An intelligent algorithm-based full-automatic sand moisture content online detection method

Abstract

The application discloses a kind of full-automatic sandstone moisture content online detection method based on intelligent algorithm, comprising the following steps: S1, determine the multiple key features of the influence sandstone moisture content value;S2, the current value of each key feature is collected by acquisition equipment;S3, the moisture content preliminary prediction value of the natural sand for current production concrete is obtained by deep extreme limit learning machine;S4, the moisture content preliminary prediction value obtained in step S3 is corrected, and the moisture content correction prediction value is obtained;S5, the moisture content of the natural sand for current production concrete is measured by moisture content detection sensor to obtain the moisture content determination value;S6, the moisture content correction prediction value obtained in step S4 is fitted with the moisture content determination value obtained in step S5, and the moisture content determination value is obtained.The application adopts multiple intelligent algorithms to correct the moisture content determination value determined by sensor, and effectively improves the measurement accuracy of sandstone moisture content online detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of sand and gravel moisture content online detection method. More specifically, the present application relates to a kind of full-automatic sand and gravel moisture content online detection method based on intelligent algorithm. BACKGROUND

[0002] In the process of transformation and upgrading of concrete industry digitalization and digital industrialization, the underlying sensing and online detection technology of data acquisition is still in the initial stage. On the one hand, due to the late start of the research, the basic research and development of the sensor is relatively lagging behind. At present, the data processing and process improvement are mainly based on related foreign sensors. With the fluctuation of raw material quality, the rise of labor cost and the improvement of quality control requirements, the method of online detection and automatic acquisition of key data of quality in the whole process of concrete production needs to be researched and applied. Among them, the detection of concrete sand and gravel moisture content is crucial to the production quality control of concrete production mixing station. In the process of concrete production, affected by the cleaning process of sand manufacturers and the rainwater accumulation of on-site sand storage, the moisture content of sand used in each production process has certain fluctuations. Therefore, the moisture content of sand used in each production process needs to be accurately tested to produce concrete that meets the quality requirements.

[0003] At present, the sand and gravel moisture content test on the construction site mainly adopts manual sampling and drying method. This method has long test time, poor sampling representativeness and test result lag, which is difficult to meet the requirements of real-time adjustment of on-site concrete mixing ratio parameters and cannot realize effective control of the stability of concrete mixture. The real-time detection method for sand and gravel moisture content test includes resistance method, capacitance method, infrared method and microwave method. The resistance method measures the moisture content by using the conductivity of water in wet sand. Although the measuring instrument is relatively low in price, it is easily affected by environmental temperature and salt ion concentration in water, resulting in low measurement accuracy. The capacitance method measures the moisture content by using capacitance. The principle is simple and the instrument is low in price. However, the capacitance is greatly affected by temperature and salt ions in water, so the measurement accuracy is also low. Although the infrared method can realize non-contact online measurement, it is easily affected by the color of the material, and the measurement result is difficult to reflect the true moisture content inside the material. Moreover, the instrument is high in cost and difficult to maintain. The microwave method can measure the moisture content in contact or non-contact online mode. Moreover, the microwave has strong penetration, and after adding filter and temperature compensation circuit, the influence of environmental factors is significantly reduced, and the instrument is lower in cost than infrared method. However, the microwave method for measuring moisture content also has some problems, such as calibration and test environment of measuring instrument. The research on microwave method is mostly to establish a moisture content prediction model under single frequency condition, without fully considering the influence of key factors such as sand particle morphology, particle size distribution, clay content and installation parameters on the detection result.

[0004] The microwave method is mainly based on the principle of microwave moisture measurement, and the moisture of the material is measured by measuring the effective dielectric constant of the material. The existing detection sand moisture content sensor commonly has Hydro-Probe II type digital microwave sensor produced by Hydronix company and industry 4.0 probe produced by Rudolph company. Among them, the reading of the moisture sensor is linear for most materials, and the linearity of the moisture sensor must be calibrated before the test. Figure 1 The sensor calibration diagram is shown in the horizontal and vertical coordinates of the sensor output and the sand standard moisture content value measured by the laboratory drying. Taking the calibration curve calibrated by collecting two points as an example, for the sand sample with a moisture content of about 3%, two groups of unconverted values and moisture content values are measured. Assuming that the actual calibration curve and the theoretical calibration curve have an error of 0.3% in the moisture content value at a certain point when calibrating. When the measured data is too concentrated, other points cannot compensate for the error, so the error will be amplified proportionally in the high and low moisture content regions, as shown in Figure 2 (a); on the contrary, if the moisture content range of the material is large and the data in the high moisture content range also has an error of 0.3%, the error of 0.3% in other moisture content regions will be proportionally reduced, so that the influence of the error can be effectively controlled, as shown in Figure 2 (b). Therefore, the calibration method will produce a large error, which is not conducive to the quality control of the on-site concrete production. In order to ensure the real-time and accuracy of the moisture content detection of natural sand for concrete mixing station production, the humidity value collected by the microwave sensor needs to be corrected to improve the reliability of the online detection of sand moisture content. SUMMARY

[0005] An object of the present application is to solve at least the above problems and provide at least the advantages to be described later.

[0006] In order to achieve these objects and other advantages according to the present application, a full-automatic sand moisture content online detection method based on intelligent algorithm is provided, comprising the following steps:

[0007] S1, determining a plurality of key characteristics affecting the sand moisture content value;

[0008] S2, collecting the current values of each of the key characteristics corresponding to the natural sand for the currently produced concrete by the collecting device;

[0009] S3, taking the current values of each of the key characteristics collected in step S2 as input, and obtaining the preliminary prediction value of the moisture content of the natural sand for the currently produced concrete by a deep extreme learning machine;

[0010] S4, correcting the water content preliminary prediction value obtained in step S3 by a deep gated recurrent unit to obtain a water content corrected prediction value;

[0011] S5, measuring the water content of the natural sand for the currently produced concrete by a water content detection sensor to obtain a water content measured value;

[0012] S6, fitting the water content corrected prediction value obtained in step S4 and the water content measured value obtained in step S5 to obtain a water content determined value of the natural sand for the currently produced concrete.

[0013] Preferably, the following steps are further included:

[0014] S7, adjusting the water consumption of the currently produced concrete according to the water content determined value to make the water content determined value of the currently produced concrete meet the design requirements;

[0015] S8, cleaning the water content detection sensor after the completion of the production of the current batch of concrete;

[0016] S9, repeating steps S1-S8 to produce the next batch of concrete.

[0017] Preferably, in step S1, the water content corresponding to different natural sands and the characteristic values affecting the water content value are taken as input data, the characteristics affecting the water content value are screened by a grey correlation degree analysis algorithm, the characteristics are sorted according to the correlation degree, and multiple characteristics with large correlation degrees are selected as key characteristics.

[0018] Preferably, step S3 further includes optimizing the parameters of the deep extreme learning machine by a cuckoo algorithm, and step S3 specifically includes the following steps:

[0019] S3-1, obtaining historical data as a training set, the historical data including the water content corresponding to n natural sand samples and the corresponding key characteristic values;

[0020] S3-2, establishing a deep extreme learning machine sandstone water content prediction model;

[0021] S3-3, optimizing the weights and thresholds of the deep extreme learning machine sandstone water content prediction model by a cuckoo optimization algorithm to obtain a CS-DELM sandstone water content prediction model;

[0022] S3-4, training the CS-DELM sandstone water content prediction model by using the sample set;

[0023] S3-5, inputting the current values of the key characteristics collected in step S2 into the CS-DELM sandstone water content prediction model trained in step S3-4 to obtain the water content preliminary prediction value.

[0024] Preferably, step S4 specifically comprises the following steps:

[0025] S4-1, establishing a GRU moisture content prediction model, the output of which is the prediction error value of the CS-DELM sand moisture content prediction model;

[0026] S4-2, training the GRU moisture content prediction model with the sample set;

[0027] S4-3, inputting the current value of each of the key features collected in step S2 into the GRU moisture content prediction model to obtain the prediction error value of the preliminary moisture content prediction value;

[0028] S4-4, superimposing the prediction error value of the preliminary moisture content prediction value and the preliminary moisture content prediction value to obtain the corrected moisture content prediction value.

[0029] Preferably, in step S6, the corrected moisture content prediction value and the moisture content measured value are fitted by using a random forest algorithm.

[0030] Preferably, in step S8, a cleaning device is used to clean the moisture content detection sensor; the cleaning device is arranged on a belt conveyor used to transport natural sand for the currently produced concrete, and the cleaning device comprises:

[0031] a bracket fixedly arranged on the belt conveyor and used to mount the moisture content detection sensor, the measuring end of the moisture content detection sensor facing the belt conveyor;

[0032] a lifting cylinder vertically fixedly arranged on the bracket, and the movable end of the lifting cylinder facing the belt conveyor;

[0033] a mounting plate fixedly connected with the movable end of the lifting cylinder;

[0034] a plurality of nozzles spaced apart in the vertical direction on the mounting plate, the nozzles spraying pulse air flow to the moisture content detection sensor to clean the moisture content detection sensor.

[0035] Preferably, the mounting plate comprises two mutually perpendicular mounting surfaces, a plurality of nozzles are spaced apart in the vertical direction on each of the mounting surfaces, and the nozzles on the two mounting surfaces are staggered in the vertical direction.

[0036] Preferably, the included angle between the pulse air flows sprayed by the nozzles on the two mounting surfaces is 120°-150°.

[0037] The present application at least has the following beneficial effects:

[0038] 1. The full-automatic sand moisture content online detection method based on intelligent algorithm provided in the application, by screening the key features affecting the sand moisture content value, cooperating with various intelligent algorithms to obtain the moisture content correction prediction value considering various influencing factors; under the condition of not changing the original calibration operation condition of the sensor, combining the moisture content determination value determined by the sensor to obtain the moisture content determination value, the measurement precision of the sand moisture content online detection is effectively improved.

[0039] 2. The cleaning device provided in the application, by high-pressure pulse air knife to clean the impurities on the surface of the sensor, while efficiently cleaning the impurities attached to the sensor, without bringing other substances affecting the detection precision of the sensor and corroding the sensor.

[0040] 3. The application combines the multi-factor intelligent correction algorithm and the cleaning device, and by the technical measures from the software and hardware two aspects, the tested moisture content accuracy-average error is improved from 1.2% to 0.5% or less.

[0041] Other advantages, objects and features of the application will be partly embodied in the following description, and partly understood by those skilled in the art through research and practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 is a schematic diagram of the calibration curve of Hydro-Probe II type digital microwave sensor;

[0043] Figure 2 is a schematic diagram of the calibration curve of Hydro-Probe II type digital microwave sensor under different moisture content data concentration;

[0044] Figure 3 is a schematic diagram of the structure of the cleaning device provided in the application;

[0045] Figure 4 is another schematic diagram of the structure of the mounting plate provided in the application;

[0046] Figure 5 is a top view of the mounting plate provided in the application; Figure 4 DETAILED DESCRIPTION

[0047] The application will be further described in detail below with reference to the drawings, so that those skilled in the art can implement the application according to the description and drawings.

[0048] ​It should be noted that the experimental methods described in the following embodiments are conventional methods, and the reagents and materials are commercially available unless otherwise specified. In the description of the present application, the terms "lateral", "longitudinal", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. indicate the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0049] The present application provides a kind of full-automatic sand moisture content online detection method based on intelligent algorithm, comprising the following steps:

[0050] S1, determine a plurality of key features affecting the sand moisture content value;

[0051] S2, the current value of each of the key features corresponding to the natural sand for current production of concrete is collected by acquisition equipment;

[0052] S3, the current value of each of the key features collected in step S2 is used as input, and the moisture content preliminary prediction value of the natural sand for current production of concrete is obtained by deep extreme learning machine;

[0053] S4, the moisture content preliminary prediction value obtained in step S3 is corrected by deep gate recurrent unit to obtain the moisture content correction prediction value;

[0054] S5, the moisture content of the natural sand for current production of concrete is measured by moisture content detection sensor to obtain the moisture content determination value;

[0055] S6, the moisture content correction prediction value obtained in step S4 is fitted with the moisture content determination value obtained in step S5 to obtain the moisture content determination value of the natural sand for current production of concrete.

[0056] In the technical scheme, the readings of the Hydro-Probe II type digital microwave sensor are linear, and the linearity of the moisture sensor must be calibrated before the test, and the curve is calibrated by collecting information of two points, but the calibration method will cause a large error of the output sand water content. The reason for this situation is that it does not refine the factors that affect the accuracy of the sensor test during measurement. Therefore, the present application first determines a plurality of key features that actually have a significant impact on the sand water content value from a plurality of features that may affect the sand water content value. Specifically, in step S1, the known different natural sand corresponding water content and the feature value affecting the water content value are taken as input data, the feature affecting the water content value is screened through the grey correlation degree analysis algorithm, the features are sorted according to the correlation degree, and the plurality of features with large correlation degree are selected as key features. The main process is as follows:

[0057] Firstly, the water content value is taken as a reference sequence, and various influence characteristic values are taken as comparison sequences;

[0058] Secondly, the water content value and the influence characteristic value are dimensionless, and the grey correlation coefficient and the correlation degree between them are calculated;

[0059] Grey correlation coefficient:

[0060]

[0061] Correlation degree:

[0062]

[0063] Wherein, X o (k) is a reference sequence, X i (k) is a comparison sequence, p is a resolution coefficient, generally 0.5, and g i [0, 1], the closer g i is to 1, the higher the correlation between the two is. X o (k) is the water content value, and X i (i) is the value of the sensor installation angle, the fineness modulus, the silt content, the environmental temperature, the environmental relative humidity and the like. The correlation degree values of different influence characteristics are calculated by bringing the values into formula (1) and formula (2).

[0064] Finally, the correlation degrees are sorted according to the correlation degrees, and the influence characteristics with larger correlation degrees are selected as key characteristics.

[0065] After determining the plurality of key features affecting the water content value, the current values of the key features corresponding to the currently produced concrete are collected by a collection device. The collection device is a collection device commonly used for each key feature. Then, the current values of the key features corresponding to the currently produced concrete using natural sand are taken as input, and the preliminary prediction value of the water content of the currently produced concrete using natural sand is obtained by the deep extreme learning machine. Specifically, step S3 further comprises optimizing the parameters of the deep extreme learning machine by the cuckoo algorithm, and step S3 specifically comprises the following steps:

[0066] S3-1, obtain historical data as a training set, the historical data including known water content values of n natural sand samples and corresponding key feature values;

[0067] S3-2, establish a deep extreme learning machine sandstone water content prediction model;

[0068] S3-3, optimize the weight and threshold of the deep extreme learning machine sandstone water content prediction model by the cuckoo optimization algorithm to obtain a CS-DELM sandstone water content prediction model;

[0069] S3-4, train the CS-DELM sandstone water content prediction model using the sample set;

[0070] S3-5, input the current values of the key features collected in step S2 into the CS-DELM sandstone water content prediction model trained in step S3-4 to obtain the preliminary prediction value of the water content.

[0071] In step S3-3, first, the parameters are initialized, the population size, the dimension of the search space and the maximum number of iterations in the cuckoo algorithm are set, the position of the bird nest is randomly set, and the root mean square error (MSE) is taken as the fitness function; secondly, the fitness function value of each bird nest position is calculated, and the optimal bird nest position is recorded; then, the update probability P is set, the random number r and P are compared to determine whether the bird nest position needs to be updated, all bird nest positions are updated according to the Levy flight, and the position update from the i th generation to the i+1 th generation is:

[0072]

[0073] μ~N(0,σ 2 ), γ~N(0, 1) (4)

[0074] wherein, is the update step, x(i) is the current solution, x best is the current optimal solution, and σ is the gamma function. Finally, the positions of all bird nests are updated, and the optimal solution bird nest position is updated and recorded, and the iteration is repeated.

[0075] After obtaining the preliminary prediction value of the moisture content, the preliminary prediction value of the moisture content is error-corrected by a deep gated recurrent unit (GRU), and step S4 specifically comprises the following steps:

[0076] S4-1, a GRU moisture content prediction model is established, and the output of the GRU moisture content prediction model is a prediction error value of the CS-DELM sandstone moisture content prediction model;

[0077] S4-2, the GRU moisture content prediction model is trained by using the sample set;

[0078] S4-3, the current value of each of the key features collected in step S2 is input into the GRU moisture content prediction model, and a prediction error value of the preliminary prediction value of the moisture content is obtained;

[0079] S4-4, the prediction error value of the preliminary prediction value of the moisture content is superimposed on the preliminary prediction value of the moisture content, and a corrected prediction value of the moisture content is obtained.

[0080] The update gate in the deep gated recurrent unit determines which information at the last time is retained to the current time, and the reset gate determines the degree of discarding information. The update gate and the reset gate are defined as:

[0081] z t =σ(W Z X t +U Z h t-1 +b Z ) (5)

[0082] r t =σ(W r X t +U r h t-1 +b r ) (6)

[0083] wherein z t is the update gate, r t is the reset gate, X t represents the input at the current time, W Z and U Z are weight matrices of the update gate, W r and U r are weight matrices of the reset gate, b Z and b r are bias vectors, and σ is a sigmoid activation function.

[0084] The output at the current time is defined as:

[0085]

[0086] wherein, tanh represents an activation function, h t-1 is the output of the previous time.

[0087] The error value of the moisture content preliminary prediction value obtained by the CS-DELM sand moisture content prediction model through the deep gated recurrent unit (GRU) achieves the purpose of correcting the moisture content preliminary prediction value.

[0088] Then, the moisture content of the natural sand for the currently produced concrete is measured by a moisture content detection sensor to obtain a moisture content determination value of the natural sand for the currently produced concrete; in this embodiment, the detection is mainly performed by a Hydro-Probe II type digital microwave sensor, and before detection, the sensor is calibrated once by using a test software attached thereto, and the moisture content value obtained by the moisture content detection sensor is the moisture content determination value.

[0089] Then, the moisture content determination value of the natural sand for the currently produced concrete is obtained by fitting the moisture content corrected prediction value obtained in step S4 and the moisture content determination value obtained in step S5. In step S6, the random forest algorithm is used to fit the moisture content corrected prediction value and the moisture content determination value. By using the nonlinear mapping capability of the random forest (RF) algorithm, the weight coefficient is adjusted, the prediction value of the moisture content is fitted with the sensor determination value, and the final moisture content value is obtained.

[0090] Y = αY i + βY j (8)

[0091] wherein, Y is the final moisture content value, α and β are weight coefficients, Y i is the prediction value of the moisture content, and Y j is the sensor determination value.

[0092] Then, the water quantity used for the currently produced concrete is adjusted according to the moisture content determination value, so that the quality of the currently produced concrete meets the design requirements. After the production of the current batch of concrete is completed, the moisture content detection sensor is cleaned. Steps S1-S8 are repeated to produce the next batch of concrete.

[0093] In another embodiment, in step S8, a cleaning device is used to clean the moisture content detection sensor; the cleaning device is arranged on a belt conveyor used to transport the currently produced concrete, and with reference to Figure 3 , the cleaning device comprises:

[0094] a support 1 fixedly arranged on the belt conveyor and used to mount the moisture content detection sensor 3, wherein a measurement end of the moisture content detection sensor 3 faces the belt conveyor;

[0095] a lifting cylinder 2 vertically fixedly arranged on the support 1, and a movable end of the lifting cylinder 2 facing the belt conveyor;

[0096] a mounting plate 5 fixedly connected with the movable end of the lifting cylinder 2;

[0097] a plurality of nozzles 4 arranged in a vertical direction on the mounting plate 5, and the nozzles 4 spraying pulse air flow to the water content detection sensor 3 to clean the water content detection sensor 3.

[0098] In this technical scheme, the support 1 can be arranged as a portal support arranged in a transverse direction of the belt conveyor and fixedly connected with a rack of the belt conveyor. Considering that water, sand, mud and other sundries can be attached to the surface of the sensor during the working process of the water content detection sensor, which greatly affects the detection accuracy of the sensor. Therefore, the cleaning device is used to clean the water content detection sensor 3 before the next batch of concrete production. When the current batch of concrete production is completed, the lifting cylinder 2 drives the mounting plate 5 to move towards the belt conveyor until all the nozzles 4 face the water content detection sensor 3, and the pulse intermittent high-pressure air flow emitted by the nozzles 4 forms an air knife to clean the sundries on the surface of the water content detection sensor 3. The lifting cylinder 2 and the nozzles 4 can be connected with a PLC controller of the concrete mixing station controlling the belt conveyor. When the belt conveyor is working normally, the lifting cylinder 2 is in a contracted state, and at this time, the mounting plate 5 does not interfere with the working of the water content detection sensor 3. When the belt conveyor stops conveying concrete, the lifting cylinder 2 is controlled by the PLC controller to move downward, and the nozzles 4 are opened to start cleaning. The nozzles 4 and the lifting cylinder 2 can use the same compressed air pump as the air source.

[0099] In another embodiment, referring to Figure 4 and Figure 5 the mounting plate 5 includes two mutually perpendicular mounting surfaces 51, and a plurality of nozzles 4 are arranged in a vertical direction on each of the mounting surfaces 51, and the nozzles 4 on the two mounting surfaces 51 are arranged in a vertical direction in a staggered manner. Further, the included angle between the pulse air flows emitted by the nozzles 4 on the two mounting surfaces 51 is 120°-150°.

[0100] In the technical scheme, the installation plate 5 is used for fixing the nozzles 4, and is bent into a 90° steel plate, that is, the installation plate 5 comprises two perpendicular installation surfaces 51, so that the installation plate 5 can well separate the sand and gravel on the belt conveyor. The nozzles 4 on the two installation surfaces 51 are arranged at an angle, so that each point on the moisture content detection sensor 3 can be hit by multi-directional pulse air flow during cleaning, the angle air flow is fully utilized, the horizontal impact force distributed on the surface of the sensor is used to break the adhesion between impurities and the sensor, and the vertical impact force distributed on the surface of the sensor blows the adhered impurities downward quickly, so that the cleaning effect is improved, and other substances affecting the detection accuracy of the sensor and corroding the sensor are not brought in.

[0101] In the production process of a certain concrete mixing station, the intelligent algorithm-based full-automatic sand moisture content online detection method is used for online detection of moisture content, and the algorithm program involved in the intelligent algorithm-based full-automatic sand moisture content online detection method is installed into a control unit of the mixing station. The control unit can be a computer system of the mixing station.

[0102] First, the moisture content corresponding to different natural sands and the characteristic values affecting the moisture content value are taken as input data, the characteristics affecting the sand moisture content value are screened through a grey correlation degree analysis algorithm, and the key characteristics obtained are:

[0103] Installation angle: the angle between the ceramic panel of the moisture content detection sensor 3 and the material conveying direction, the moisture content detection sensor 3 selects a Hydro-Probe II type digital microwave sensor;

[0104] Fineness modulus: an index representing the fineness degree and category of the particle size of natural sand, which is determined by field measurement, and the fineness modulus range of the natural sand commonly used in engineering projects is 2.2-3.2;

[0105] Silt content: the mass percentage of the part with a diameter less than 0.020 mm in the natural sand, which is determined by field measurement, and the silt content range of the natural sand commonly used in engineering projects is 0.5%-3.5%;

[0106] Ambient temperature: the ambient temperature of the sensor, which is measured and collected by a corresponding temperature sensor on site;

[0107] Relative humidity: the relative humidity of the sensor, which is measured and collected by a corresponding humidity sensor on site;

[0108] Natural sand temperature: the temperature of natural sand, measured on site by corresponding temperature sensors;

[0109] The cleaning device is installed on the belt conveyor in the production site, and the lifting cylinder 2 and each nozzle 4 in the cleaning device are connected with the PLC controller of the mixing station. The above six key characteristic values are measured on site and written into the algorithm packaging program, and the water content correction value is obtained by calculation; the water content detection sensor 3 is calibrated once according to the test software of the water content detection sensor 3, and the output of the water content detection sensor 3 is connected to the algorithm packaging program. The concrete mixing station starts production, and the water content detection sensor 3 works during the transmission of natural sand, tests the water content to obtain the water content measurement value, and transmits the water content measurement value to the algorithm packaging program to obtain the water content determination value through calculation. The control unit of the mixing station adjusts the water consumption during the concrete production process according to the water content determination value. After the transmission of natural sand for the current batch of concrete production is completed, the belt conveyor is stopped, the PLC controller starts the cleaning device, the compressed air pump is started to provide power, the lifting cylinder 2 is driven to elongate, the mounting plate 5 is driven to move downward, and at the same time, pulse airflow is sprayed from each nozzle 4 to clean the sensor. The cleaning time can be adjusted according to the interval of the belt conveyor on site, generally 15-20s; after cleaning, the production of the next batch of concrete is started, the cleaned sensor detects the water content of the currently transmitted natural sand, and the above process is repeated; when the fineness modulus and the clay content of the natural sand are found to have large fluctuations during the measurement of the key characteristics on site, the latest measurement data can be written into the algorithm packaging program through testing to improve the accuracy of the sand and stone water content detection. Table 1 shows the results of the sand and stone water content detection by the intelligent algorithm-based full-automatic sand and stone water content online detection method during the production of a concrete mixing station.

[0110] Table 1 shows the results of the sand and stone water content detection by the intelligent algorithm-based full-automatic sand and stone water content online detection method during the production of a concrete mixing station.

[0111]

[0112] Although the embodiments of the present application have been disclosed as above, they are not limited to the application listed in the specification and the embodiments, and can be fully applied to various fields suitable for the present application, and additional modifications can be easily realized by those skilled in the art, and therefore the present application is not limited to specific details and the figures shown and described herein, without departing from the general concept defined by the claims and the equivalent scope.

Claims

1. A full-automatic sand moisture content online detection method based on an intelligent algorithm, characterized in that, The method comprises the following steps: S1, determining a plurality of key features affecting the water content value of sand and gravel; in step S1, the water content corresponding to different natural sands known and the characteristic values affecting the water content value are taken as input data, the features affecting the water content value are screened through a gray correlation degree analysis algorithm, the features are sorted according to the correlation degree, and a plurality of features with large correlation degrees are selected as key features, including a sensor installation angle, a fineness modulus, a silt content, an environmental temperature, an environmental relative humidity, and a natural sand temperature; S2, collecting current values of the key features corresponding to the natural sand for the current production of concrete through a collection device; S3, taking the current values of the key features collected in step S2 as input, and obtaining a preliminary prediction value of the water content of the natural sand for the current production of concrete through a deep extreme learning machine; step S3 further comprises optimizing parameters of the deep extreme learning machine through a cuckoo algorithm, and step S3 specifically comprises the following steps: S3-1, obtaining historical data as a training set, wherein the historical data comprises water contents corresponding to n natural sand samples known and corresponding key feature values; S3-2, establishing a deep extreme learning machine sand and gravel water content prediction model; S3-3, optimizing weights and thresholds of the deep extreme learning machine sand and gravel water content prediction model through a cuckoo optimization algorithm to obtain a CS-DELM sand and gravel water content prediction model; S3-4, training the CS-DELM sand and gravel water content prediction model by using the sample set; S3-5, inputting the current values of the key features collected in step S2 into the CS-DELM sand and gravel water content prediction model trained in step S3-4 to obtain the preliminary prediction value of the water content; S4, correcting the preliminary prediction value of the water content obtained in step S3 through a deep gated recurrent unit to obtain a corrected prediction value of the water content; S5, measuring the water content of the natural sand for the current production of concrete through a water content detection sensor to obtain a water content measurement value; S6, fitting the corrected prediction value of the water content obtained in step S4 with the water content measurement value obtained in step S5 to obtain a determined value of the water content of the natural sand for the current production of concrete.

2. The full-automatic sand moisture content online detection method based on an intelligent algorithm according to claim 1, characterized in that, Further comprising the following steps: S7, adjusting the water quantity of the current production of concrete according to the determined value of the water content, so that the determined value of the water content of the current production of concrete meets the design requirements; S8, cleaning the water content detection sensor after the production of the current batch of concrete is completed; S9, repeating steps S1-S8 to produce the next batch of concrete.

3. The fully automatic sand moisture content online detection method based on intelligent algorithm according to claim 1, characterized in that, Step S4 specifically comprises the following steps: S4-1, establishing a GRU water content prediction model, wherein the output of the GRU water content prediction model is a prediction error value of the CS-DELM sand and gravel water content prediction model; S4-2, training the GRU water content prediction model by using the sample set; S4-3, inputting the current values of the key features collected in step S2 into the GRU water content prediction model to obtain a prediction error value of the preliminary prediction value of the water content; S4-4, superimposing the prediction error value of the water content preliminary prediction value and the water content preliminary prediction value to obtain the water content correction prediction value.

4. The fully automatic sand moisture content online detection method based on intelligent algorithm according to claim 1, characterized in that, In step S6, the water content correction prediction value and the water content measured value are fitted by using the random forest algorithm.

5. The fully automatic sand moisture content online detection method based on intelligent algorithm according to claim 2, characterized in that, In step S8, the water content detection sensor is cleaned by using a cleaning device. The cleaning device is arranged on a belt conveyor for conveying natural sand for currently produced concrete, and the cleaning device comprises: a support fixedly arranged on the belt conveyor and used for mounting the water content detection sensor, a measuring end of the water content detection sensor facing the belt conveyor; a lifting cylinder vertically fixedly arranged on the support, and a movable end of the lifting cylinder facing the belt conveyor; a mounting plate fixedly connected with the movable end of the lifting cylinder; a plurality of nozzles arranged on the mounting plate in a vertical direction, the nozzles spraying pulse air flow to the water content detection sensor to clean the water content detection sensor.

6. The fully automatic sand moisture content online detection method based on intelligent algorithm according to claim 5, characterized in that, The mounting plate comprises two mutually perpendicular mounting surfaces, and a plurality of nozzles are arranged on each mounting surface in a vertical direction, and the nozzles on the two mounting surfaces are arranged in a vertical direction in a staggered manner.

7. The fully automatic sand moisture content online detection method based on intelligent algorithm according to claim 6, characterized in that, An included angle between the pulse air flows sprayed by the nozzles on the two mounting surfaces is 120°-150°.

Images