface of a GSM module.

[0190] The processor executes one or more programs, which implement the following steps:

[0191] S101, receiving the time series data of the filter tension in the no-load state collected by the intelligent particle screening device;

[0192] S102, based on the time series data of the tension of the filter screen under no-load condition, using the autocorrelation function to determine the period of the time series data of the tension under no-load condition P ; Wherein, the autocorrelation function is defined as:

[0193] ,

[0194] In the formula, is the lag time after the start time. is the filter tension sequence in the The value of the timing point, is the mean tension, is the length of the sequence;

[0195] S103, according to the cycle P , the Fourier series is used to fit the time series data of the filter tension under the no-load state of the equipment to obtain the fitting periodic function;

[0196] S104, subtracting the value of the fitting periodic function at the same time point from the collected time series data of the filter tension in the no-load state of the equipment to obtain a fitting error;

[0197] S105, if the fitting error is less than a preset error threshold, the fitting periodic function is used to perform periodic signal fitting reduction on the tension data collected under the working state to obtain the reduced tension time series data of the filter screen; wherein the tension time series data collected under the working state includes the tension time series data when the filter screen is intact and the tension time series data of the filter screen under different degrees of damage;

[0198] S106, normalizing the reduced filter screen tension time series data, and calculating the upper and lower tension ratios of the normalized filter screen tension time series data;

[0199] S107, inputting the normalized upper and lower filter tension time series data and the upper and lower tension ratio into the long short-term memory network for training to obtain a trained model;

[0200] S108, classifying the degree of damage of the filter into multiple categories according to the actual degree of damage of the filter, and using the multiple categories as damage degree labels;

[0201] S109, adding the damage degree label to the test data set of the long short-term memory network;

[0202] S110, using the trained model, calculating the damage degree prediction value of the test data, and calculating the optimal split point according to the damage degree prediction value of the test data and its damage degree label, and constructing a decision tree;

[0203] During the model application phase:

[0204] S111, normalizing the upper and lower filter screen tension time series data collected when the equipment is in working state and calculating the upper and lower tension ratio;

[0205] S112, loading the trained model, and inputting the normalized upper and lower filter screen tension time series data and the upper and lower tension ratio into the trained model, and calculating the damage degree prediction value in real time;

[0206] S113. According to the predicted value of the damage degree, the decision tree is used to make a real-time judgment on the damage degree to obtain a damage degree category.

[0207] The processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc. It may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application may be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0208] (1) Mobile communication devices: These devices are characterized by having mobile communication functions and their main purpose is to provide voice and data communications. These terminals include: smart phones (e.g. iPhone), multimedia phones, functional phones, and low-end phones.

[0209] (2) Ultra-mobile personal computer devices: These devices fall into the category of personal computers, have computing and processing functions, and generally also have mobile Internet access features. These terminals include: PDA, MID and UMPC devices, such as iPad.

[0210] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players (e.g. iPod), handheld game consoles, e-books, smart toys and portable car navigation devices.

[0211] (4) Server: A device that provides computing services. The server consists of a processor, hard disk, memory, system bus, etc. The server is similar to a general computer architecture, but because it needs to provide highly reliable services, it has higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability.

[0212] (5) Other electronic devices with data interaction functions.

[0213] In an exemplary embodiment, the embodiment of the present application also provides an electronic device, including a processor and a memory for storing a computer program that can be run on the processor, wherein the processor is used to execute the steps of any of the above-mentioned filter damage monitoring methods when running the computer program.

[0214] An embodiment of the present application also provides a computer-readable storage medium, that is, a storage medium, for example, including a memory for storing a computer program, on which a computer program is stored. The above-mentioned computer program can be executed by a processor of an electronic device to complete the steps described in the aforementioned filter damage monitoring method.

[0215] The computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface storage, optical disk, or CD-ROM.

[0216] The embodiment of the present application also provides a computer program product, which includes computer executable instructions, and the computer executable instructions are used to implement the steps of any of the above-mentioned filter damage monitoring methods when executed by a computer.

[0217] It should be pointed out that, according to the needs of implementation, the various components / steps described in the embodiments of the present application can be split into more components / steps, or two or more components / steps or partial operations of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of the present application.

[0218] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for monitoring filter damage of an intelligent particle screening device, the intelligent particle screening device comprising: Upper and lower filter screens, two tension and pressure sensors, power module, data acquisition and transmission module; The mesh size of the two layers of filter screens is the same and is used for screening particles; The two layers of the filter screen are fixedly connected to the two tension and pressure sensors respectively; the other end of the tension and pressure sensor is fixed to the transmission rod, and the transmission rod is driven by the power module; the data acquisition and transmission module is used to receive the filter screen tension time series data collected by the tension and pressure sensor and forward it to the background server; it is characterized in that, in the model construction stage: S101, receiving the time series data of the filter tension in the no-load state collected by the intelligent particle screening device; S102, based on the time series data of the tension of the filter screen under no-load condition, using the autocorrelation function to determine the period of the time series data of the tension under no-load condition P ; Wherein, the autocorrelation function is defined as: ， In the formula, is the lag time after the start time. is the filter tension sequence in the The value of the timing point, is the mean tension, is the length of the sequence; S103, according to the cycle P , the Fourier series is used to fit the time series data of the filter tension under the no-load state of the equipment to obtain the fitting periodic function; S104, subtracting the value of the fitting periodic function at the same time point from the collected time series data of the filter tension in the no-load state of the equipment to obtain a fitting error; S105, if the fitting error is less than a preset error threshold, the fitting periodic function is used to perform periodic signal fitting reduction on the tension data collected under the working state to obtain the reduced tension time series data of the filter screen; wherein the tension time series data collected under the working state includes the tension time series data when the filter screen is intact and the tension time series data of the filter screen under different degrees of damage; S106, normalizing the reduced filter screen tension time series data, and calculating the upper and lower tension ratios of the normalized filter screen tension time series data; S107, inputting the normalized upper and lower filter tension time series data and the upper and lower tension ratio into the long short-term memory network for training to obtain a trained model; S108, classifying the degree of damage of the filter into multiple categories according to the actual degree of damage of the filter, and using the multiple categories as damage degree labels; S109, adding the damage degree label to the test data set of the long short-term memory network; S110, using the trained model, calculating the damage degree prediction value of the test data, and calculating the optimal split point according to the damage degree prediction value of the test data and its damage degree label, and constructing a decision tree; During the model application phase: S111, normalizing the upper and lower filter screen tension time series data collected when the equipment is in working state and calculating the upper and lower tension ratio; S112, loading the trained model, and inputting the normalized upper and lower filter screen tension time series data and the upper and lower tension ratio into the trained model, and calculating the damage degree prediction value in real time; S113. According to the predicted value of the damage degree, the decision tree is used to make a real-time judgment on the damage degree to obtain a damage degree category.

2. An intelligent particle screening device, characterized in that: include: Two layers of filters, two tension and pressure sensors, a power module, a data acquisition and transmission module, a control module and an outer frame; The mesh size of the two layers of filter screens is the same and is used for screening particles; The two layers of the filter screen are fixedly connected to the two tension and pressure sensors respectively; the other end of the tension and pressure sensor is fixed to the transmission rod, and the transmission rod is driven by the power module; The data acquisition and transmission module is used to receive the filter tension time series data collected by the tension and pressure sensor and forward it to the background server, and at the same time receive the analysis results and control instructions returned by the background server, and send the control instructions to the control module; The backend server is used to execute the filter damage monitoring method according to claim 1 to monitor the filter damage; The control module is used to control the working state of the intelligent particle screening device according to the control instruction.

3. The intelligent particle screening device according to claim 2, characterized in that: The outer frame comprises an outer frame bracket, an upper frame, a middle frame and a lower frame; The upper frame and the upper filter screen, and the middle frame and the lower filter screen are connected through soft materials, and are connected to the corresponding flow guide ports of the two filter screens.

4. The intelligent particle screening device according to claim 2, characterized in that: The intelligent particle screening device further includes: a base; The base is a square or rectangular frame with a storage space inside for accommodating the data acquisition and transmission module and the control module; The base is connected to the power module via a shock-absorbing spring.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the filter damage monitoring method according to claim 1 are implemented.

6. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the filter damage monitoring method as claimed in claim 1 when executing the computer program.

7. A computer program product, characterized in that It contains computer executable instructions, and the computer executable instructions are used to implement the steps of the filter damage monitoring method as claimed in claim 1 when executed by a computer.

Images