Scribd
Upload
84 views6 pages

Machine Learning: UAS - Paper Review Adityapati Rangga - 20MTTKB 2056102016

1. The document reviews 4 papers related to machine learning and data analytics applications in machinery fault detection, predictive asset management, oil production forecasting, and power systems. 2. Key findings from the papers include that autoencoders are more fault tolerant than other methods for machinery fault detection, integrating diverse asset data streams can improve decision making, segmented regression models performed well for oil production forecasting with uncertainties, and big data analytics can provide new insights into power systems operations. 3. Gaps identified include the need for more comprehensive machinery fault identification, testing predictive analytics in operational control, considering additional uncertainty variables, and addressing challenges of legacy energy sector data systems.

Uploaded by

kurnia rahman agus
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
84 views6 pages

Machine Learning: UAS - Paper Review Adityapati Rangga - 20MTTKB 2056102016

1. The document reviews 4 papers related to machine learning and data analytics applications in machinery fault detection, predictive asset management, oil production forecasting, and power systems. 2. Key findings from the papers include that autoencoders are more fault tolerant than other methods for machinery fault detection, integrating diverse asset data streams can improve decision making, segmented regression models performed well for oil production forecasting with uncertainties, and big data analytics can provide new insights into power systems operations. 3. Gaps identified include the need for more comprehensive machinery fault identification, testing predictive analytics in operational control, considering additional uncertainty variables, and addressing challenges of legacy energy sector data systems.

Uploaded by

kurnia rahman agus
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 6

Machine Learning

UAS - Paper Review

Adityapati Rangga – 20MTTKB


2056102016
Paper list :
1. Teguh Handjoyo Dwiputranto, et al. Machinery Early Fault Detection Using
Artificial Neural Network Based Autoencoder, 3rd International Conference on
Science and Technology - Computer (ICST) IEEE 2017
2. R.A. Platfoot, Predictive Analytics Combining Multi-stream Data Sources,
Springer Proceedings of the 13th World Congress on Engineering Asset
Management 2020
3. Danielle D. Monteiro, et al. Using data analytics to quantify the impact of
production test uncertainty on oil flow rate forecast, Oil and Gas Science and
Technology, 2020
4. Hossein Akhavan-Hejazi, et. al. Power systems big data analytics : An assessment
of paradigm shift barriers and prospects, ScienceDirect Energy Report 4 (2018)
Paper 1 : Machinery Early Fault Detection Using Artificial
Neural Network Based Autoencoder
Objective : Compare performance of ANN Autoencoder with
Industrial practice-Similarity Based Modelling (SBM), for
machinery early fault detection.
Method : RMSE for performance Measurement
Result :

Conclusion : SBM is very sensitive to data anomaly, while


autoencoder is more fault tolerance.

Gap : Only using single machinery data with multi-sensor, and did not provide
fault identification or type. There are still many items that were not done in this
research, for example tuning the parameters of the SBM such as number of
vectors for the state matrix, the technique in choosing the vectors for state matrix,
number of hidden layers for the autoencoder, number of hidden nodes, applying
neuron activation functions other than sigmoid, and another training algorithm.
Paper 2 : Predictive Analytics Combining Multi-stream
Data Sources Objective : Integrate data which requires a complex view of the assets,
linking network or functional connectivity with regional locations to build
up an understanding of how the extensive fleet of assets must work with
each other to achieve the operational
objectives.
Method : The multiple data streams are processed by schema to support reporting
of both current and future risks across the assets, which will improve decision
making for both
current operations and the investment portfolio.
• Expert Judgement
• correlating the timing of transaction data with changes in the real time data

Key Input : Conclusion : This work has identified the essential requirements for integrating
• Datalake, formed by a mix of asset asset information
identifiers, transaction, data and data sets including real time operational data:
realtime data • Improvement of data management
• Transaction data consist of discrete • Refinement of the analysis rules
• Testing of the communication of statistics
transaction such as individual
faults, work orders,
Gap : this work is not an operational control process, but an asset management
• purchase orders or condition process.
measurements.
Paper 3 : Using data analytics to quantify the impact of production test uncertainty
on oil flow rate forecast
Objective : This study aims to introduce uncertainties in oil flow
rate forecast.
Method : This study comprises two main areas: data analytics
using R and Random multiphase flow simulation.
Result : Based on the 52 forecasted dates, 42 presented
accurate results, indicating 80.7% of total dates.
Conclusion : Segmented regression model presented good
results when applied in the statistical analysis, so as the
nonlinear model SVR especially when linear and polynomial
regressions not presented satisfactory results.
Gap : Only select water-cut as uncertainty variable, due to strong
correlation .
Paper 4 : Power systems big data analytics : An assessment of paradigm
shift barriers and prospects Objective :
• This paper provides an assessment of the distinguished aspects in big data analytics
developments in the domain of power systems.
• Perform several taxonomy of the existing and the missing elements in the structures and
methods associated with big data analytics in power systems.
• Provide a holistic outline, classifications, and concise discussions on the technical
approaches, research opportunities, and application areas for energy big data analytics.
Methodology :
• Complex Event Processing
• Data Mining and Computation within Databases
• Integrating Statistical Packages with Databases
• Massive Parallelism
• Models that scale up
Summary : the foundation of the emerging and potential applications of BDA in power systems are expected to be
built upon four new features:

• Higher volumes of data become available, which enables more robust statistical or data mining analysis that
allows increased process accuracy and enhanced control.

• New types of data are emerging, which allows for creating new feedback loops for planning and operation.

• Data can now be better managed, which provides clear, actionable information.

• Advanced analytics is emerging, which allows uncovering new facts about systems to support complex
decision-making.

Gap :
• Although the energy domain data have been growing immensely, a majority of power system data is
yet to be exploited.
• Many of energy domain legacy measurement devices and data management systems are based on
the traditional concept of enterprise data warehousing, whereas under Big Data Analytic (BDA)
approach some of its key fundamentals have been reconsidered.

You might also like