Description:DESCRIPTION:
Field of the invention:
[0001] The present disclosure generally relates to the technical field of a system for evaluating liquefaction potential of soil and, in particular, relates to a system for evaluating liquefaction potential of subsoil layers using an artificial neural network module.
Background of the invention:
[0002] Liquefaction is a phenomenon that happens mostly when saturated fine sands and silts are exposed to earthquakes or dynamic loads. Because of increasing pore water pressure, the saturated cohesion less soils entirely lose their strength during the occurrence. Since the 1964 Niigata and Alaska earthquakes in Japan, the phenomenon has been frequently noticed all over the world. Liquefaction has disastrous consequences for infrastructure projects, such as foundation and settlement issues, lateral spreading of embankments, sand boils and ground vibrations, and so on.

[0003] Several researchers have undertaken extensive studies and suggested techniques for forecasting the likelihood of liquefaction based on field and laboratory test data. It was discovered that there are several limits for utilizing laboratory test results to forecast soil liquefaction potential. One of the most significant disadvantages of laboratory test findings is that they do not account for actual soil features such as fabric, soil structure, prior strain history and over-consolidation.

[0004] Another significant disadvantage is that laboratory equipment is prohibitively expensive, time-consuming, and labor-intensive. Analytical approaches such as the finite element method are sometimes impeded because they require a large number of parameters to adequately describe complicated geotechnical engineering issues such as liquefaction. These disadvantages are solved by utilizing field test data from the SPT test, the CPT test, and other field tests that are routinely used for analyzing the liquefaction potential of soils.

[0005] The most prevalent technique proposed based on field data is the Seed and Idriss method, and they proposed a simpler method that takes into account all elements that impact liquefaction. Iwasaki et al. suggested a strategy based on liquefaction resistance and prospective variables for evaluating soil liquefaction susceptibility. Idriss and Boulanger offered a number of possibly relevant correlations and suggestions related to seismically induced soil liquefaction. However, due to the complexity of the problem and the uncertainty in soil properties, these empirical and semi-empirical techniques have considerable drawbacks.

[0006] However, the major disadvantage of conventional methods is their reliance on empirical correlations and simplified assumptions. These approaches often overlook critical factors that can impact the liquefaction potential of soil, such as soil composition, stress history and groundwater conditions. As a result, the assessments produced by these methods may not accurately reflect the true liquefaction potential of the subsoil layers.

[0007] Another issue with traditional techniques is their inability to account for site-specific conditions. The methods are typically based on generalized data and do not consider the unique characteristics of a particular site. This can lead to overestimation or underestimation of the liquefaction potential, which can have serious consequences for the safety and stability of structures built on the subsoil layers.

[0008] Furthermore, conventional methods can be time-consuming and costly. They often require extensive field testing and laboratory analysis, which can delay construction projects and increase expenses. Additionally, the results of these methods may not be available in real-time, which can further delay decision-making and increase project costs.

[0009] Therefore, there is a need for a system to incorporate advanced technologies and data analysis methods for evaluating the liquefaction potential of subsoil layers using an artificial neural network module. There is also a need for a system to offer more accurate and reliable assessments of liquefaction potential of subsoil layers using an artificial neural network module. There is also a need for a system that can ensure safety and stability of construction projects by reducing costs and improving efficiency using an artificial neural network module.
Objectives of the invention:
[0010] The primary objective of the invention is to develop a system for evaluating liquefaction potential of subsoil layers using an artificial neural network module.

[0011] Another objective of the invention is to develop a system that can ensure safety and stability of construction projects by evaluating the liquefaction potential of subsoil layers using an artificial neural network module.

[0012] Yet another objective of the invention is to develop a system that offers more accurate and reliable assessments of liquefaction potential of subsoil layers using an artificial neural network module.

[0013] Further objective of the invention is to develop a system that reduces cost and improves efficiency for evaluating the liquefaction potential of subsoil layers using an artificial neural network module.
Summary of the invention:
[0014] The present disclosure proposes a system and method for predicting liquefaction susceptibility of subsoil layers using artificial neural network. The following presents a simplified summary in order to provide a basic understanding of some aspects of the claimed subject matter. This summary is not an extensive overview. It is not intended to identify key/critical elements or to delineate the scope of the claimed subject matter. Its sole purpose is to present some concepts in a simplified form as a prelude to the more detailed description that is presented later.

[0015] In order to overcome the above deficiencies of the prior art, the present disclosure is to solve the technical problem to provide a system for evaluating liquefaction potential of subsoil layers using an artificial neural network module.

[0016] According to one aspect, the system offers more accurate and reliable assessments of liquefaction potential of subsoil layers. In one embodiment herein, the system can ensure safety and stability of construction projects by evaluating the liquefaction potential of subsoil layers. The system reduces cost and improves efficiency for evaluating the liquefaction potential of subsoil layers. In one embodiment herein, the system comprises a computing device and an artificial neural network module.

[0017] In one embodiment herein, the computing device is having a controller and a memory for storing information to be processed and instructions executable by the controller. The controller is in communication with the artificial neural network module. In one embodiment herein, the artificial neural network module is trained with a dataset, thereby accessing the dataset for evaluating the liquefaction potential of subsoil layers.

[0018] In one embodiment herein, the artificial neural network module is a single-layer feed-forward artificial neural network. In one embodiment herein, the artificial neural network module is trained, validated and tested from the dataset. The dataset is divided into 60:20:20 proportions, which means 60% of the data is used in training, 20% of the data is used in validation and 20% of the data is used in testing for prediction.

[0019] In one embodiment herein, the artificial neural network module is having plurality of epochs to train the artificial neural network module. Each epoch is fed with a set of training data that is used to adjust weights of the artificial neural network module. In one embodiment herein, the artificial neural network module comprises an input layer, an output layer and a hidden layer.

[0020] In one embodiment herein, the input layer is having plurality of neurons configured to receive an experimental data as input parameters. In one embodiment herein, the input parameters include a study area, a depth, an SPT N-value, a zone and an earthquake intensity magnitude. In one embodiment herein, the output layer is configured to output a liquefaction result in form of classifications and predictions based on the input data.

[0021] In one embodiment herein, the hidden layer is configured to extract data from the input layer and provide the output to the output layer based on the stored dataset. The hidden layer comprises one or more neurons based on an average of the neurons in the input layer and output layer.

[0022] According to one aspect, a method is disclosed for evaluating liquefaction potential of subsoil layers using a system. First, at one step, the artificial neural network module is trained with the dataset and accessing the dataset for evaluating liquefaction potential of subsoil layers. At another step, the input layer of the artificial neural network module is inputted with the experimental data as input parameters.

[0023] At another step, the data is extracted from the input layer and provides output to the output layer of the artificial neural network module based on the dataset through the hidden layer of the artificial neural network module. At another step, a liquefaction result is outputted in the form of classifications and predictions based on the dataset through the output layer.

[0024] Further, objects and advantages of the present invention will be apparent from a study of the following portion of the specification, the claims, and the attached drawings.
Detailed description of drawings:
[0025] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate an embodiment of the invention, and, together with the description, explain the principles of the invention.

[0026] FIG. 1 illustrates a block diagram of a system for evaluating liquefaction potential of subsoil layers, in accordance with an exemplary embodiment of the invention.

[0027] FIG. 2 illustrates a schematic view of an artificial neural network module of the system, in accordance with an exemplary embodiment of the invention.

[0028] FIG. 3 illustrates a graphical representation between accuracy percentage of the artificial neural network module predictions and study areas, in accordance with an exemplary embodiment of the invention.

[0029] FIG. 4 illustrates a flowchart of a method evaluating liquefaction potential of subsoil layers using the system, in accordance with an exemplary embodiment of the invention.
Detailed invention disclosure:
[0030] Various embodiments of the present invention will be described in reference to the accompanying drawings. Wherever possible, same or similar reference numerals are used in the drawings and the description to refer to the same or like parts or steps.

[0031] The present disclosure has been made with a view towards solving the problem with the prior art described above, and it is an object of the present invention to provide to a system for evaluating liquefaction potential of subsoil layers using an artificial neural network module.

[0032] According to an exemplary embodiment of the invention, FIG. 1 refers to a schematic view of a system for evaluating liquefaction potential of subsoil layers. In one embodiment herein, the system 100 offers more accurate and reliable assessments of liquefaction potential of subsoil layers using the artificial neural network module 101. In one embodiment herein, the system 100 can ensure safety and stability of construction projects by evaluating the liquefaction potential of subsoil layers using the artificial neural network module 101. The system 100 reduces cost and improves efficiency for evaluating the liquefaction potential of subsoil layers artificial neural network module 101. In one embodiment herein, the system 100 comprises a computing device 102 and the artificial neural network module 101.

[0033] In one embodiment herein, the computing device 102 having a controller 104 and a memory 106 for storing information to be processed and instructions executable by the controller 104. The controller 104 is in communication with the artificial neural network module 101. In one embodiment herein, the artificial neural network module 101 is trained with a dataset, thereby accessing the dataset whenever needed for evaluating liquefaction potential of subsoil layers. In one embodiment herein, the artificial neural network module 101 is a tool for solving problems that are complex and unsolved. The artificial neural network module 101 works in the same manner as human brain by taking past experiences and measurements to solve issues and face situations.

[0034] According to an exemplary embodiment of the invention, FIG. 2 illustrates a schematic view of the artificial neural network module 101 of the system 100. The artificial neural network module 101 comprises an input layer 108, an output layer 110 and a hidden layer 112. In one embodiment herein, the input layer 108 is having plurality of neurons configured to receive an experimental data as input parameters. In one embodiment herein, the input parameters include a study area, a depth, an SPT N-value, a zone and an earthquake intensity magnitude. In one embodiment herein, the output layer 110 is configured to output a liquefaction result in form of classifications and predictions based on the input data.

[0035] In one embodiment herein, the hidden layer 112 is configured to extract data from the input layer 108 and provide the output to the output layer 110 based on the extracted data. In one embodiment herein, the hidden layer 112 can be more than one based on the requirement. The hidden layer 112 comprises one or more neurons based on an average of the neurons in the input layer 108 and the output layer 110.

[0036] In one embodiment herein, an operation of the artificial neural network module 101 is done in two phases, the first one is a training or a learning phase and the second one is a recall or a retrieval phase. In the first phase, the data is supplied to the artificial neural network module 101 for training or learning, and the learning phase is highly time-consuming yet seeks the best performance. The retrieval phase can be rapid once the network is trained because processing can be distributed. The artificial neural network module 101 learns by adjusting the interconnection weights between layers.

[0037] In one embodiment herein, a proper dataset plays a vital role in solving any complex problem precisely using the artificial neural network module 101. The dataset may consist of characteristics namely reliability and sufficient data that covers all affected parameters. In the present study, 10 selected study areas are considered in the region of Visakhapatnam City in Andhra Pradesh state, India. Visakhapatnam is a coastal city comprising of saturated fine sands and silts at shallow depths.

[0038] In one embodiment herein, the dataset required to constitute the artificial neural network module should account for various factors that affect complexity of the relationship among them. Hence study area, depth of soil, the SPT N-value, earthquake zone (Z2, Z3, Z4 &Z5), and earthquake magnitude intensity (5M, 5.5M, 6M &6.5M) are the parameters used as dataset to develop the artificial neural network module. The dataset is imported into the artificial neural network module as a dot CSV file and 592 data points are used.

[0039] In one embodiment herein, the artificial neural network module 101 is trained, validated and tested from the same dataset. In one embodiment herein, the data is divided in 60:20:20 proportions, which means 60% of the data is used in training and 20% of the data is used in validation and 20% of the data is used in testing for prediction. After grouping the information, the data is pre-processed to avoid the dimensional dissimilarities of different input parameters and to get better results. As the dataset shall be wholly numerical, liquefaction is indicated with “1”, and non-liquefaction is indicated with “0”.

[0040] In one embodiment herein, the artificial neural network module 101 is developed by Python with Keras deep learning technique. In one embodiment herein, the artificial neural network module 101 is a single-layer feed-forward neural network. The single-layer feed-forward neural network represents the most simple form of the neural network, in which there is only one layer of input nodes that send weighted inputs to a subsequent layer of receiving nodes, or in some cases, one receiving node. In one embodiment herein, the artificial neural network module 101 is having plurality of epochs to train the artificial neural network module 101. Each epoch is fed with a set of training data that is used to adjust weights of the artificial neural network module 101.

[0041] In one embodiment herein, the data is fed in form of input parameters to the artificial neural network module 101 by five input neurons and the output is represented by a single node as shown in FIG. 2. The input parameters considered for this study area (SA), depth, SPT N-value, zone, and earthquake intensity magnitude. At the end of the training and validation process, 20% of the data is used for testing. From the total input of 592 data points, 119 points are tested. Out of which 26 points are from the study area-1 (SA-1) and 6,6,11,12,5,9,21,14,8 points from the remaining study areas respectively for different conditions.

[0042] In one embodiment herein, summary of observed and predicted values of selected study areas are mentioned in Table 2.
Table 2:
Study areas Observed Values by IS 1893 part 1 (2016) (Non-Liquefaction-0 /Liquefaction-1 ) ANN Predicted Non-Liquefaction ANN Predicted Liquefaction
SA-1 (26) 0 13 0
1 1 12
SA-2(6) 0 2 1
1 0 4
SA-3(6) 0 2 2
1 0 2
SA-4(11) 0 6 0
1 0 5
SA-5(12) 0 9 1
1 0 2
SA-6(5) 0 2 2
1 0 1
SA-7(9) 0 0 0
1 3 6
SA-8(21) 0 6 4
1 0 11
SA-9(14) 0 0 3
1 1 10
SA-10(8) 0 1 0
1 0 7
Total 0 41 13
1 5 60

[0043] In one embodiment herein, from Table 2, it is observed that the artificial neural network module 101 predicted 41 non-liquefaction points as non-liquefaction points and 13 non liquefaction points as liquefaction points when compared to the observed values by IS 1893 part 1 (2016). Similarly, five liquefaction points as non-liquefaction points, and 60 liquefaction points as liquefaction points. In one embodiment herein, five liquefaction points predicted as non-liquefaction points by the artificial neural network module 101 are noticed as critical among all predictions when compared to the observed values by the IS 1893 part 1 (2016).

[0044] According to an exemplary embodiment of the invention, FIG. 3 illustrates a graphical representation 300 between accuracy percentage of the artificial neural network module predictions and study areas. Regarding to the Table 2, the percentage accuracy of ANN predictions for all study areas are represented in Fig.3. From the results, it is observed that for study areas 4 and 10, the observed and predicted results are 100% similar and for study areas 1 and 5 the prediction accuracy is observed to be greater than 95%. For remaining areas the accuracy of prediction varied from 67% to 90% due to less training and testing data fed to the network.

[0045] In one embodiment herein, from the output results, it is observed that the developed is capable enough to predict the liquefaction susceptibility of subsoil layers in Visakhapatnam city region. The input parameters viz. depth, SPT N-value, seismic zone, and earthquake intensity magnitude formed a strong correlation and effectively predicted the liquefaction potential of soils. The efficiency and accuracy of the prediction is almost 90% and it can be further improved by feeding more input data. The predicted results are compared with standard simplified procedures and the predictions are within acceptable confidence level to the IS 1893 Part-1 (2016) method. The artificial neural network module 101 can be used as simpler and more reliable over the conventional methods of evaluation the liquefaction potential of subsoil layers.

[0046] According to an exemplary embodiment of the invention, FIG. 4 illustrates a flowchart 400 of a method for evaluating liquefaction potential of subsoil layers using the system 100. First, at step 402, the artificial neural network module 101 is trained with the dataset and accessing the dataset whenever needed for evaluating liquefaction potential of subsoil layers. At step 404, the input layer 108 of the artificial neural network module 101 is inputted with the experimental data as input parameters.

[0047] At step 406, the data is extracted from the input layer 108 and provides output to the output layer 110 of the artificial neural network module 101 based on the dataset through the hidden layer 112 of the artificial neural network module 101. At step 408, a liquefaction result is outputted in the form of classifications and predictions based on the dataset through the output layer 112.

[0048] Numerous advantages of the present disclosure may be apparent from the discussion above. In accordance with the present disclosure, the system 100 offers more accurate and reliable assessments of liquefaction potential of subsoil layers using the artificial neural network module 101. The proposed system 100 can ensure safety and stability of construction projects by evaluating the liquefaction potential of subsoil layers using the artificial neural network module 101. The system 100 reduces cost and improves efficiency for evaluating the liquefaction potential of subsoil layers using the artificial neural network module 101.

[0049] It will readily be apparent that numerous modifications and alterations can be made to the processes described in the foregoing examples without departing from the principles underlying the invention, and all such modifications and alterations are intended to be embraced by this application.
, Claims:CLAIMS:
I/We Claim:
1. A system (100) for evaluating liquefaction potential of subsoil layers, comprising:
a computing device (102) having a controller (104) and a memory (106) for storing information to be processed and instructions executable by the controller (104), wherein the controller (104) is in communication with an artificial neural network module (101),
wherein the artificial neural network module (101) is trained with a dataset and access the dataset for evaluating liquefaction potential of subsoil layers, wherein the artificial neural network module (101) comprises:
an input layer (108) having plurality of neurons configured to receive an experimental data as input parameters;
an output layer (110) configured to output a liquefaction result in form of classifications and predictions based on the input data; and
a hidden layer (112) configured to extract data from the input layer (108) and provide the output to the output layer (110) based on the dataset.
2. The system (100) as claimed in claim 1, wherein the artificial neural network module (101) is a single-layer feed-forward artificial neural network.
3. The system (100) as claimed in claim 1, wherein the artificial neural network module (101) is trained, validated and tested from the dataset, wherein the dataset is divided in 60:20:20 proportions, which means 60% of data is used in training, 20% of data is used in validation and 20% of data is used in testing for prediction.
4. The system (100) as claimed in claim 1, wherein the input parameters include a study area, a depth, an SPT N-value, a zone and an earthquake intensity magnitude.
5. The system (100) as claimed in claim 1, wherein the hidden layer (112) comprises one or more neurons based on an average of the neurons in the input layer (108) and the output layer (110).
6. The system (100) as claimed in claim 1, wherein the artificial neural network module (101) is having plurality of epochs to train itself, wherein each epoch is fed with a set of training data that is used to adjust weights of the artificial neural network module (101).
7. A method for evaluating liquefaction potential of subsoil layers using a system (100), comprising:
training an artificial neural network module (101) with a dataset and accessing the dataset for evaluating liquefaction potential of subsoil layers;
inputting experimental data as input parameters to the input layer (108) of the artificial neural network module (101);
extracting the data from the input layer (108) and providing output to an output layer (110) of the artificial neural network module (101) based on the dataset through a hidden layer (112) of the artificial neural network module (101); and
outputting a liquefaction result in form of classifications and predictions based on the input data through the output layer (110).