Pea protein production device with filtering structure and filtering method

Abstract

The invention relates to the technical field of pea protein production, and particularly discloses a pea protein production device with a filtering structure and a filtering method.The pea protein production device comprises an upper cover plate, a lower cover plate, a first side plate, a second side plate, a third side plate and a fourth side plate, a feeding port is formed in the upper cover plate, and a discharging port is formed in the fourth side plate; a coarse filter screen, a fine filter screen and a nano film screen are sequentially arranged in the direction from the upper cover plate to the lower cover plate, the bottom of each layer of filter screen is fixedly connected with a vibrator, a plurality of convex rods are fixedly arranged on the inner wall of the fourth side plate, a groove is formed between every two convex rods and is matched with one end of the filter screen, and a pressure sensor, a turbidity measuring device and a temperature sensor are arranged at the position of the filter screen. According to the method, a vibration adjustment prediction model is constructed, the optimal vibration frequency and vibration duration are output in real time through monitoring data of each layer of filter screen sensor, accurate adjustment and control of filter screen blockage are achieved, it is ensured that the whole filtering process is in an efficient filtering state, and the production efficiency is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pea protein production, and particularly relates to a pea protein production device with a filtering structure and a filtering method. Background Art

[0002] As a high-quality plant protein source, the production demand for pea protein is continuously increasing. During the production process of pea protein, it is necessary to filter pea liquid to separate pea residues and slurry, and then precipitate and purify the remaining slurry to obtain high-quality pea protein. Among them, the filtering link directly affects the purity and production efficiency of pea protein.

[0003] Traditional filtering devices achieve solid-liquid separation through static screening or mechanical vibration at a single frequency. However, due to the presence of impurities such as starch and fiber in pea liquid, it is easy to cause the filter screen to become blocked, reducing the filtering efficiency and even affecting protein activity. Although some devices can relieve the blockage by electric beating or centrifugal vibration of the filter plate, the fixed vibration parameters cannot respond to dynamic working conditions such as the degree of filter screen blockage and slurry viscosity. When the vibration parameters are set too large, local overload is likely to occur, resulting in deformation of the filter screen and reducing the service life of the equipment. When the vibration parameters are set too small, the filtering may be incomplete, reducing the filtering efficiency. Summary of the Invention

[0004] To solve the above problems, the present invention provides a pea protein production device with a filtering structure. Its main body is a cuboid, including an upper cover plate, a lower cover plate, a first side plate, a second side plate, a third side plate, and a fourth side plate. There is a feed inlet on the upper cover plate and a discharge outlet on the fourth side plate. A coarse filter screen, a fine filter screen, and a nano-film screen are sequentially arranged from the upper cover plate to the lower cover plate. A vibrator a is fixedly connected to the bottom of the coarse filter screen, a vibrator b is fixedly connected to the bottom of the fine filter screen, and a vibrator c is fixedly connected to the bottom of the nano-film screen. A plurality of convex rods are fixedly arranged on the inner wall of the fourth side plate, and all the convex rods are perpendicular to the first side plate. A groove is formed between every two convex rods, and the groove is adapted to one side of each layer of filter screen. A turbidity detector a is horizontally arranged at the upper convex rod of the coarse filter screen, and a pressure sensor a is vertically arranged at the lower convex rod. A turbidity detector b is horizontally arranged at the upper convex rod of the fine filter screen, and a pressure sensor b is vertically arranged at the lower convex rod. A temperature sensor is horizontally arranged at the upper convex rod of the nano-film screen, and a pressure sensor c is vertically arranged at the lower convex rod.

[0005] Further, handle a, handle b, and handle c are respectively provided on the outer wall of the second side plate. Handle a is fixedly connected to the coarse filter screen through connecting block a, handle b is fixedly connected to the fine filter screen through connecting block b, and handle c is fixedly connected to the nano-membrane net through connecting block c. By pulling the handle, the filter screen is driven to take out each layer of the filter screen from the device for cleaning. The two sides of each layer of the filter screen are closely attached to the first side plate and the third side plate respectively, and a cavity is formed between every two layers of the filter screen. A buffer plate is fixedly provided at the lower part of the nano-membrane net to buffer and collect the filtered pea liquid. There is an included angle between the buffer plate and the fourth side plate, and the range of the included angle is 45° to 60°. The two sides of the buffer plate are hermetically connected to the first side plate and the third side plate respectively.

[0006] Further, the aperture size of the coarse filter screen is 100 - 300 um, the aperture size of the fine filter screen is 10 - 50 um, and the aperture size of the nano-membrane net is 2 - 50 nm.

[0007] Further, a flip cover is fixedly connected to the upper part of the feed inlet, and a control panel is provided on the first side plate for controlling all devices. Connectors are provided on vibrator a, vibrator b, and vibrator c for facilitating connection to an external power supply. The vibration frequencies of vibrator a, vibrator b, and vibrator c are 10 - 150 Hz, and different frequency values are output in real time according to the system command.

[0008] A filtering method for a pea protein production device with a filtering structure includes the following steps: I. Open the flip cover, pour the pea soybean milk from the feed inlet, and close the flip cover.

[0009] II. Start the switch. As the pea soybean milk flows into the coarse filter screen layer, pressure sensor a at the coarse filter screen monitors the pressure difference before and after the coarse filter screen , turbidity detector a monitors the concentration of suspended substances in the inlet pea soybean milk , then the pea soybean milk flows into the fine filter screen layer, pressure sensor b at the fine filter screen monitors the pressure difference before and after the fine filter screen , turbidity detector b monitors the concentration of suspended substances in the outlet pea soybean milk , and finally the pea soybean milk flows into the nano-membrane net layer, pressure sensor c at the nano-membrane net monitors the transmembrane pressure difference , and the temperature sensor monitors the frictional heat .

[0010] III. Based on the historical database, the random forest algorithm is selected to explore the non-linear relationship between vibration adjustment and variables such as filter screen pressure difference and suspended substance concentration, and a vibration adjustment prediction model is constructed.

[0011] IV. Use the test sample data to evaluate the prediction result. Assume the test set is , where each sample All correspond to a set of actual values of vibration frequency and vibration duration , the prediction results obtained by predicting each sample through the model , if the prediction result and the actual value The error is within 5%, which means the prediction result is accurate. Finally, the accuracy rate is calculated to be 89.3%, and the model prediction result is good.

[0012] V. The six parameters of the monitored , , , , , are used as the input layer and input into the vibration adjustment prediction model in the control panel. The prediction model outputs the vibration frequency and vibration duration indicators. At this time, the vibrators a, b, and c are adjusted to work according to the output indicators.

[0013] VI. The vibration adjustment prediction model dynamically adjusts the vibration frequency and vibration duration of each vibrator according to the real-time monitoring data until the filtration ends.

[0014] The beneficial effects of the present invention are as follows: 1. The present invention embeds the vibration adjustment prediction model in the device system, proposes an intelligent vibration control system, constructs the vibration adjustment prediction model based on the historical experimental database using the random forest algorithm, monitors the pressure difference, turbidity, and temperature changes of each filter screen layer, and the prediction model outputs the optimal vibration frequency and vibration duration in real time. The system adjusts the vibrators of each filter screen according to the output indicators, realizes the precise control of the filter screen blockage, ensures that the entire filtration process maintains an efficient filtration state, avoids the blockage of the filter screen, and improves the production efficiency.

[0015] 2. Compared with the traditional continuous fixed vibration mode, the present invention not only reduces the occurrence of local overload or incomplete filtration, but also increases or decreases the vibration frequency and duration according to the multi-dimensional working conditions of the filter screen, significantly reducing the energy consumption.

[0016] 3. The present invention improves the product quality by setting three-level filter screens. The coarse filter screen layer intercepts large particle fibers such as pea dregs, the fine filter screen layer removes micron-level starch, and the nano-film screen intercepts colloidal impurities, realizing gradual and refined separation and improving the purity of pea protein. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a schematic cross-sectional structure diagram of a pea protein production device with a filtration structure according to the present invention; Figure 2 is a schematic side structure diagram of a pea protein production device with a filtration structure according to the present invention; Figure 3Schematic diagram of the overall structure of the pea protein production device with a filtering structure according to the present invention; Figure 4 Schematic diagram of the bottom structure of the filter screen of the pea protein production device with a filtering structure according to the present invention; Figure 5 Schematic diagram of the structure of vibrator a of the pea protein production device with a filtering structure according to the present invention; Figure 6 Flowchart for evaluating the test results of the random forest algorithm according to the present invention.

[0018] Reference numerals: 1, upper cover plate; 2, lower cover plate; 3, first side plate; 4, second side plate; 5, third side plate; 6, fourth side plate; 7, feed inlet; 8, discharge outlet; 9, coarse filter screen; 10, fine filter screen; 11, nano-film screen; 12, vibrator a; 13, vibrator b; 14, vibrator c; 15, convex rod; 16, turbidity detector a; 17, pressure sensor a; 18, turbidity detector b; 19, pressure sensor b; 20, temperature sensor; 21, pressure sensor c; 22, handle a; 23, handle b; 24, handle c; 25, connecting block a; 26, connecting block b; 27, connecting block c; 28, buffer plate; 29, flip cover; 30, control panel; 31, joint. Detailed implementation manners

[0019] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0020] A pea protein production device with a filtering structure, as Figures 1-5As shown in the figure, its main body is a cuboid, including an upper cover plate 1, a lower cover plate 2, a first side plate 3, a second side plate 4, a third side plate 5 and a fourth side plate 6. There is a feed inlet 7 on the upper cover plate 1 and a discharge outlet 8 on the fourth side plate 6. A coarse filter screen 9, a fine filter screen 10 and a nano-membrane screen 11 are arranged in sequence from the upper cover plate 1 to the lower cover plate 2. A vibrator a 12 is fixedly connected to the bottom of the coarse filter screen 9, a vibrator b 13 is fixedly connected to the bottom of the fine filter screen 10, and a vibrator c 14 is fixedly connected to the bottom of the nano-membrane screen 11. The vibrators are used to adjust the vibration frequency and vibration duration in real time. A plurality of convex rods 15 are fixedly arranged on the inner wall of the fourth side plate 6. All the convex rods 15 are perpendicular to the first side plate 3. A groove is formed between every two convex rods 15. A turbidity detector a 16 is horizontally arranged at the upper convex rod 15 of the coarse filter screen 9, and a pressure sensor a 17 is vertically arranged at the lower convex rod 15. A turbidity detector b 18 is horizontally arranged at the upper convex rod 15 of the fine filter screen 10, and a pressure sensor b 19 is vertically arranged at the lower convex rod 15. A temperature sensor 20 is horizontally arranged at the upper convex rod 15 of the nano-membrane screen 11, and a pressure sensor c 21 is vertically arranged at the lower convex rod 15. Among them, the pressure sensors of each layer of filter screen are used to monitor the pressure difference before and after each layer of filter screen. The turbidity detectors are used to measure the concentration of suspended substances. The temperature sensor is used to monitor the frictional heat between the transmembrane layers. The size of one end of each layer of filter screen plate is adapted to the size of the groove formed between the upper convex rod 15 and the pressure sensor.

[0021] On the basis of the above scheme, the present invention can be further improved. Handles a 22, b 23 and c 24 are respectively arranged on the outer wall of the second side plate 4. The handle a 22 is fixedly connected to the coarse filter screen 9 through a connecting block a 25, the handle b 23 is fixedly connected to the fine filter screen 10 through a connecting block b 26, and the handle c 24 is fixedly connected to the nano-membrane screen 11 through a connecting block c 27. By pulling the handle, the filter screen is driven to take out each layer of filter screen from the device, which is convenient for cleaning; the two sides of each layer of filter screen are closely attached to the first side plate 3 and the third side plate 5 respectively. A cavity is formed between every two layers of filter screens. A buffer plate 28 is fixedly arranged at the lower part of the nano-membrane screen 11, which plays a role in buffering and collecting the filtered pea liquid to avoid liquid splashing. There is an included angle between the buffer plate 28 and the fourth side plate 6, and the included angle range is 45° - 60°. The two sides of the buffer plate 28 are hermetically connected to the first side plate 3 and the third side plate 5 respectively.

[0022] On the basis of the above scheme, the present invention can be further improved. The aperture size of the coarse filter screen 9 is 100 - 300um, which is used to filter out large particle impurities such as pea dregs. The aperture size of the fine filter screen 10 is 10 - 50um, which is used to filter out fine particles such as starch. The aperture size of the nano-membrane screen 11 is 2 - 50nm, which is used to intercept colloidal impurities.

[0023] On the basis of the above solution, the present invention can be further improved. A flip cover 29 is fixedly connected to the upper part of the feed inlet 7. A control panel 30 is provided on the first side plate 3, which is embedded with a vibration adjustment prediction model for controlling all electrical equipment. Connectors 31 are provided on the vibrators a12, b13 and c14 for facilitating connection to an external power supply. The vibration frequencies of the vibrators a12, b13 and c14 are 10 - 150 Hz, and different frequency values are output in real time according to the system command.

[0024] A filtering method for a pea protein production device with a filtering structure includes the following steps: I. Open the flip cover 29, pour the pea soybean milk from the feed inlet 7, and close the flip cover 29.

[0025] II. Start the switch. As the pea soybean milk flows into the coarse filter screen 9 layer, the pressure sensor a17 at the coarse filter screen 9 monitors the pressure difference before and after the coarse filter screen 9 , and the turbidity detector a16 monitors the concentration of suspended substances in the inlet pea soybean milk . Subsequently, the pea soybean milk flows into the fine filter screen 10 layer. The pressure sensor b19 at the fine filter screen 10 monitors the pressure difference before and after the fine filter screen 10 , and the turbidity detector b18 monitors the concentration of suspended substances in the outlet pea soybean milk . Finally, the pea soybean milk flows into the nano - membrane network 11 layer. The pressure sensor c21 at the nano - membrane network 11 monitors the transmembrane pressure difference , and the temperature sensor 20 monitors the frictional heat .

[0026] III. Based on the historical database, the random forest algorithm is selected to explore the non - linear relationship between vibration adjustment and variables such as filter screen pressure difference and suspended substance concentration, and a vibration adjustment prediction model is constructed. Among them, the core of the present invention lies in the construction of the vibration adjustment prediction model. According to the actual process of pea protein filtration production, the vibration frequency and vibration duration of the vibrator and variables such as the pressure difference before and after the filter screen belong to a complex data set, and there is a complex non - linear relationship between the variables. Therefore, machine learning - random forest is selected as the core algorithm of the vibration adjustment prediction model. The random forest algorithm randomly selects training data and feature subsets, generates multiple decision trees, and averages or votes on the prediction results of each decision tree to obtain the final prediction result.

[0027] The process of establishing the model includes collecting the data set, sample sampling, feature sampling, constructing decision trees, generating the decision tree model and integrating the prediction results. First, read the pressure difference before and after the coarse filter screen , the inlet suspended substance concentration , the pressure difference before and after the fine filter screen , the outlet suspended substance concentration , the transmembrane pressure difference and temperature Taking the vibration amplitude as the independent variable, the vibration frequency and vibration duration as the predicted dependent variables, and a complete set of index data as a training sample value. After removing abnormal data, all sample data are read into a set, and training samples and test samples are constructed, as shown in Table 1 and Table 2. A sample includes a complete row of data. Assuming the number of rows of the training samples is N rows, a sample is randomly selected from the training samples and put into the sampling set, and then the sample is put back into the original training samples. After N random sampling operations, a training subset containing N samples is obtained.

[0028] Table 1. Training sample data

[0029] Table 2. Test sample data

[0030] IV. Randomly select m features from the feature set, sort the data records according to this feature, and split the data for recording according to the sorted feature values. If all records have the same category, mark it as a leaf node. If the records have different categories, continue to recursively check this node until the number of decisions reaches the default value of 100, complete the construction of the vibration adjustment prediction model, use the test sample data to evaluate the test results, and calculate the prediction accuracy rate of the model.

[0031] Assume the test set is , where each sample corresponds to a set of actual values of vibration frequency and vibration duration , and the prediction result obtained by predicting each sample through the model . If the prediction result has an error within 5% of the actual value , finally, the value of the counter is divided by the number of samples in the test set , and the prediction accuracy rate of the vibration adjustment prediction model can be obtained. The flow chart is as Figure 6 shown. After training, the calculated accuracy rate is , where represents the number of samples with correct predictions, and represents the number of samples in the test set.

[0032] V. When it is monitored that the equipment has run for 98 seconds, at this time, , , , , , , the parameters are automatically input into the vibration adjustment prediction model in the control panel 30. The prediction model outputs the optimal vibration frequency of 38 Hz and the vibration duration of 9 s. The vibrator operates according to the optimal vibration value. When the device runs to the 107th second, it is monitored that , , , , , , the prediction model outputs the optimal vibration frequency of 20 Hz and the vibration duration of 4 s. The vibrator continues to operate according to the new optimal vibration value,... This filtering process is cycled.

[0033] VI. The vibration adjustment prediction model dynamically adjusts the vibration frequency and vibration duration of each vibrator according to the real-time monitoring data until the filtering ends.

[0034] All the empty components of the present invention are connected to the control panel 30 by existing electrical connection means. The control panel 30 is encapsulated with a vibration adjustment prediction model and is equipped with an automatic control system, a data acquisition and storage component, and a touch display screen.

Claims

1. A pea protein production device with a filtering structure, wherein the main body is a rectangular parallelepiped, comprising an upper cover plate (1), a lower cover plate (2), a first side plate (3), a second side plate (4), a third side plate (5) and a fourth side plate (6), characterized in that: The upper cover plate (1) is provided with a feed inlet (7), the fourth side plate (6) is provided with a discharge port (8), a coarse filter (9), a fine filter (10) and a nano-membrane net (11) are arranged in sequence from the upper cover plate (1) to the lower cover plate (2), the bottom of the coarse filter (9) is fixedly connected to a vibrator a (12), the bottom of the fine filter (10) is fixedly connected to a vibrator b (13), and the bottom of the nano-membrane net (11) is fixedly connected to a vibrator c (14), and a plurality of protruding rods (15) are fixedly provided on the inner wall of the fourth side plate (6), and every two protruding rods (15) form a A concave groove is adapted to one side of each layer of filter screen, a turbidity meter a (16) is horizontally provided at the upper protruding rod (15) of the coarse filter screen (9), and a pressure sensor a (17) is vertically provided at the lower protruding rod (15); a turbidity meter b (18) is horizontally provided at the upper protruding rod (15) of the fine filter screen (10), and a pressure sensor b (19) is vertically provided at the lower protruding rod (15); a temperature sensor (20) is horizontally provided on the upper protruding rod (15) of the nano-membrane net (11), and a pressure sensor c (21) is vertically provided at the lower protruding rod (15).

2. The pea protein production device with a filtering structure according to claim 1, characterized in that: The outer wall of the second side panel (4) is provided with a handle a (22), a handle b (23) and a handle c (24), respectively; the handle a (22) is fixedly connected to the coarse filter (9) via a connecting block a (25); the handle b (23) is fixedly connected to the fine filter (10) via a connecting block b (26); the handle c (24) is fixedly connected to the nano-membrane net (11) via a connecting block c (27); the two side edges of each layer of the filter net are respectively tightly fitted to the first side panel (3) and the third side panel (5); a buffer plate (28) is fixedly provided at the bottom of the nano-membrane net (11); an angle is formed between the buffer plate (28) and the fourth side panel (6), the angle ranging from 45° to 60°; the two side edges of the buffer plate (28) are respectively tightly connected to the first side panel (3) and the third side panel (5).

3. The pea protein production device with a filtering structure according to claim 1, characterized in that: The pore size of the coarse filter (9) is 100-300 um, the pore size of the fine filter (10) is 10-50 um, and the pore size of the nano-membrane net (11) is 2-50 nm.

4. The pea protein production device with a filtering structure according to claim 1, characterized in that: The upper portion of the feed port (7) is fixedly connected to a flap (29); a control panel (30) is provided on the first side panel (3); and joints (31) are provided on the vibrator a (12), the vibrator b (13), and the vibrator c (14); and the vibration frequencies of the vibrator a (12), the vibrator b (13), and the vibrator c (14) are 10 to 150 Hz.

5. A filtering method for a pea protein production device with a filtering structure, based on the pea protein production device with a filtering structure according to any one of claims 1 to 4, characterized in that: The following steps are involved:

1. Open the flap (29), pour the pea slurry into the container from the feed port (7), and close the flap (29); 2. Start the switch, and the pressure sensor a (17) at the coarse filter (9) starts to monitor the pressure difference before and after the coarse filter (9). , turbidity meter a (16) monitors the concentration of suspended matter in the inlet pea slurry The pressure sensor b (19) at the fine filter (10) monitors the pressure difference before and after the fine filter (10). , turbidity meter b (18) monitors the concentration of suspended matter in the outlet pea slurry , the pressure sensor c (21) at the nanomembrane network (11) monitors the transmembrane pressure difference , the temperature sensor (20) monitors the friction heat ; 3. Based on the historical database, the random forest algorithm is used to explore the nonlinear relationship between vibration regulation and variables such as filter pressure difference and suspended solids concentration, and a vibration regulation prediction model is constructed; Fourth, use the test sample data to evaluate the prediction results. Assume that the test set is , where each sample Each corresponds to a set of actual values ​​of vibration frequency and vibration duration , the prediction result obtained by predicting each sample through the model If the prediction result With actual value The error is within 5%, which means the prediction result is accurate. Finally, the accuracy is calculated to be 89.3%, and the model prediction result is good; 5. The monitored , , , , , The six parameters are used as an input layer and input into a vibration adjustment prediction model in a control panel (30). The prediction model outputs vibration frequency and vibration duration indicators. At this time, vibrator a (12), vibrator b (13) and vibrator c (14) are activated according to the output indicators to work; 6. The vibration regulation prediction model dynamically adjusts the vibration frequency and vibration duration of each vibrator according to the real-time monitoring data until the filtering is completed.

Images