Concrete slump control system and method combining AI monitoring and ERP

Abstract

The invention provides a concrete slump control system and method combining AI monitoring and ERP, and relates to the technical field of building materials, the system comprises: a proportioning parameter acquisition module for acquiring concrete proportioning design parameters; the path determining module is used for determining a transportation path according to the transportation vehicle-mounted end and the mixing station end; the attenuation prediction module is used for acquiring vehicle-mounted operation condition data of the vehicle-mounted end to determine a transportation attenuation index; the output prediction module is used for collecting a stirring image and inputting the stirring image into a double-flow neural network to obtain predicted output slump; and the compensation control module is used for transmitting the transportation attenuation index to the stirring station end to carry out compensation control on the predicted slump of the machine. The technical problem that the slump does not meet the design requirement when the concrete reaches a construction site due to the fact that the influence of the transportation process on the slump of the concrete cannot be comprehensively considered in the prior art is solved, and the technical effects of accurately controlling the slump of the concrete and improving the quality stability of the concrete are achieved.

Description

Technical Field

[0001] The present application relates to the field of building materials technology, and specifically to a concrete slump control system and method combining AI monitoring and ERP. Background Art

[0002] Concrete slump, a key indicator of concrete's workability and fluidity, has a significant impact on the final construction quality. Current concrete slump control methods largely rely on preset parameters at the mixing plant and operator experience. During transportation, losses are often estimated based on fixed time or distance, and adjustments are made through simple methods such as adding admixtures. These methods fail to fully consider the impact of various actual working conditions, such as the transportation route and the vehicle's operating status. This results in deviations in the concrete's slump upon arrival at the construction site, making it difficult to meet design requirements and impacting construction quality and efficiency. Summary of the Invention

[0003] This application provides a concrete slump control system and method that combines AI monitoring and ERP, which solves the technical problem that the existing technology is unable to comprehensively consider the impact of multiple factors such as transportation routes and vehicle operating conditions on the slump of concrete, resulting in the slump of concrete not meeting the design requirements when it arrives at the construction site. It achieves the technical effect of accurately controlling the slump of concrete and improving the stability of concrete quality.

[0004] In view of the above problems, on the one hand, the present application provides a concrete slump control system that combines AI monitoring and ERP, and the system includes: a mix parameter acquisition module, which is used to access the ERP work order system in the cloud platform to collect the concrete mix design parameters of the target work order; a path determination module, which is used to determine the transport vehicle end and the mixing station end of the target work order, and determine the transport path according to the transport vehicle end and the mixing station end; an attenuation prediction module, which is used to collect the on-board operating condition data corresponding to the transport vehicle end according to the transport path, and predict the concrete attenuation according to the on-board operating condition data to obtain the transport attenuation index of the transport path; an outbound prediction module, which is used to collect mixing images according to the mixing station end, and input the mixing images into a dual-stream neural network to predict the concrete outbound slump to obtain the predicted outbound slump; a compensation control module, which is used to transmit the transportation attenuation index to the mixing station end to perform compensation control on the predicted outbound slump until the target slump in the concrete mix design parameters is met.

[0005] On the other hand, the present application also provides a concrete slump control method combining AI monitoring and ERP, the method including: accessing the ERP work order system in the cloud platform to collect the concrete mix design parameters of the target work order; determining the transport vehicle end and the mixing station end of the target work order, and determining the transportation route according to the transport vehicle end and the mixing station end; collecting the on-board operating condition data corresponding to the transport vehicle end according to the transportation route, and performing concrete attenuation prediction based on the on-board operating condition data to obtain the transportation attenuation index of the transportation path; collecting mixing images according to the mixing station end, and inputting the mixing images into a dual-stream neural network to predict the concrete out-of-machine slump to obtain the predicted out-of-machine slump; transmitting the transportation attenuation index to the mixing station end to compensate for the predicted out-of-machine slump until the target slump in the concrete mix design parameters is met.

[0006] One or more technical solutions provided in this application have at least the following beneficial effects: The mix parameter acquisition module accesses the ERP work order system from the cloud platform to collect concrete mix design parameters for the target work order, providing foundational data for the entire solution and ensuring a clear target basis for subsequent control. The route determination module identifies the transport vehicle and mixing plant for the target work order and determines the transport route, laying the foundation for subsequent concrete condition monitoring and prediction during specific transport processes, enabling focus on specific transport links. The attenuation prediction module collects on-board operating condition data corresponding to the transport vehicle based on the transport route and predicts concrete attenuation to obtain a transport attenuation index. This allows for early estimation of concrete slump loss during transport, providing critical predictive data for subsequent compensation control. The delivery prediction module captures mixing images at the mixing plant and inputs these images into a dual neural network to predict the delivery slump of concrete. This accurately determines the actual slump of concrete at the source, enabling correlation analysis with attenuation during subsequent transport. The compensation control module transmits the transport attenuation index to the mixing plant for compensation control of the predicted delivery slump. Through dynamic adjustment, the concrete slump is precisely controlled until the target slump is met, ensuring that concrete quality meets design requirements.

[0007] In summary, this application realizes the accurate prediction and dynamic compensation control of concrete slump during transportation. Through the synergistic effect of various modules, from collecting mix parameters to determining the transportation route, to predicting transportation attenuation and slump out of the machine, and finally performing compensation control, a complete closed-loop control system is formed, which can ensure that the slump of concrete is stable within the design requirements when it arrives at the construction site, greatly improving the quality stability of concrete, reducing construction quality problems caused by slump not meeting requirements, and improving the overall efficiency and reliability of concrete production and construction, which has a positive role in promoting the intelligent and precise development of the concrete industry.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a schematic diagram of the structure of a concrete slump control system combining AI monitoring and ERP provided in an embodiment of the present application.

[0010] Figure 2 A flow chart of a concrete slump control method combining AI monitoring and ERP provided in an embodiment of the present application.

[0011] Description of reference numerals: ratio parameter acquisition module 10 , path determination module 20 , attenuation prediction module 30 , engine output prediction module 40 , compensation control module 50 . DETAILED DESCRIPTION

[0012] The embodiments of the present application provide a concrete slump control system and method that combines AI monitoring and ERP, thereby solving the technical problem in the prior art that the slump of concrete does not meet the design requirements when it arrives at the construction site due to the inability to comprehensively consider the impact of various factors such as the transportation route and vehicle operating conditions on the slump of concrete. This achieves the technical effect of accurately controlling the slump of concrete and improving the stability of concrete quality.

[0013] Example 1, as Figure 1 As shown, the embodiment of the present application provides a concrete slump control system combining AI monitoring and ERP, the system comprising: The mix parameter collection module 10 is used to access the ERP work order system in the cloud platform to collect the concrete mix design parameters of the target work order.

[0014] Specifically, the ERP work order system records the concrete mix design parameters for each batch of concrete, including the formula, pouring location, quantity, required strength grade, and slump target. The module connects the work order data in the ERP system to the cloud platform through an API or data interface. For example, data synchronization is achieved using a RESTful API or database middleware connection. It automatically reads the mix design parameters corresponding to the target work order and uses them as benchmark data for subsequent prediction and control. For example, a work order "Order No. 2025041801" is generated in the ERP system. The mix parameters are: cement 350kg / m³, water 165kg / m³, sand 630kg / m³, stone 1200kg / m³, admixture 5kg / m³, and the target slump is 180mm. This information is synchronized to the cloud platform for use by subsequent modules.

[0015] The path determination module 20 is used to determine the transport vehicle end and the mixing station end of the target work order, and determine the transport path according to the transport vehicle end and the mixing station end.

[0016] Specifically, the transport vehicle side refers to the concrete mixer truck carrying concrete. The concrete mixing plant side refers to the concrete mixing plant, typically the location where concrete leaves the factory. Based on the concrete transport vehicle and mixing plant location specified in the work order, the module uses GIS (Geographic Information System) and navigation services to determine the actual route the transport vehicle travels from the mixing plant to the construction site, known as the transport route. The module also records the route length, estimated time, and road conditions along the way. This provides a reference for transport conditions that can be used to predict subsequent changes in concrete performance, enabling more accurate predictions of concrete slump degradation.

[0017] The attenuation prediction module 30 is used to collect the vehicle-mounted operating condition data corresponding to the transport vehicle-mounted terminal according to the transport path, perform concrete attenuation prediction based on the vehicle-mounted operating condition data, and obtain the transport attenuation index of the transport path.

[0018] Specifically, the module collects operating information about the transport vehicle, such as speed, vibration, and ambient temperature, based on the transport route using OBD (on-board device), gyroscopes, accelerometers, and other devices. This information is then used to obtain onboard operating condition data. Using a pre-built transport attenuation prediction model, the module predicts concrete attenuation based on this onboard operating condition data. This model estimates the slump loss due to factors such as vibration, time, and temperature, generating a transport attenuation index (i.e., a slump loss value), such as "ΔS = 25mm." This index quantifies changes in concrete performance during transport, providing accurate data support for subsequent compensation adjustments.

[0019] The concrete slump prediction module 40 is used to collect mixing images from the mixing station end, input the mixing images into a double-stream neural network to predict the concrete slump out of the machine, and obtain the predicted concrete slump out of the machine.

[0020] Specifically, a high-definition industrial camera and vibration sensor are installed at the mixing station to capture the concrete outflow process from the discharge port and collect vibration data. This produces a mixing image, which includes a video image sequence and a vibration spectrum from multiple angles. These mixing images are then input into a pre-trained two-stream neural network to predict the slump of concrete leaving the mixing station. The two-stream neural network comprises a spatial stream, which processes images, and a temporal stream, which processes dynamically changing data. The spatial stream network of the two-stream neural network performs feature analysis on the video image sequence, while the temporal stream network performs feature analysis on the vibration spectrum, outputting image features and temporal features. These image and temporal features are then fused and predicted to produce the predicted slump of concrete leaving the mixing station.

[0021] This module can provide a relatively accurate estimate of the slump of concrete before it leaves the station, which helps to promptly detect possible problems in the mixing process and provides another important data source for subsequent compensation control, thereby improving the accuracy of the entire concrete slump control.

[0022] The compensation control module 50 is used to transmit the transport attenuation index to the mixing station to perform compensation control on the predicted slump until the target slump in the concrete mix design parameters is met.

[0023] Specifically, the module transmits the transport attenuation index obtained by the attenuation prediction module 30 to the mixing station end. At the mixing station end, based on this index and the predicted slump of the concrete out of the machine obtained by the out-of-machine prediction module 40, the feedback mechanism is used to adjust the amount of water added, extend the mixing time, or add admixtures and other operations for compensation control until the target slump is met. The control logic can be automatically executed through the PLC system or the MES system. For example, if the transport attenuation index shows that the slump of concrete will decrease by 20mm during transportation, and the predicted out-of-machine slump is 180mm and the target slump is 200mm, then it is necessary to add an appropriate amount of admixture at the mixing station end to increase the slump. This module effectively avoids insufficient slump of concrete on site through a feedforward compensation mechanism based on prediction, realizes closed-loop intelligent control of the entire process, and improves project adaptability and concrete quality consistency.

[0024] Furthermore, the transport path includes a first transport path, wherein the first transport path is a path for supplying concrete raw materials from the first transport vehicle end to the mixing station end, and the path determination module 20 includes: The first construction module is used to establish a raw material transportation attenuation model, and the raw material transportation attenuation model is obtained by training the historical operating condition data of the first transportation vehicle-mounted terminal.

[0025] The first prediction module is used to collect the first vehicle-mounted operating condition data of the first transport vehicle-mounted terminal, call the raw material transportation attenuation model to predict the concrete raw material attenuation of the first vehicle-mounted operating condition data, and output the transport raw material attenuation index.

[0026] The first compensation module is used to transmit the attenuation index of the transported raw materials to the mixing station end to perform ratio compensation control at the raw material stage.

[0027] Specifically, the first transport vehicle-mounted terminal is the transport vehicle and its related equipment responsible for supplying concrete raw materials (such as cement, sand, gravel, etc.) to the mixing station. The first transport route refers to the transport route from the raw material vehicle (such as a vehicle transporting sand, gravel, cement, etc.) to the mixing station. The historical operating condition data of the first transport vehicle-mounted terminal is obtained from the database, including information such as vehicle speed, driving time, road conditions, and initial raw material data (raw material type, moisture content) during the historical transportation process. The historical raw material receipt records corresponding to each historical operating condition record, such as the raw material moisture content, are extracted from the raw material receipt records of the mixing station. This collected historical data is constructed into a training dataset, and a supervised learning method (such as XGBoost regression and support vector regression) is used. With the historical operating condition data as input data and the historical raw material receipt records as supervised output, a raw material transportation attenuation model is trained to evaluate the loss or change of raw material properties during transportation.

[0028] During the actual transportation process, the first vehicle operating condition data is collected through sensors and other equipment installed on the first transport vehicle, and then these vehicle operating condition data are input into the constructed raw material transportation attenuation model to predict the attenuation of concrete raw materials and output the transportation raw material attenuation index, such as "sand moisture content drops by 1.5%".

[0029] The transport raw material attenuation index obtained by the first prediction module is transmitted to the mixing station. At the mixing station, the proportion of concrete raw materials is adjusted according to this index. For example, if the transport raw material attenuation index shows that the water content of sand and gravel has decreased, the amount of water used during mixing is increased accordingly to ensure the final performance of the concrete. For example, during transportation, the cement raw materials are not sealed in the vehicle, the temperature reaches 35°C, and some water evaporates. The model predicts that the water content will decrease by 0.8%. This compensation index is transmitted to the mixing station, and the mixing station control system automatically increases the originally planned water addition amount from 165kg / m³ to 170kg / m³, thereby ensuring that the slump of the concrete remains at the design target.

[0030] The above module solves the problem of imperceptible hidden quality changes during raw material transportation, realizes feedforward compensation adjustment of the pre-mixing ratio, effectively improves the input quality consistency of concrete mixture, and enhances the stability and intelligent control level of the production stage.

[0031] Furthermore, the transport path includes a second transport path, which is a path for the second transport vehicle to transport the finished concrete from the mixing station. The path determination module 20 further includes: The second construction module is used to establish a finished product transportation attenuation model, which is obtained by training the historical operating condition data of the second transportation vehicle terminal and the slump difference label before and after transportation.

[0032] The second prediction module is used to collect the second vehicle-mounted operating condition data of the second transport vehicle-mounted terminal, call the finished product transportation attenuation model to predict the concrete slump attenuation of the second vehicle-mounted operating condition data, and output the transportation slump attenuation index.

[0033] The second compensation module is used to transmit the transportation slump attenuation index to the mixing station end to perform finished product compensation control at the machine exit stage.

[0034] Specifically, the secondary transport vehicle is responsible for transporting the finished concrete produced at the mixing plant to the construction site, along with its associated equipment. The secondary transport route refers to the route taken by the finished concrete from the mixing plant to the construction site. The finished concrete transport attenuation model uses slump as its output, combined with onboard operational data, to predict slump reduction during transport.

[0035] The second construction module collects historical operating condition data from the second transport vehicle, noting the difference in concrete slump before and after each transport. This data is then supervised using a long-short-term memory network (LSTM). The historical operating condition data from the second transport vehicle serves as the input vector, and the actual slump difference before and after transport measured during historical transports serves as the regression label. The mean squared error (MSE) is used as the loss function, and the Adam optimizer is used for weight training. The model parameters are continuously optimized over multiple rounds of iterations to ensure that the predicted output is as close to the actual attenuation value as possible. This process continues until the model converges and the constructed finished product transport attenuation model is output.

[0036] When actually transporting finished concrete, the second prediction module collects the second vehicle operating condition data, such as transportation time, ambient temperature, vehicle speed, vehicle vibration amplitude, etc., inputs these data into the finished product transportation attenuation model, and predicts and outputs the transportation slump attenuation index.

[0037] The second compensation module transmits the transport slump attenuation indicator to the mixing station. At the mixing station, adjustments are made to the finished concrete about to leave the station based on this indicator. This involves appropriately increasing the slump of the concrete leaving the machine during production, or adjusting the amount of admixtures, to ensure that the slump of the concrete meets construction requirements after it is transported to the construction site, thus achieving compensation control at the finished product stage. For example, a truckload of concrete needs to be transported 30 minutes to a construction site 9 kilometers away from the mixing station. The model predicts that transportation will result in a 28mm slump drop. Therefore, the mixing station control system changes the slump of the concrete leaving the machine from the initial 180mm to 208mm in advance.

[0038] The above module realizes feedforward prediction and dynamic compensation of the slump of finished products, effectively avoiding rework or waste at the construction site and ensuring construction quality and efficiency.

[0039] Furthermore, the second vehicle-mounted operating condition data includes transportation time, vehicle vibration amplitude, acceleration change, turning frequency and ambient temperature; the finished product transportation attenuation model is used to extract transportation attenuation time series characteristics of the second vehicle-mounted operating condition data, predict slump attenuation based on the extracted transportation attenuation time series characteristics, and output a transportation slump attenuation index.

[0040] Specifically, transport time is the total time it takes for finished concrete to travel from the batching plant to the construction site, measured in minutes or seconds. Vehicle vibration amplitude is the intensity of the vehicle's vibration in all directions (especially the vertical direction) during transport, measured by an accelerometer or gyroscope. Acceleration changes indicate fluctuations in vehicle speed during transport and are primarily used to identify maneuvers such as sudden acceleration and braking. Turning frequency represents the number of turns a vehicle makes per unit time, indirectly reflecting the degree of curvature change along the transport route. Ambient temperature is the external air temperature during concrete transport and directly impacts the rate at which concrete slump evolves.

[0041] The second prediction module collects the second vehicle operating condition data in real time through sensors deployed on the transport vehicle (such as accelerometers, gyroscopes, thermometers, etc.), and performs preprocessing such as noise filtering, data synchronization and timestamp correction, and outlier removal. The data is then input into the finished product transportation attenuation model to predict the concrete slump attenuation.

[0042] The finished product transportation attenuation model adopts a structural design based on a long-short-term memory network and has strong time-series dependency modeling capabilities. After receiving the second vehicle's operating condition data, a time series analysis is first performed. Using methods such as sliding window methods, FFT, wavelet transform, and time series statistical analysis, the transportation attenuation time series characteristics from the transportation start to the end point are extracted from the original time series data. These include average vibration amplitude, vibration frequency spectrum density, number of sudden accelerations, mean turning time interval, temperature change trend slope, etc. Slump attenuation is then predicted based on the extracted transportation attenuation time series characteristics, and a transportation slump attenuation index is output. For example, in the model prediction, if a transportation batch experiences significant vibration between the 10th and 15th minute period, combined with the temperature increase trend at the time, the model predicts that a slump loss of 10 mm will occur during this period. This prediction result will be used at the mixing station for precise compensation control of the outgoing slump.

[0043] By introducing the second vehicle operating condition data and extracting the transport attenuation time series characteristics, accurate modeling of the complex dynamic conditions of the concrete transportation process is achieved, and the model can maintain adaptability under different climate, road and vehicle conditions.

[0044] Furthermore, the stirring image includes a video image sequence and a vibration spectrum diagram at multiple angles, and the machine failure prediction module 40 is further configured to: The video image sequence and the vibration spectrum diagram are input into the dual-stream neural network, feature analysis is performed on the video image sequence according to the spatial stream network of the dual-stream neural network, and feature analysis is performed on the vibration spectrum diagram according to the temporal stream network of the dual-stream neural network, and image features and time features are output; feature fusion prediction is performed on the image features and the time features to obtain the predicted machine slump.

[0045] Specifically, in order to accurately predict the slump of concrete leaving the machine, the mixing image includes a video image sequence from multiple angles and a corresponding vibration spectrum diagram. The video image is used to intuitively reflect the appearance state of the concrete mixing process, such as fluidity, boundary adhesion, vortex formation, etc. The vibration spectrum diagram is a frequency domain image generated by Fourier transforming the vibration data collected by the sensor during the mixing process, reflecting the dynamic energy changes during the operation of the mixing equipment.

[0046] The machine exit prediction module 40 employs a dual-stream neural network architecture, consisting of a spatial network and a temporal network. The spatial network uses a multi-layer convolutional neural network (such as ResNet50) to process video image sequences from multiple angles, extracting spatial features between image frames, such as concrete color distribution, particle aggregation, and edge diffusion. The temporal network inputs the vibration spectrum formed after Fourier transformation and uses a one-dimensional or two-dimensional convolutional neural network (such as a 1D-CNN or Inception module) to extract dynamic features such as spectral intensity distribution, energy band peaks, and dominant frequency variation trends. These spatial and temporal features are then fused using methods such as fully connected concatenation, weighted averaging, and attention mechanism fusion. The fused feature vector is input into a regression prediction layer (such as an MLP or Transformer decoder), which outputs the predicted machine exit slump value. For example, if the image sequence shows a relatively smooth concrete surface and the vibration spectrum has strong high-frequency components, the model predicts that the current concrete slump is at a low level (e.g., 120mm). If the image shows a lot of concrete rolling and the low-frequency components of the spectrum are prominent, the predicted slump is 180mm. This dual-stream neural network model enables the integrated perception and analysis of the visual and dynamic states of the concrete mixing process, significantly improving the prediction accuracy of the slump at the mixer and providing a reliable basis for subsequent compensation control.

[0047] Furthermore, the compensation control module 50 includes: The detection module is used to detect the real-time slump of concrete at the second transport vehicle end.

[0048] The judgment module is used to judge whether the real-time slump of the concrete meets the target slump in the concrete mix design parameters. If the real-time slump of the concrete does not meet the target slump in the concrete mix design parameters, a navigation instruction is sent to the second transport vehicle-mounted terminal to control the second transport vehicle-mounted terminal to return to the mixing station terminal.

[0049] The third compensation module is used to control the mixing station to perform concrete slump compensation according to the transportation attenuation index.

[0050] Specifically, to ensure timely remediation even if concrete slump anomalies occur during transportation, the compensation control module 50 includes a detection module, a judgment module, and a third compensation module. The detection module, deployed on the second transport vehicle, collects concrete slump parameters in real time and uploads them to the cloud platform. For example, the current slump data is acquired using an onboard automatic slump measurement device (e.g., a drop distance measurement device using image processing) and uploaded to the cloud platform.

[0051] After receiving the real-time slump data, the judgment module compares the real-time slump value with the target slump in the concrete mix design parameters of the ERP design work order, and determines whether the real-time slump of the concrete meets the target slump, that is, whether the deviation between the two is within the allowable range. If it exceeds the allowable deviation (such as ±10mm), the return compensation control mechanism is triggered, and the on-board communication system is used to send navigation instructions to the second transport vehicle-mounted terminal to control the second transport vehicle-mounted terminal to return to the mixing station end.

[0052] When the second transport vehicle returns to the mixing station, the third compensation module automatically adjusts the proportion of ingredients based on the previously recorded transportation attenuation index (for example, a loss of 20mm slump), such as adding an appropriate amount of water reducer, controlling the amount of water added, etc., re-mixing or local addition adjustments, thereby compensating for the slump loss and ensuring that the concrete meets the original design requirements when it is discharged again.

[0053] Through the above modules, concrete that does not meet the requirements can be effectively adjusted to ensure that even if performance fluctuations occur during dynamic transportation, closed-loop corrections can be achieved, thereby improving the stability of concrete quality and ensuring that the performance of the concrete meets the requirements of the engineering design.

[0054] Furthermore, the navigation instruction is sent via the cloud platform, which includes multiple mixing station terminals and multiple transport vehicle terminals, and the judgment module includes: A position acquisition module is used to obtain the real-time position of the second transport vehicle terminal based on the cloud platform if the real-time slump of the concrete does not meet the target slump in the concrete mix design parameters.

[0055] The neighboring mixing station end identification module is used to detect the relative distance between the second transport vehicle end and the multiple mixing station ends based on the real-time position, and obtain the neighboring mixing station end, where the neighboring mixing station end is the mixing station end with the smallest relative distance.

[0056] An instruction generation module is used to generate the navigation instruction based on the position of the adjacent mixing station end.

[0057] Specifically, when the real-time slump of concrete does not meet the target slump in the concrete mix design parameters, the position acquisition module obtains the GPS positioning information of the current second transport vehicle terminal through the cloud platform, identifies the real-time position of the second transport vehicle terminal (expressed in coordinate form), and feeds it back to the neighboring mixing station terminal identification module.

[0058] The neighboring mixing station identification module obtains the location information of multiple mixing stations from the cloud platform. This location information also exists in the form of geographic coordinates. For each mixing station, the Euclidean distance formula is used to calculate the distance between it and the second transport vehicle. The relative distances of multiple mixing stations are compared, and the nearest mixing station is selected as the compensation site.

[0059] The instruction generation module generates navigation instructions according to the position of the neighboring mixing station, and sends them to the on-board system of the transport vehicle through the cloud platform to guide it to return.

[0060] Furthermore, the system described in the embodiment of the present application also includes a training module, which is used to: Initialize the dual-stream neural network; collect multiple mixing control parameter sample groups from the multiple mixing station ends, perform transfer learning on the dual-stream neural network according to the multiple mixing control parameter sample groups, and output multiple dual-stream neural networks; extract multiple network parameters of the multiple dual-stream neural networks, perform fusion learning on the multiple network parameters, and obtain updated multiple dual-stream neural networks; download the updated multiple dual-stream neural networks to the multiple mixing station ends based on the cloud platform.

[0061] Specifically, the training module uses a deep learning framework (such as TensorFlow or PyTorch) to create a two-stream neural network, initialize the weight parameters, and prepare for subsequent training.

[0062] To account for environmental differences across multiple mixing stations (such as equipment model and mixer structure), mixing image sequences and vibration spectrum data, along with corresponding actual slump labels, are collected from the control systems or databases of these stations to construct multiple mixing control parameter sample groups, such as sample group A1 for station A and sample group B1 for station B. For each mixing control parameter sample group, transfer learning is performed on an initialized two-stream neural network, using the mixing image sequence and vibration spectrum data as input and the corresponding mixing result (actual outgoing slump) as label data. During the transfer learning process, for each sample group at each mixing station, some layers of the spatial and temporal networks (such as the feature extraction layers, which typically learn general image and temporal features) are frozen, and only the subsequent fully connected layers are retrained to adapt them to the concrete slump prediction task at that station. In this way, the initial two-stream neural network is personalized using data from different mixing stations, resulting in multiple two-stream neural networks tailored to different mixing stations.

[0063] Extract parameter weights (such as convolutional layer weights and LSTM gating weights) from each site's two-stream neural network model. Use methods such as weighted averaging, attention mechanism fusion, or meta-learning to perform ensemble learning on these parameters to obtain a more generalized set of network parameters. These parameters are then used to update the two-stream neural network model at each site, resulting in multiple updated two-stream neural networks. For example, by evaluating the predictive performance of each network on a validation set, calculating its weight coefficient, and then performing a weighted average of the network parameters according to the weights to obtain the fused network parameters. Using these fused network parameters, the original multiple two-stream neural networks are updated, allowing them to comprehensively consider the characteristics of multiple mixing stations and improve the model's generalization capabilities.

[0064] Based on the cloud platform, the updated dual-stream neural networks are downloaded to multiple concrete mixing stations. The cloud platform transmits the network model files to the equipment at each concrete mixing station through a data transmission protocol, enabling local deployment of the model.

[0065] The training module trains and optimizes the dual-stream neural network through transfer learning and fusion learning, and distributes the updated network model to each mixing station, so that the dual-stream neural network can adapt to the concrete production characteristics of different mixing stations and improve the prediction performance.

[0066] Furthermore, the compensation control module 50 is further configured to: After obtaining the predicted slump out of the machine, the actual slump out of the machine at the mixing station end is obtained, the error of the actual slump out of the machine is compared with the predicted slump out of the machine, and the prediction residual is calculated; the prediction residual is used as the error reverse input signal to optimize the parameters of the dual-stream neural network.

[0067] Specifically, the prediction residual refers to the difference between the actual outgoing slump and the model's predicted value, reflecting the model's prediction accuracy. After obtaining the predicted outgoing slump, the compensation control module 50 uses the slump detection equipment installed at the mixing station's outlet to obtain the actual outgoing slump. The actual value is compared with the predicted value, and the difference between the two is calculated, which is the prediction residual.

[0068] The prediction residual is used as the error input signal, and the parameters of the two-stream neural network are optimized using the backpropagation algorithm. The backpropagation algorithm calculates the gradient of the loss function (such as the mean squared error) with respect to the weights and biases of each network layer. Based on this gradient information, the network parameters are adjusted to reduce the prediction residual. For example, gradient descent is used to update the network parameters at a specific learning rate. The update formula is: new weight = old weight - learning rate × gradient. This process is repeated repeatedly. As training data accumulates and the number of iterations increases, the prediction accuracy of the two-stream neural network gradually improves.

[0069] The compensation control module 50 uses the prediction residual to further optimize the dual-stream neural network to form a closed-loop feedback control, continuously improve the prediction accuracy of the model, and enhance its adaptability and generalization ability to different working conditions.

[0070] In summary, the concrete slump control system combining AI monitoring and ERP provided in the embodiments of the present application has the following beneficial effects: The mix parameter acquisition module 10 accesses the ERP work order system from the cloud platform to collect the concrete mix design parameters of the target work order, providing basic data for the entire plan and ensuring that subsequent control has a clear target basis. The path determination module 20 determines the transport vehicle end and the mixing station end of the target work order and determines the transportation path, laying the foundation for subsequent monitoring and prediction of the concrete status during a specific transportation process, and can focus on specific transportation links. The attenuation prediction module 30 collects the on-board operating condition data corresponding to the transport vehicle end based on the transportation path, and performs concrete attenuation prediction to obtain a transport attenuation index, which can estimate the loss of concrete slump during transportation in advance and provide key prediction data for subsequent compensation control. The out-of-machine prediction module 40 collects mixing images at the mixing station end and inputs the images into a dual neural flow network to predict the slump of concrete out of the machine. It can accurately obtain the actual slump of concrete at the production source, so as to correlate it with the attenuation during subsequent transportation. The compensation control module 50 transmits the transport attenuation index to the mixing station to perform compensation control on the predicted slump of the machine, and realizes precise control of the concrete slump through dynamic adjustment until the target slump is met, ensuring that the concrete quality meets the design requirements.

[0071] In summary, this application realizes the accurate prediction and dynamic compensation control of concrete slump during transportation. Through the synergistic effect of various modules, from collecting mix parameters to determining the transportation route, to predicting transportation attenuation and slump out of the machine, and finally performing compensation control, a complete closed-loop control system is formed, which can ensure that the slump of concrete is stable within the design requirements when it arrives at the construction site, greatly improving the quality stability of concrete, reducing construction quality problems caused by slump not meeting requirements, and improving the overall efficiency and reliability of concrete production and construction, which has a positive role in promoting the intelligent and precise development of the concrete industry.

[0072] Example 2, as Figure 2 As shown, based on the same inventive concept as the aforementioned embodiment 1, the embodiment of the present application provides a concrete slump control method combining AI monitoring and ERP, the method comprising: Step S1: Connect to the ERP work order system in the cloud platform to collect the concrete mix design parameters of the target work order.

[0073] Step S2: Determine the transport vehicle end and the mixing station end of the target work order, and determine the transport route according to the transport vehicle end and the mixing station end.

[0074] Step S3: collecting the on-board operating condition data corresponding to the transport vehicle terminal according to the transport route, performing concrete attenuation prediction based on the on-board operating condition data, and obtaining a transport attenuation index for the transport route.

[0075] Step S4: collecting mixing images from the mixing station end, inputting the mixing images into a dual-stream neural network to predict the slump of concrete leaving the machine, and obtaining a predicted slump of concrete leaving the machine.

[0076] Step S5: transmitting the transport attenuation index to the mixing station to perform compensation control on the predicted slump until the target slump in the concrete mix design parameters is met.

[0077] Furthermore, the transport path includes a first transport path, wherein the first transport path is a path for the first transport vehicle-mounted end to supply concrete raw materials to the mixing station end; a raw material transport attenuation model is established, and the raw material transport attenuation model is obtained by training the historical operating condition data of the first transport vehicle-mounted end; the first vehicle-mounted operating condition data of the first transport vehicle-mounted end is collected, and the raw material transport attenuation model is called to predict the concrete raw material attenuation of the first vehicle-mounted operating condition data, and output a transport raw material attenuation index; the transport raw material attenuation index is transmitted to the mixing station end for ratio compensation control of the raw material stage.

[0078] Furthermore, the transport path includes a second transport path, which is the path for the second transport vehicle-mounted end to transport the finished concrete from the mixing station end; a finished product transport attenuation model is established, and the finished product transport attenuation model is obtained by training the historical operating condition data of the second transport vehicle-mounted end and the slump difference label before and after transportation; the second vehicle-mounted operating condition data of the second transport vehicle-mounted end is collected, and the finished product transport attenuation model is called to perform AI prediction of concrete slump attenuation on the second vehicle-mounted operating condition data, and output a transport slump attenuation index; the transport slump attenuation index is transmitted to the mixing station end for finished product compensation control in the delivery stage.

[0079] Furthermore, the finished product transportation attenuation model is called to predict the concrete slump attenuation based on the second vehicle-mounted operating condition data, including: The second vehicle-mounted operating condition data includes transportation time, vehicle vibration amplitude, acceleration change, turning frequency and ambient temperature; the finished product transportation attenuation model is used to extract the transportation attenuation time series characteristics of the second vehicle-mounted operating condition data, predict the slump attenuation based on the extracted transportation attenuation time series characteristics, and output the transportation slump attenuation index.

[0080] Furthermore, the mixing image is input into a dual-stream neural network to predict the slump of concrete out of the machine, and the predicted slump out of the machine is obtained, including: The stirring image includes a video image sequence and a vibration spectrum diagram at multiple angles; the video image sequence and the vibration spectrum diagram are input into the dual-stream neural network, feature analysis is performed on the video image sequence according to the spatial stream network of the dual-stream neural network, and feature analysis is performed on the vibration spectrum diagram according to the temporal stream network of the dual-stream neural network, and image features and time features are output; feature fusion prediction is performed on the image features and the time features to obtain the predicted machine slump.

[0081] Furthermore, after transmitting the transport attenuation index to the mixing station to perform compensation control on the predicted machine slump, the method includes: Detect the real-time slump of concrete at the second transport vehicle-mounted terminal; determine whether the real-time slump of concrete meets the target slump in the concrete mix design parameters; if the real-time slump of concrete does not meet the target slump in the concrete mix design parameters, send a navigation instruction to the second transport vehicle-mounted terminal to control the second transport vehicle-mounted terminal to return to the mixing station terminal; the mixing station terminal compensates for the slump of concrete according to the transport attenuation index.

[0082] Furthermore, a navigation instruction is sent to the second transport vehicle-mounted terminal, wherein the navigation instruction is sent through the cloud platform, and the cloud platform includes multiple mixing station terminals and multiple transport vehicle-mounted terminals; if the real-time slump of the concrete does not meet the target slump in the concrete mix design parameters, the cloud platform obtains the real-time position of the second transport vehicle-mounted terminal; based on the real-time position, the relative distance between the second transport vehicle-mounted terminal and the multiple mixing station terminals is detected, and the neighboring mixing station terminal is obtained, and the neighboring mixing station terminal is the mixing station terminal with the smallest relative distance; the navigation instruction is generated based on the position of the neighboring mixing station terminal.

[0083] Furthermore, the cloud platform includes multiple mixing station terminals, and the method further includes: Initialize the dual-stream neural network; collect multiple mixing control parameter sample groups from the multiple mixing station ends, perform transfer learning on the dual-stream neural network according to the multiple mixing control parameter sample groups, and output multiple dual-stream neural networks; extract multiple network parameters of the multiple dual-stream neural networks, perform fusion learning on the multiple network parameters, and obtain updated multiple dual-stream neural networks; the cloud platform downloads the updated multiple dual-stream neural networks to the multiple mixing station ends.

[0084] Furthermore, after obtaining the predicted slump out of the machine, the actual slump out of the machine at the mixing station end is obtained, the error of the actual slump out of the machine is compared with the predicted slump out of the machine, and the prediction residual is calculated; the prediction residual is used as the error reverse input signal to optimize the parameters of the dual-stream neural network.

[0085] Through the above detailed description of the concrete slump control system combining AI monitoring and ERP in this specification, those skilled in the art can clearly understand the concrete slump control method combining AI monitoring and ERP in this embodiment. For the method disclosed in Example 2, since it corresponds to the system disclosed in Example 1 and has corresponding execution steps and beneficial effects, the relevant details can be referred to the system part description.

[0086] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. The concrete slump control system combining AI monitoring and ERP is characterized by: The system comprises: The mix parameter collection module is used to access the ERP work order system in the cloud platform to collect the concrete mix design parameters of the target work order; A path determination module is used to determine the transport vehicle end and the mixing station end of the target work order, and determine the transport path according to the transport vehicle end and the mixing station end; an attenuation prediction module, configured to collect on-board operating condition data corresponding to the transport vehicle terminal according to the transport route, perform concrete attenuation prediction based on the on-board operating condition data, and obtain a transport attenuation index for the transport route; A concrete slump prediction module is used to collect mixing images from the mixing station end, input the mixing images into a dual-stream neural network to predict the concrete slump out of the machine, and obtain a predicted concrete slump out of the machine; The compensation control module is used to transmit the transportation attenuation index to the mixing station end to perform compensation control on the predicted slump of the machine until the target slump in the concrete mix design parameters is met.

2. The concrete slump control system combining AI monitoring and ERP as claimed in claim 1, characterized in that: The transport path includes a first transport path, wherein the first transport path is a path for supplying concrete raw materials from a first transport vehicle end to the mixing station end, and the path determination module includes: A first construction module is used to establish a raw material transportation attenuation model, wherein the raw material transportation attenuation model is obtained by training the historical operating condition data of the first transport vehicle terminal; a first prediction module, configured to collect first vehicle-mounted operating condition data of the first transport vehicle-mounted terminal, invoke the raw material transport attenuation model to perform concrete raw material attenuation prediction on the first vehicle-mounted operating condition data, and output a transport raw material attenuation index; The first compensation module is used to transmit the attenuation index of the transported raw materials to the mixing station end to perform ratio compensation control at the raw material stage.

3. The concrete slump control system combining AI monitoring and ERP as claimed in claim 1, characterized in that: The transport path includes a second transport path, which is a path for the second transport vehicle to transport the finished concrete from the mixing station. The path determination module further includes: A second construction module is used to establish a finished product transportation attenuation model, wherein the finished product transportation attenuation model is obtained by training the historical operating condition data of the second transport vehicle terminal and the slump difference label before and after transportation; A second prediction module is configured to collect second vehicle-mounted operating condition data of the second transport vehicle-mounted terminal, call the finished product transport attenuation model to perform concrete slump attenuation prediction on the second vehicle-mounted operating condition data, and output a transport slump attenuation index; The second compensation module is used to transmit the transportation slump attenuation index to the mixing station end to perform finished product compensation control at the machine exit stage.

4. The concrete slump control system combining AI monitoring and ERP as claimed in claim 3, characterized in that: The second vehicle operating condition data includes transportation time, vehicle vibration amplitude, acceleration change, turning frequency and ambient temperature; The finished product transportation attenuation model is used to extract transportation attenuation time series features of the second vehicle-mounted operating condition data, perform slump attenuation prediction based on the extracted transportation attenuation time series features, and output a transportation slump attenuation index.

5. The concrete slump control system combining AI monitoring and ERP as claimed in claim 1, characterized in that: The stirring image includes a video image sequence and a vibration spectrum diagram at multiple angles, and the machine failure prediction module is further used to: Inputting the video image sequence and the vibration spectrum graph into the two-stream neural network, performing feature analysis on the video image sequence according to the spatial stream network of the two-stream neural network, and performing feature analysis on the vibration spectrum graph according to the temporal stream network of the two-stream neural network, and outputting image features and temporal features; Perform feature fusion prediction on the image features and the time features to obtain the predicted machine slump.

6. The concrete slump control system combining AI monitoring and ERP as claimed in claim 3, characterized in that: The compensation control module includes: A detection module, configured to detect the real-time slump of concrete at the second transport vehicle end; a judgment module, configured to judge whether the real-time slump of the concrete satisfies the target slump in the concrete mix design parameters; if the real-time slump of the concrete does not satisfy the target slump in the concrete mix design parameters, sending a navigation instruction to the second transport vehicle-mounted terminal to control the second transport vehicle-mounted terminal to return to the mixing station; The third compensation module is used to control the mixing station to perform concrete slump compensation according to the transportation attenuation index.

7. The concrete slump control system combining AI monitoring and ERP according to claim 6, characterized in that: The navigation instruction is sent through the cloud platform, which includes multiple mixing station terminals and multiple transport vehicle terminals. The judgment module includes: a position acquisition module, configured to acquire the real-time position of the second transport vehicle terminal based on the cloud platform if the real-time slump of the concrete does not meet the target slump in the concrete mix design parameters; a neighboring mixing station end identification module, configured to detect the relative distance between the second transport vehicle end and the plurality of mixing station ends based on the real-time position, and obtain a neighboring mixing station end, wherein the neighboring mixing station end is the mixing station end with the smallest relative distance; An instruction generation module is used to generate the navigation instruction based on the position of the adjacent mixing station end.

8. The concrete slump control system combining AI monitoring and ERP according to claim 7, characterized in that: The system further comprises a training module, wherein the training module is configured to: Initializing the two-stream neural network; Collecting multiple mixing control parameter sample groups from the multiple mixing station ends, performing transfer learning on the two-stream neural network according to the multiple mixing control parameter sample groups, and outputting multiple two-stream neural networks; Extracting multiple network parameters of the multiple two-stream neural networks, performing fusion learning on the multiple network parameters, and obtaining multiple updated two-stream neural networks; Based on the cloud platform, the updated multiple dual-stream neural networks are downloaded to the multiple mixing station ends.

9. The concrete slump control system combining AI monitoring and ERP according to claim 1, characterized in that: The compensation control module is further configured to: After obtaining the predicted slump out of the machine, obtaining the actual slump out of the machine at the mixing station end, performing an error comparison between the actual slump out of the machine and the predicted slump out of the machine, and calculating the prediction residual; The prediction residual is used as an error reverse input signal to optimize the parameters of the two-stream neural network.

10. The concrete slump control method combining AI monitoring and ERP is characterized in that: The method is performed by the concrete slump control system combining AI monitoring and ERP according to any one of claims 1 to 9, comprising: Connect the ERP work order system to the cloud platform to collect the concrete mix design parameters of the target work order; Determine the transport vehicle end and the mixing station end of the target work order, and determine the transport route according to the transport vehicle end and the mixing station end; Collecting vehicle-mounted operating condition data corresponding to the transport vehicle-mounted terminal according to the transport route, performing concrete attenuation prediction based on the vehicle-mounted operating condition data, and obtaining a transport attenuation index for the transport route; Acquire mixing images at the mixing station end, input the mixing images into a double-stream neural network to predict the slump of concrete leaving the machine, and obtain a predicted slump of concrete leaving the machine; The transport attenuation index is transmitted to the mixing station to perform compensation control on the predicted slump until the target slump in the concrete mix design parameters is met.

Images