Premixed concrete working performance detection method based on acoustoelectric signal multi-domain feature fusion convolutional neural network
Through the multi-domain feature fusion convolutional neural network of acoustic and electrical signals, the problems of low accuracy and high cost of premixed concrete detection in the prior art are solved, and high accuracy and low cost of working performance detection are achieved, which simplifies equipment maintenance.
Patent Information
- Application Number
- CN202411053699.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-26
AI Technical Summary
The existing working performance detection methods for ready-mixed concrete rely on manual experience or visual recognition and mixer current judgment, which have problems such as low accuracy, high cost, and difficult equipment maintenance, making it difficult to achieve efficient quality control.
A convolutional neural network is used to collect current and acoustic signals on the mixer, perform preprocessing and feature fusion, and construct a convolutional neural network model to detect concrete working performance.
It improves detection accuracy, reduces calculation volume and cost, facilitates installation and maintenance of equipment, avoids overfitting and underfitting, and achieves efficient concrete working performance detection.
Smart Images

Figure CN119025960A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of ready-mixed concrete production detection, and more specifically, relates to a ready-mixed concrete working performance detection method based on a convolutional neural network with multi-domain features fused from acoustic and electrical signals. Background Art
[0002] Ready-mixed concrete is one of the most widely used building materials, but the material is often prone to fluctuations during the production process, which puts higher demands on the quality control of the ready-mixed concrete plant. Usually, in order to maintain the stability of concrete quality, the ready-mixed concrete plant mainly relies on the experience of quality inspectors or manual slump tests before it leaves the factory, which affects production efficiency. At present, market competition is becoming increasingly fierce, and it is particularly important for concrete companies to strengthen the degree of intelligence in production and quality control.
[0003] With the continuous development of deep learning, AI learning based on a large amount of sample data can already match or even surpass humans in empirical and repetitive work. However, in terms of concrete working performance detection, current research mainly uses camera visual recognition or mixer current for judgment. Due to the problems of single concrete color, low visual distinction, easy overfitting, and large amount of computation of video stream data, the current judgment accuracy is not high, the implementation cost is high, and it is difficult to apply and deploy. In addition, the camera is set inside the mixer, which is difficult to clean and maintain. The mixer current judgment method contains less information and the calculation data is difficult to converge because the time domain information of the current only contains the amplitude changes caused by the change of mixer power. Summary of the invention
[0004] In view of the above defects or improvement needs of the prior art, the present invention proposes a method for detecting the working performance of ready-mixed concrete based on a convolutional neural network with multi-domain features fused from acoustic and electrical signals. The method has high detection accuracy, small system calculation amount, convenient equipment installation and maintenance, and low cost.
[0005] To achieve the above object, the present invention provides a method for detecting the working performance of ready-mixed concrete based on a convolutional neural network with multi-domain features of acoustic and electrical signals, comprising:
[0006] In the production of ready-mixed concrete, current signals and sound signals are used to collect the current signals of the mixer during concrete mixing and the sound signals generated by the collision between the raw materials and the machine body. The corresponding concrete is sampled and tested for working performance for classification;
[0007] The collected raw data of current signals and sound signals are preprocessed, and a sample data set is constructed by corresponding one-to-one with the category labels obtained by sampling tests;
[0008] The acoustic and electrical signal features are fused with a convolutional neural network to build a ready-mixed concrete working performance detection model. The model is trained based on the training set in the sample data set to obtain a trained model.
[0009] The test set in the sample data set is used to evaluate the model and obtain a ready-mixed concrete working performance detection model based on multi-signal feature fusion convolutional neural network.
[0010] In some optional implementation schemes, current signals are collected by installing a current transformer on the input cable of the concrete mixer, and sound signals are collected by installing a sound sensor on the side wall of the mixer or beside the mixer.
[0011] In some optional implementation schemes, the preprocessing of the collected raw data of the current signal and the sound signal includes:
[0012] The original data of the collected current signal and sound signal are filtered and cleaned respectively, and then the filtered and cleaned current signal and sound signal are processed in sections respectively, and data standardization is performed after the section processing.
[0013] Among them, the signal-to-noise ratio is improved by filtering and removing impurities, and the influence of power frequency interference, electromagnetic interference, and environmental interference of acoustic signals in the concrete production process are removed; the longer time series data collected in the concrete mixing process is converted into shorter, fixed-length segments by cutting and segmenting, which reduces the computational complexity of the model and is conducive to the extraction and analysis of fine-grained features. In addition, since the concrete mixing process is accompanied by dynamic changes in signal amplitude and frequency, segmentation is conducive to the extraction of features of time-varying characteristics; the purpose of data standardization is to normalize the data to between [0,1], so that the data converges faster during the training iteration process, thereby improving the stability and accuracy of the model.
[0014] In some optional implementation schemes, the acoustic and electrical signal feature fusion convolutional neural network structure includes: a convolution layer, a pooling layer, a feature splicing layer, a fully connected layer, and a softmax layer in sequence.
[0015] In some optional implementation schemes, the training set data in the sample data set is subjected to convolution and pooling operations to obtain the time-frequency characteristics of the sound signal, and the time domain characteristics of the current signal and the sound signal, respectively. The time-frequency characteristics of the sound signal, and the time domain characteristics of the current signal and the sound signal are then fused by splicing, and finally the forward propagation of the neural network is completed through the fully connected layer and the softmax classifier output.
[0016] In some optional embodiments, the acoustic and electrical signal feature fusion convolutional neural network structure calculates the loss using a cross entropy loss function and updates the weight parameters through back propagation.
[0017] In some optional implementations, the cross entropy loss function is: Among them, N B represents the number of samples, N represents the number of categories of the softmax demodulator, and y ij is the true category value, To calculate category values.
[0018] In general, compared with the prior art, the above technical solution conceived by the present invention proposes a method for detecting the working performance of ready-mixed concrete by using a convolutional neural network with multi-domain features of acoustic and electrical signals, and its beneficial effects and principles are as follows:
[0019] (1) The present invention has a small amount of calculation, is easy to install and maintain, and has low cost, and has obvious advantages in practicality. First, the present invention uses acoustic and electrical one-dimensional signal sources for deep learning, and its data volume is much lower than the graphic or video signals used in visual recognition, which significantly reduces the number of parameters and calculation complexity, and greatly reduces the calculation cost. The previous method used a graphic sensor that needed to be installed inside the mixer, which was troublesome to install, clean and maintain. The acoustic and electrical sensors used in the present invention are all placed outside the mixer.
[0020] (2) The present invention greatly improves the accuracy of concrete working performance testing, which is mainly reflected in two aspects.
[0021] First, the present invention adds acoustic signals for deep learning to judge the working performance of concrete. Compared with previous methods, the characteristics of acoustic signals are more distinguishable and model training is more effective, thus avoiding the occurrence of overfitting or underfitting. Concrete with different working performance corresponds to different slurry encapsulation and flow rate. As the mixer rotates, concrete with different encapsulation or different flow rate rubs and collides with the mixer body, thus reflecting different acoustic signal characteristics.
[0022] Second, the present invention adopts a multi-signal and multi-domain feature fusion convolutional neural network to splice and fuse the features of acoustic and electrical signals, time domain, and time-frequency domain signals, utilizing the information complementarity between multiple signals and multiple domains to improve model recognition performance. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 It is a structural schematic diagram of a convolutional neural network for fusion of acoustic and electrical signal features provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0024] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0025] like Figure 1 As shown, the present invention provides a method for detecting the working performance of ready-mixed concrete based on acoustic and electrical signal feature fusion convolutional neural network, comprising the following steps:
[0026] (1) During the production of ready-mixed concrete, current data is collected by installing a current transformer on the input cable of the concrete mixer, and a sound sensor is installed on the side wall of the mixer to collect sound signals. The corresponding concrete is manually sampled to test the working performance;
[0027] (2) Preprocess the collected raw data, including filtering and removing impurities from acoustic and current signals, segmentation, and data standardization. Data standardization uses the Min-Max method to normalize the data to between [0, 1], and the formula is: The sample data set is constructed one by one, and the data of the manual sampling test are divided into 5 categories (slump: 140-160mm, 160-180mm, 180-200mm, 200-220mm, 220mm-240mm) as label values to assign category labels to the sample data set;
[0028] (3) A convolutional neural network is used to fuse the features of acoustic and electrical signals to construct a ready-mixed concrete working performance detection model. The model is trained using the training set in the sample data set to obtain a trained model. The input training data is processed in the model through operations such as convolution-pooling, etc. to obtain the time-frequency features of the acoustic signal and the time domain features of the current and sound. The features are fused by splicing, and finally the classification is output through the fully connected layer and the softmax classifier to complete the forward propagation of the neural network. After calculating the loss using the cross entropy loss function, the weight parameters are updated through back propagation. The formula for the cross entropy loss function is as follows:
[0029]
[0030] Among them, N B y represents the number of samples, and N represents the number of categories of the softmax demodulator. ij is the true category value, To calculate category values.
[0031] (4) The test set in the sample data set is used to evaluate the model and obtain a ready-mixed concrete working performance detection model based on multi-signal feature fusion convolutional neural network.
[0032] (5) The model is configured in the computer of the control room of the mixing station. During production, the mixer current signal and the collected sound signal are transmitted to the computer and input into the model to complete the working performance test of the ready-mixed concrete.
[0033] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, and two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.
[0034] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for detecting the working performance of ready-mixed concrete based on a convolutional neural network with multi-domain features of acoustic and electrical signals, characterized in that: include: In the production of ready-mixed concrete, current signals and sound signals are used to collect the current signals of the mixer during concrete mixing and the sound signals generated by the collision between the raw materials and the machine body, and the corresponding concrete is sampled to test the working performance for classification; The collected raw data of current signals and sound signals are preprocessed, and a sample data set is constructed by corresponding one-to-one with the category labels obtained by sampling tests; The acoustic and electrical signal features are fused with a convolutional neural network to build a ready-mixed concrete working performance detection model. The model is trained based on the training set in the sample data set to obtain a trained model. The test set in the sample data set is used to evaluate the model and obtain a ready-mixed concrete working performance detection model based on multi-signal feature fusion convolutional neural network.
2. The method according to claim 1, characterized in that Current signals are collected by installing a current transformer on the input cable of the concrete mixer, and sound signals are collected by installing a sound sensor on the side wall or next to the mixer.
3. The method according to claim 1, characterized in that: The preprocessing of the collected raw data of the current signal and the sound signal includes: The original data of the collected current signal and sound signal are filtered and cleaned respectively, and then the filtered and cleaned current signal and sound signal are processed in sections respectively, and data standardization is performed after the section processing.
4. The method according to any one of claims 1 to 3, characterized in that: The acoustic and electrical signal feature fusion convolutional neural network structure includes: a convolution layer, a pooling layer, a feature splicing layer, a fully connected layer, and a softmax layer.
5. The method according to claim 4, characterized in that After convolution and pooling operations, the training set data in the sample data set obtains the time-frequency characteristics of the sound signal and the time domain characteristics of the current signal and the sound signal respectively. Then, the time-frequency characteristics of the sound signal and the time domain characteristics of the current signal and the sound signal are fused by splicing. Finally, the forward propagation of the neural network is completed through the fully connected layer and the softmax classifier output.
6. The method according to claim 5, characterized in that The acoustic and electrical signal feature fusion convolutional neural network structure calculates the loss using a cross entropy loss function and updates the weight parameters through back propagation.
7. The method according to claim 6, characterized in that The cross entropy loss function is: Among them, N B represents the number of samples, N represents the number of categories of the softmax demodulator, and y ij is the true category value, To calculate category values.
Citation Information
Patent Citations
Single frequency alarm sound feature detection method on basis of STFT (Short-Time Fourier Transform)
CN103674235A
Method and device for monitoring state of concrete in mixer truck based on voice recognition
CN112378997A
Rolling bearing acoustic emission intelligent diagnosis method based on optimal wavelet basis and multi-dimensional depth feature fusion
CN114048682A
Method and device for diagnosing void defect of concrete filled steel tube structure based on sound signal
CN115356397A
Rotary machinery intelligent fault diagnosis method and system based on feature selection
CN118312832A
Cited By
Building structure health state assessment method using environmental micro-vibration signals and deep learning
CN122196673A
