Labquake Prediction

• 

  TCN model used in the original paper

Title: Using TCN for earthquake prediction and forecasting of fault zone stress

Who: Maria Wang mwang272 Eric Long Him Ko eko10 Eleanor Park enpark

Introduction: The objectives of the paper are to develop deep learning models for predicting laboratory earthquakes, apply autoregressive forecasting to predict fault zone stress and then evaluate the performance of the models in terms of accuracy and reliability so as to better understand earthquake mechanisms and to develop predictive models that can assist in early warning systems and earthquake hazard assessments.

We chose this paper mainly due to the recent events of devastating earthquakes in both Japan and Taiwan, and wish to use deep learning to explore how such natural disasters can be better predicted and then avoided.

The problem is primarily a regression problem, as it involves predicting continuous values (earthquake magnitudes, stress levels) based on input features.

Related Work:

In another paper, titled Attention Network Forecasts Time-to-Failure in Laboratory Shear Experiments, the research focuses on forecasting synthetic earthquakes or "labquakes" created in a controlled laboratory setting. Rocks under stress emit small bursts of energy known as acoustic emissions (AE), which are believed to contain predictive information about impending failure. The primary objective is to harness this AE data to forecast the time-to-failure (TTF) of labquakes. In the study, they used Conscience Self-Organizing Map (CSOM) to perform topologically ordered vector quantization based on AE waveform properties, leading to the creation of clusters. These clusters were analyzed over time to identify those with predictive capabilities. Subsequently, Long Short-Term Memory (LSTM) networks and attention-based networks were utilized to test the forecasting power of the selected AE clusters. The final findings was that the attention network was able to forecast TTF and shear stress within a reasonable amount of error for the majority of the test set and overall worked better than the LSTM model. Furthermore, through clustering, it is found that a subset of acoustic emissions contain sufficient information for these tasks and hence network training does not require continuous seismic data, which would make collecting the data and building the model easier.

Original paper implementation: https://github.com/lauralaurenti/DNN-earthquake-prediction-forecasting/blob/main/README.md

Data: Our model requires acoustic emissions data, which is sound released when the ground undergoes stress. In the majority of previously created models that use acoustic emissions to predict earthquakes, the data is lab generated. Therefore, we plan to use a dataset compiled by the US Department of Energy that contains lab-generated fault zone acoustic emissions data. This dataset has already been processed and normalized, so our preprocessing will mostly involve taking out unneeded data columns, converting the dataset into a usable form, and splitting the data into 80% training and 20% testing.

Methodology: The architecture of our model is based on a Temporal Convolutional Network (TCN). TCN is a type of Convolutional Neural Network (CNN) specifically designed for sequence modeling tasks. It consists of causal and dilated 1D convolutional layers with the same input and output lengths.

In the existing implementation, the model is composed of three convolutional 1D layers. The sequence is scanned in a causal fashion, with dilation set to 1. The first layer has a hidden size of 64, while the second layer has a hidden size of 256. The last layer has the dimension of the output, which in our case is 1 because the model forecasts the next step. Training the TCN model involves optimizing its parameters to minimize a chosen loss function, such as mean squared error (MSE) or mean absolute error (MAE). We use an optimizer, such as stochastic gradient descent (SGD) or Adam, to update the model parameters during training.

One potential challenge in implementing the TCN model could be tuning the hyperparameters, such as the number of layers, hidden sizes, and dilation settings, to achieve optimal performance. Additionally, ensuring efficient handling of variable-length sequences and managing computational resources during training may also be areas of focus. Overall, the TCN-based model offers a promising approach for sequence modeling tasks, such as earthquake prediction and forecasting, leveraging the capabilities of convolutional neural networks to capture temporal dependencies in the data effectively.

Metrics: Since this is a regression project, accuracy is the best metric to analyze the success of our project, which is also what the author used to quantify the results of their model.

Our base goal is to implement the best of the models that were tried in the research paper, which is the Temporal Convolutional Network, with an 85% accuracy. We would convert the model into a TensorFlow model and use a different dataset and apply data augmentation. Our target goal would be to make our TCN network have an accuracy above 93%. Our stretch goal would be to implement all the models that the research tested out, including Long short-term memory and the Transformer Network, then determine which has the highest accuracy.

Ethics: Broader Societal Issues: While the intent behind earthquake predictions is noble, aiming to save lives and minimize property damage, there are potential downsides to consider. False alarms resulting from inaccurate predictions can lead to unnecessary panic, evacuations, or trauma within communities. This can erode public trust in the prediction systems and create unnecessary strain on emergency services. Additionally, there's a disparity concern: regions with fewer financial resources or less developed infrastructure might not benefit equally from accurate earthquake predictions due to limited access to early warning systems or preventive measures.

Data Collection Issues: Regarding data collection, the current reliance on laboratory-generated datasets raises certain considerations. While such data collection methods do not infringe upon individuals' privacy, there's a risk that the data may not accurately represent real-world earthquake dynamics. This could lead to the development of flawed prediction models, ultimately impacting the efficacy of disaster preparedness efforts. Furthermore, if future data collection endeavors were to involve actual earthquake data, there's a possibility that the methods used could be intrusive, potentially raising concerns about consent, privacy, and the ethical treatment of communities affected by earthquakes.

Division of labor: We anticipate data preprocessing to be very difficult, so we will work together to convert the data into a usable form. In general, we expect to work on all of the parts together.

Built With

• python