| What is included as part of preventative maitenance? | -predetermined maintenance (follows factory schedule)
-condition-based maintenance (occurs when a situation of condition indicates maintenance is needed) |
| What is included as part of preventative maitenance? | -predetermined maintenance (follows factory schedule)
-condition-based maintenance (occurs when a situation of condition indicates maintenance is needed) |
| What is included as part of preventative maitenance? | -predetermined maintenance (follows factory schedule)
-condition-based maintenance (occurs when a situation of condition indicates maintenance is needed) |
| What is included as part of preventative maitenance? | -predetermined maintenance (follows factory schedule)
-condition-based maintenance (occurs when a situation of condition indicates maintenance is needed) |
| What is included as part of preventative maitenance? | -predetermined maintenance (follows factory schedule)
-condition-based maintenance (occurs when a situation of condition indicates maintenance is needed) |
| NA | NA |
| What is included as part of preventative maitenance? | -predetermined maintenance (follows factory schedule)
-condition-based maintenance (occurs when a situation of condition indicates maintenance is needed) |
| What is artificial intelligence (AI)? | Artificial intelligence (AI) is a popular branch of computer science that concerns with building “intelligent” smart machines capable of performing intelligent tasks. |
| What is machine learning? | Field of study that gives computer the ability to learn without being explicitly programmed. |
| What is deep learning? | A subfield of machine learning that is concerned with algorithms inspired by the brain’s structure, such as artificial neural network. |
| What are some applications of machine learning? | -Deep learning applications.
-Predictive analytics.
-Translation.
-Information extraction. |
| What are the ways in which machines can learn? | Supervised learning and unsupervised learning. |
| What is supervised learning? | An approach to creating artificial intelligence (AI) based on labelled training data consisting of input objects and a desired output value.
After trained by using labelled data, it can be used to predict the outcomes based on the unlabelled data (only inputs are known). |
| How can supervised learning be achieved for simple engineering problems? | Regression. |
| How can supervised learning be achieved for complex engineering problems? | By combining different regressions together, namely artificial neural networks (ANN). |
| What are the approaches of supervised learning using ANN? | -ANN using activation function
-Fuzzy neural network (FNN) using membership function |
| What are artificial neural networks (ANN)? | Series of algorithms that mimic the operations of a human brain to recognise relationships between vast amounts of data. |
| What is a fuzzy set defined by? | Its unclear boundaries. |
| When is fuzzy set theory used? | When the activation function has a very steep change, fuzzy set theory is needed for problems relating to imprecise judgements. |
| How is the element in fuzzy set theory presented? | ? = {?, pf(?)}
where:
-x is the element value,
-pf(x) is membership value showing the degree of probability of x belonging to A (e.g., a high cost category) |
| What are support-vector machines (SVMs) in machine learning? | Supervised learning models with associated learning algorithms that analyse data for classification and regression analysis. |
| What is the process of supervised training using Support Vector Machine (SVM)? | Labelled data for training -> determine hyperplane (&unlabelled data) -> trained support vector machine -> risk level prediction |
| How is the best decision boundary (hyperplane) determined? | The boundary line with the maximum margin with a high fault tolerance. |
| What are the vectors that sit on the hyperplane called? | support vectors |
| What do recurrent neural networks (RNN) do? | Recognise sequential characteristics of data and use patterns to predict the next likely scenario.
Neural networks which has memory. |
| How are recurrent neural networks (RNN) sequential? | The outputs of hidden layer of RNN are stored in the memory and considered as inputs as well. |
| What is the process in which sequential data is input into a recurrent neural network (RNN) after training? | • The process of inputting sequential data into an RNN after training involves pre-processing the data into a suitable format that can be fed into the network.
• As RNNs are designed to capture temporal dependencies in sequential data, they may not be able to capture any meaningful temporal dependencies with a single data point. |
| What are the limitations of recurrent neural networks (RNN)? | Can not maintain information in memory over a long period.
Therefore, it is not suitable to predict data consequences over a long period of time. |
| What are the advantages of long short-term memory (LSTM) networks compared to recurrent neural networks (RNN)? | • Retain information for longer periods, making them more effective in modelling long-term dependencies.
• Better control of the flow of information using the gating mechanism (input gate, forget gate and output gate).
• Better performance on sequence-to-sequence tasks.
• More effectively filter out irrelevant information and noise in
a sequence due to their gating mechanism. |
| How does the gating mechanism of LSTM work? | Information in LSTMs can be stored, written, or read via gates that open and close. These gates store the memory in the analogue format, implementing element-wise multiplication by sigmoid ranges between 0-1. |
| Why is preventative maintenance of infrastructure projects important? | Ensure the proper functioning and longevity of infrastructure. |
| What can LTSM be used for? | Predict and identify potential maintenance issues before they become critical. |
| How can LSTM networks be implemented as part of preventative maitenance? | • Step 1 - Data collection (e.g., historical inspection data)
• Step 2 - Data preprocessing (e.g., handle missing values, and normalize the data)
• Step 3 - Design an LSTM network (e.g., input layers, LSTM layers, and output layers)
• Step 4 – Training the model (e.g., Split the preprocessed data into training and validation sets)
• Step 5 - Model evaluation (e.g., Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) )
• Step 6 - Predictive maintenance (e.g., schedule maintenance based on residual life prediction)
• Step 7 - Continuous improvement (e.g., continuously collect new data for training) |
| What is included as part of preventative maitenance? | -predetermined maintenance (follows factory schedule)
-condition-based maintenance (occurs when a situation of condition indicates maintenance is needed) |
