Method for predicting compressive strength of flow-state solidified soil based on BP (Back Propagation) neural network

Abstract

The invention belongs to the technical field of geotechnical engineering, and particularly relates to a flow-state solidified soil compressive strength prediction method based on a BP neural network. In order to predict the compressive strength of the fluid-state solidified soil more accurately and stably with lower errors, the method is based on a BP neural network, a mapping relation between input indexes and output compressive strength is established by utilizing a large amount of training data, and the quality and maintenance age of slag, fly ash, cement and water in the fluid-state solidified soil are used as five inputs of the network, so that the compressive strength of the fluid-state solidified soil is predicted more accurately and stably. And outputting the predicted compressive strength of the flow-state solidified soil under the action of various factors.

Description

Technical Field

[0001] The invention belongs to the technical field of geotechnical engineering, and in particular relates to a method for predicting the compressive strength of fluidized solidified soil based on a BP neural network. Background Art

[0002] At present, fluidized solidified soil has been applied to relatively narrow engineering sites such as foundation pits, fertilizer troughs, and "three backs", but the strength of fluidized solidified soil has a certain uncertainty, resulting in a large amount of manpower and material resources being wasted in the construction process to measure the mechanical indicators of fluidized solidified soil, which lacks real-time performance. In addition, traditional methods cannot predict the compressive strength of fluidized solidified soil at a certain age in the future.

[0003] The performance indicators of fluidized solidified soil are closely related to the water-solid ratio and the type of curing agent. The compressive strength is not in a simple linear relationship with the components. Therefore, it is difficult to establish a relationship between the components of fluidized solidified soil using conventional regression analysis methods.

[0004] The existing method is to establish an empirical relationship between different material mix ratios, construction conditions, and fluidized solidified soil strength through the accumulation of a large amount of experimental data. However, it is costly and time-consuming, and requires a lot of manpower, material resources, and time to obtain a relatively accurate compressive strength. In addition, its scope of application is limited, and it is also unable to consider the complex interactions between various influencing factors, resulting in inaccurate results.

[0005] Therefore, a more accurate, more stable and lower error method for predicting the compressive strength of fluidized solidified soil is needed. Summary of the invention

[0006] In view of the above problems, the present invention provides a method for predicting the compressive strength of fluidized solidified soil based on BP neural network. The BP neural network uses a large amount of training data to establish a mapping relationship between input index and output compressive strength, which can effectively realize multi-index prediction of the compressive strength of fluidized solidified soil. While ensuring the scientific and accurate data, the compressive strength of fluidized solidified soil can be non-destructively predicted, saving a lot of manpower, material resources and financial resources.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] A method for predicting the compressive strength of fluidized solidified soil based on BP neural network comprises the following steps:

[0009] Step 1: Collect the quality and curing age of raw materials required for fluidized solidified soil and construct a data set;

[0010] Step 2, input the data in the data set into a pre-trained BP neural network model, and output the predicted value of the compressive strength of the fluidized solidified soil, wherein the BP neural network model is a BP neural network with a single hidden layer, including an input layer, a hidden layer and an output layer, wherein the input layer is used to input the quality of the raw materials required for the fluidized solidified soil and the curing age, the hidden layer is used to extract the characteristics of the input data, and the output layer is used to output the predicted value of the compressive strength of the fluidized solidified soil. The results of the present invention are more accurate, more stable, and have lower errors.

[0011] Furthermore, the construction of the data set in step 1 is specifically as follows:

[0012] By collecting the different qualities of raw materials required for fluidized solidification soil and the unconfined compressive strength at different curing ages, a reliable and diverse data set is constructed, and then the data set is divided into a training set and a prediction set according to a certain ratio.

[0013] Furthermore, the raw materials required for the fluidized solidified soil are slag, fly ash, cement and water.

[0014] Furthermore, the ratio of the training set to the prediction set is 8:2.

[0015] Furthermore, the method also includes training of the BP neural network model, which includes two stages. The first stage is the forward propagation of the signal, from the input layer through the hidden layer, and finally to the output layer. The second stage is the back propagation of the error, from the output layer to the hidden layer, and finally to the input layer.

[0016] Further, the training of the BP neural network model is specifically as follows:

[0017] The forward propagation is a process of obtaining the predicted value of compressive strength through forward calculation based on the input data in the training set, as shown in the following formula:

[0018]

[0019] Among them, H in is the input value of the hidden layer, H out is the output value of the hidden layer, f is the sigmoid activation function, x is the data of the input layer, O in is the input value of the output layer, O out is the output value of the output layer, ω1 and ω2 are weight matrices, b1 and b2 are bias matrices, and T is the transpose of the matrix;

[0020] After the forward propagation is completed, if the error between the obtained output value and the actual value is large, the error is back-propagated, the weights are updated, and the output is recalculated. By continuously adjusting the network parameters, a BP neural network model is established, in which the total error is calculated as: target is the real value, output is the output value.

[0021] The present invention realizes the prediction of compressive strength of fluidized solidified soil based on BP neural network model, which includes three aspects:

[0022] Sample information collection: Raw materials such as slag, fly ash, cement and water are mixed evenly with soil in a certain proportion, added into the mold, and after standard curing, the strength is measured at a specified age to form a certain number of training sample libraries;

[0023] Training model construction: Through BP neural network, a training model for predicting the strength of fluidized solidified soil is constructed. The quality and curing age of slag, fly ash, cement and water are used as input units of the input layer, and the unconfined compressive strength of fluidized solidified soil is used as the output unit of the output layer. Reasonable parameters such as the number of hidden layers, activation function, and learning rate are set. Through continuous training of sample library samples, a prediction model method for the compressive strength of fluidized solidified soil is established.

[0024] Application of prediction model: The prediction model obtained through sample training can predict the compressive strength value of fluidized solidified soil at a specified age by inputting the mass and curing age of slag, fly ash, cement and water to be predicted.

[0025] Compared with the prior art, the present invention has the following advantages:

[0026] The present invention fills the existing gap in the prediction of the compressive strength of fluidized solidified soil, and can also pave the way for future research on fluidized solidified soil.

[0027] The BP neural network model established in the present invention predicts the average relative percentage error of the fluidized solidified soil strength within 10%, has a higher fitting degree than the traditional neural network prediction model, and the prediction result is scientific and accurate.

[0028] The present invention is based on a BP neural network, improves the accuracy of the prediction model, and realizes the prediction of the compressive strength of fluidized solidified soil at different ages and in different proportions. By inputting the quality and curing age of the raw materials slag, fly ash, cement and water, the output of the predicted compressive strength results can be achieved, thereby reducing the waste of manpower and material resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a flow chart of the method for predicting the compressive strength of fluidized solidified soil based on BP neural network.

[0030] Figure 2 This is a structural diagram of the BP neural network model.

[0031] Figure 3 Setting parameters for the BP neural network model.

[0032] Figure 4 This is the prediction result diagram of the compressive strength prediction method of fluidized solidified soil based on BP neural network.

[0033] Figure 5 This is the error result diagram of the compressive strength prediction method of fluidized solidified soil based on BP neural network. DETAILED DESCRIPTION

[0034] In order to further illustrate the technical solution of the present invention, the present invention is further described below through embodiments.

[0035] Example 1

[0036] like Figure 1 As shown, a method for predicting the compressive strength of fluidized solidified soil based on a BP neural network in this embodiment includes the following steps:

[0037] Step 1: Collect the quality and curing age of raw materials required for fluidized solidified soil and construct a data set;

[0038] This embodiment constructs a data set by collecting different masses of raw materials required for fluidized solidified soil and unconfined compressive strength at different curing ages, and then divides the data set into a training set and a prediction set.

[0039] The raw materials required for fluidized solidified soil are slag, fly ash, cement and water, and the ratio of training set to prediction set is 8:2.

[0040] Step 2: Input the data in the data set into a pre-trained BP neural network model to output the predicted value of the compressive strength of the fluidized solidified soil. The BP neural network model is a single hidden layer BP neural network (structured as Figure 2 As shown in the figure) (including input layer, hidden layer, output layer number), including input layer, hidden layer and output layer, the input layer is used to input the quality of raw materials required for fluidized solidified soil and the curing age, the hidden layer is used to extract the characteristics of input data, and the output layer is used to output the predicted value of the compressive strength of fluidized solidified soil.

[0041] It also includes the training of the BP neural network model, which includes two stages. The first stage is the forward propagation of the signal, from the input layer through the hidden layer, and finally to the output layer. The second stage is the error back propagation, from the output layer to the hidden layer, and finally to the input layer. The specific process is:

[0042] The forward propagation is a process of obtaining the predicted value of compressive strength through forward calculation based on the input data in the training set, as shown in the following formula:

[0043]

[0044] Among them, Hin is the input value of the hidden layer, H out is the output value of the hidden layer, f is the sigmoid activation function, x is the data of the input layer, O in is the input value of the output layer, O out is the output value of the output layer, ω1 and ω2 are weight matrices, b1 and b2 are bias matrices, and T is the transpose of the matrix;

[0045] After the forward propagation is completed, if the error between the obtained output value and the actual value is large, the error is back-propagated, the weights are updated, and the output is recalculated. By continuously adjusting the network parameters, a BP neural network model is established, in which the total error is calculated as: target is the real value, and output is the output value.

[0046] Example 2

[0047] The data set X in this embodiment is {(X1, Y1), (X2, Y2), (X3, Y3), ..., (X n-1 ,Y n-1 ), (X n ,Y n )}, where X i =(X i1 ,X i2 ,X i3 ,……X in ), X i represents the i-th sample parameter, i.e., the quality of slag, fly ash, cement and water and the curing age of the i-th sample, a total of 5 input parameters, X in is the nth parameter value in the ith sample, Y i is the strength of fluidized solidified soil in the i-th sample, and the output value is y i .

[0048] For example, X1=(X 11 =126g, X 12 =42g, X 13 =70g, X 14 =737.1g, X 15 =7d), Y1=1.79MPa.

[0049] The data of some data sets are:

[0050] Slag(g) Fly ash(g) Cement(g) Water(g) Curing age (d) Compressive strength(MPa) 126 42 70 737.1 7 1.79 70 42 70 711.9 7 1.29 98 42 42 711.9 7 1.07 98 70 42 724.5 28 1.98 70 56 42 705.6 28 2.08

[0051] The prediction method of this embodiment includes:

[0052] Step 1: Construct a BP neural network structure with input layer, hidden layer and output layer. The input layer has five units, namely, slag, fly ash, cement and water quality and curing age; the output layer has one unit, namely, the predicted value of compressive strength of fluidized solidified soil.

[0053] Step 2: Construct a data set with different proportions of fluidized solidified soil components to prepare for BP neural network training, and divide the sample set into a training set and a prediction set in a ratio of 8:2.

[0054] Step 3: Perform BP neural network training based on the data in the training set. The network setting parameters are as follows: Figure 3 As shown (the figure includes the number of input layers, hidden layers, and output layers), forward propagation is performed first and then backward propagation is performed, and the network parameters are continuously adjusted and determined. The number of iterations is 5000 and the number of hidden layer nodes is 12.

[0055] Based on the BP neural network-based prediction model for the compressive strength of fluidized solidified soil, the mass and curing age of each component of fluidized solidified soil were input into the program, namely: 98g slag, 56g fly ash, 56g cement, 724g water, and 28d curing age. After calculation by the prediction model, the predicted compressive strength of fluidized solidified soil was 2.3452MPa (e.g. Figure 4 The relative percentage error is 6.3794%, thus realizing the prediction of compressive strength of fluidized solidified soil under non-destructive conditions. Figure 5 It can also be seen from the error result diagram shown that the average relative percentage error of this method is within 10%.

[0056] The above shows and describes the main features and advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes that fall within the meaning and scope of the equivalent elements of the claims are included in the present invention.

[0057] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for predicting the compressive strength of fluidized solidified soil based on BP neural network, characterized in that: The following steps are involved: Step 1: Collect the quality and curing age of raw materials required for fluidized solidified soil and construct a data set; Step 2: Input the data in the data set into a pre-trained BP neural network model and output the predicted value of the compressive strength of the fluidized solidified soil. The BP neural network model is a BP neural network with a single hidden layer, including an input layer, a hidden layer and an output layer. The input layer is used to input the quality of the raw materials required for the fluidized solidified soil and the curing age, the hidden layer is used to extract the features of the input data, and the output layer is used to output the predicted value of the compressive strength of the fluidized solidified soil.

2. The method for predicting the compressive strength of fluidized solidified soil based on BP neural network according to claim 1 is characterized in that: The construction of the data set in step 1 is specifically as follows: By collecting the different qualities of raw materials required for fluidized solidified soil and the unconfined compressive strength at different curing ages, a data set is constructed and then divided into a training set and a prediction set.

3. The method for predicting the compressive strength of fluidized solidified soil based on BP neural network according to claim 2 is characterized in that: The raw materials required for the fluidized solidified soil are slag, fly ash, cement and water.

4. The method for predicting the compressive strength of fluidized solidified soil based on BP neural network according to claim 2 is characterized in that: The ratio of the training set to the prediction set is 8:

2.

5. The method for predicting the compressive strength of fluidized solidified soil based on BP neural network according to claim 2 is characterized in that: The method also includes training of the BP neural network model, which includes two stages. The first stage is the forward propagation of the signal from the input layer through the hidden layer to the output layer, and the second stage is the reverse propagation of the error from the output layer to the hidden layer and finally to the input layer.

6. The method for predicting the compressive strength of fluidized solidified soil based on BP neural network according to claim 5 is characterized in that: The training of the BP neural network model is specifically as follows: The forward propagation is a process of obtaining the predicted value of compressive strength through forward calculation based on the input data in the training set, as shown in the following formula: H out =f(H in ) O out =f(O in ) Among them, H in is the input value of the hidden layer, H out is the output value of the hidden layer, f is the sigmoid activation function, x is the data of the input layer, O in is the input value of the output layer, O out is the output value of the output layer, ω1 and ω2 are weight matrices, b1 and b2 are bias matrices, and T is the transpose of the matrix; After the forward propagation is completed, if the error between the obtained output value and the actual value is large, the error is back-propagated, the weights are updated, and the output is recalculated. By continuously adjusting the network parameters, a BP neural network model is established, in which the total error is calculated as: Target is the output value, and output is the real value.

Images