Hasil Pencarian  ::  Simpan CSV :: Kembali

"In this research, based on actual data gathered from an industrial scale vacuum gas oil (VGO) hydrocracker and artificial neural network (ANN) method, a model is proposed to simulate yields of products including light gases, liquefied petroleum gas (LPG), light naphtha, heavy naphtha, kerosene, diesel and unconverted oil (off-test). The input layer of the ANN model consists of the catalyst, feed and recycle flow rates, and bed temperatures, while the output neurons are yields of those products. The results showed that the AAD% (average absolute deviation) of the developed ANN model for training, testing, and validating data are 0.445%, 1.131% and 0.755%, respectively. Then, by considering all operational constraints, the results confirmed that the decision variables (i.e., recycle rate and bed temperatures) generated by the optimization approach can enhance the gross profit of the hydrocracking process to more than $0.81 million annually, which is significant for the economy of the target refinery."

"Transforming seismic data into lateral sonic log properties was carried out successfully using the artificial neural network (ANN). This work is related to a detailed investigation of reservoir properties that requires complete data. The objective of this paper is to build a geological model that has vertical and lateral distribution representing the framework of geological change of sonic log properties. However, detailed well log data analysis only provides information of vertical distribution, therefore effective application of seismic data is required to construct a spatial distribution model that represents the lateral sonic log properties away from a well. This paper presents a strategy for transforming seismic data into pseudo-sonic log data by using ANN approaches rather than a simple approach of empirical relationship. The ANN approach defines a specific function that correlates a series of attributes generated from seismic data, such as amplitude envelope, instantaneous frequency, instantaneous phase, and acoustic impedance by a training mechanism based on the sonic log data as a target parameter. The probabilistic neural network (PNN) as one ANN algorithm is applied to transform seismic attributes into a lateral sonic log. An example of an ANN approach using a real data set from the Indonesian field was presented. The pseudo-sonic log shows a good agreement with the real sonic log data, which is represented with a correlation coefficient of 0.91. Further, the seismic line data was successfully transformed into the pseudo lateral sonic log data that covers the whole seismic line."

"In this research, a layered-recurrent artificial neural network (ANN) using the back-propagation method was developed for simulation of a fixed-bed industrial catalytic reforming unit called Platformer. Ninety-seven data points were gathered from the industrial catalytic naphtha reforming plant during the complete life cycle of the catalytic bed (about 919 days). Ultimately, 80% of them were selected as past horizontal data sets, and the others were selected as future horizontal ones. After training, testing, and validating the model with past horizontal data, the developed network was applied to predict the volume flow rate and research octane number (RON) of the future horizontal data versus days on stream. Results show that the developed ANN was capable of predicting the volume flow rate and RON of the gasoline for the future horizontal data sets with AAD% (average absolute deviation) of 0.238% and 0.813%, respectively. Moreover, the AAD% of the predicted octane barrel levels against the actual values was 1.447%, which shows the excellent capability of the model to simulate the behavior of the target catalytic reforming plant."

"Experience in applying a hybrid artificial neural network (ANN)-genetic algorithm for modeling and optimizing the Hall-Heroult process for aluminum extraction is described in this study. During the stage of modeling, the most important and effective process variables including temperature and cell voltage, metal and bath heights, purity of CaF2 and Al2O3, and bath ratio are chosen as input variables whilst outputs of the model are product purity, ampere efficiency, and product rate. During three years of operation, 19 points were selected for building and training, 7 points for testing, and 7 data points for validating the model. Results show that a feed-forward Artificial Neural Network (ANN) model with 3 neurons in the hidden layer can acceptably simulate the mentioned output variables with the Mean Squared Error (MSE) of 0.002%, 0.108% and 0.407%, respectively. Utilizing the validated model and multi-objective genetic algorithms, aluminum purity and the rate of production are maximized by manipulating decision variables. Results show that setting these decision variables at the optimal values can increase approximately the metal purity, ampere efficiency, and product rate by 0.007%, 0.185%, and 20kg/h, respectively."

"Road infrastructure includes toll roads developed to support mobility and economic activity. The toll road is part of the road network and is an alternative that can save travelers time and give them better service. The level of service of the toll road is strongly connected to the level of satisfaction of toll road users; therefore, customer satisfaction needs to be included in development models. The purpose of this study was to develop a model approach to customer satisfaction using an artificial neural network (ANN). Two models of customer satisfaction, SERVQUAL and Minimum Service Standards (SPM), have been used to modify the Toll Road Service Quality (TRSQ) model. This study has been able to explain that TRSQ has a value of R2, meaning the result is better than that of the other two models. The TRSQ model itself consists of seven dimensions: information, accessibility, reliability, mobility, security, rest areas, and responsiveness. Reliability is the dimension with the greatest effect on customer satisfaction."

"Pemanfaatan Artificial Intelligence (AI) terutama Machine Learning (ML) semakin banyak ditemui dalam berbagai hal termasuk pengambilan keputusan. Hal ini menimbulkan kebutuhan untuk memperoleh explanation dari prediksi model ML sebagai akuntabilitas dan kepercayaan terhadap sistem AI. Penelitian ini menggunakan abduction yang terdapat pada pendekatan logika untuk memperoleh minimal explanations yang valid secara formal dari suatu prediksi model Artificial Neural Network (ANN) berbasiskan Rectified Linear Unit (ReLU). Peneli-tian ini melakukan implementasi terhadap algoritma subset-minimal dan algoritma cardinality-minimal yang telah ada sebelumnya. Selain itu, penelitian ini mengajukan algoritma randomized-subset-minimal sebagai bentuk pengembangan dari kedua algoritma. Eksperimen menunjukkan bahwa algoritma randomized-subset-minimal dapat menghasilkan explanation dengan ukuran yang lebih kecil daripada algoritma subset-minimal, dengan waktu komputasi yang jauh lebih efisien daripada algoritma cardinality-minimal.Abstrak Berbahasa Inggris:

---

Artificial Intelligence (AI), especially Machine Learning (ML) is prevalent today in many donations, including for decision making. It raises the need for explanations of predictions by ML models to guarantee the accountability and trust of the AI system. This research exploits abduction from logic for obtaining minimal explanations of predictions by Artificial Neural Network (ANN) with rectifier activation function. This research implements both subset-minimal and cardinality-minimal algorithms for finding those explanations. Furthermore, this research proposes randomized subset-minimal algorithm for improving the algorithms. The experiment shows that the proposed algorithm is able to give explanations with a smaller size than the subset-minimal algorithm with computation time that much efficient than the cardinality-minimal algorithm.

"

"Pengukuran kadar gula darah merupakan salah satu kebutuhan utama dalam penanganan diabetes. Namun, moda pengukuran kadar gula darah yang umum saat ini, dilakukan secara invasive atau perlu melukai bagian tubuh manusia untuk mendapat nilai kadar gula darahnya. Terdapat metode pengukuran non invasive tanpa melukai manusia, namun metode ini masih belum dapat diandalkan karena banyaknya factor yang mempengaruhi glukosa tersebut. Penelitian ini mencoba untuk menganalisis akurasi dan performa dari pengukuran gula darah secara non invasive menggunakan sensor infrared pada panjang gelombang 940 nm dengan dibantu oleh Artificial Neural Network dan juga untuk mengevaluasi hubungan komponen dasar dari sinyal analog dari sensor yang bersangkutan terhadap kadar gula darah menggunakan Multiple Regression. Akurasi prediksi gula darah dievaluasi menggunakan Clark Grid Error analysis Dalam analisis ini, 81% dari 97 sampel data berada pada zona yang dapat diterima secara klinis, sedangkan sisanya berada pada zona yang tidak. Hal ini belum mencukupi kebutuhan akurasi 95% yang dapat diterima berdasarkan dari standar ISO 15197, maka hasil daripada penelitian ini masih belum memberikan hasil yang baik. Evaluasi menggunakan multiple regression sendiri menghasilkan hubungan yang tidak signifikan antara komponen dari sinyal analog dengan kadar gula darah dengan nilai R-squared sebesar 0.0174, RMSE 66.9, dan P-value keseluruhan sebesar 0.801.

---

Measuring blood sugar levels is one of the main needs in managing diabetes. However, the current common method of measuring blood sugar levels is carried out invasively or requires injuring parts of the human body to obtain blood sugar levels. There are non-invasive measurement methods without injuring humans, but this method is still not reliable because of the many factors that influence glucose. This research attempts to analyze the accuracy and performance of non-invasive blood sugar measurements using an infrared sensor at a wavelength of 940 nm assisted by an Artificial Neural Network and also to evaluate the relationship of the basic components of the analog signal from the sensor in question to blood sugar levels using Multiple Regression. The accuracy of blood sugar predictions was evaluated using Clark Grid Error analysis. In this analysis, 81% of the 97 data samples were in the clinically acceptable zone, while the rest were in the zone that was not. This does not meet the acceptable 95% accuracy requirement based on the ISO 15197 standard, thus the results of this research still do not provide relatively good results. Evaluation using multiple regression itself produced an insignificant relationship between the components of the analog signal and blood sugar levels with an R-squared value of 0.0174, RMSE 66.9, and an overall P-value of 0.801."

"Kebutuhan energi untuk rumah tangga atau bangunan di Indonesia sedang tumbuh secara signifikan. Oleh karena itu, efisiensi pada energi pendinginan sangat dibutuhkan. Penelitian ini bertujuan untuk mengembangkan model Artificial Neural Network (ANN) yang dapat memprediksi jumlah konsumsi energi pendinginan untuk pengaturan yang berbeda dari variabel kontrol sistem pendingin VRF. Bangunan dimodelkan oleh perangkat lunak Sketchup dan sistem pendinginan dimodelkan dengan EnergyPlus. MATLAB digunakan untuk training dan testing model ANN. Untuk model testing, set data dikumpulkan melalui simulasi yang sudah divalidasi dengan pengukuran lapangan. Empat langkah yang dilakukan dalam proses training yaitu pengembangan model awal, pemilihan variabel input, optimasi model, dan evaluasi kinerja. Model yang telah dioptimalkan menunjukkan akurasi prediksi yang akurat, sehingga membuktikan potensinya untuk aplikasi dalam algoritma kontrol yang diharapkan dapat menciptakan lingkungan termal ruangan yang nyaman serta energi yang efisien. Hasil analisis TOPSIS menunjukkan penghematan daya listrik sistem VRF sebesar 26% dari daya listrik observasi.

---

Energy needs for households or buildings in Indonesia are growing significantly. Therefore, efficiency in cooling energy is needed. This study aims to develop an Artificial Neural Network (ANN) model that can predict the amount of cooling energy consumption for different settings of the VRF cooling system control variable. The building is modeled by the Sketchup software and the cooling system is modeled by EnergyPlus. MATLAB is used for training and testing ANN models. For model testing, data sets are collected through simulations that have been validated with field measurements. The four steps involved in the training process are initial model development, selection of input variables, model optimization, and performance evaluation. The optimized model shows accurate prediction accuracy, thereby proving its potential for application in control algorithms that are expected to create a comfortable and energy efficient indoor thermal environment. The results of the TOPSIS analysis show that the VRF system's electrical power savings are 26% of the observed electrical power."

"Kegiatan Operasi dan Pemeliharaan Jaringan Pipa Gas (O & M) dilakukan untuk memastikan Infrastrtuktur Pipa Gas dapat berjalan dengan handal dan aman. Estimasi Biaya O & M yang digunakan untuk menyusun anggaran O & M Rutin, Non Rutin serta Pendukung O & M menggunakan pendekatan deterministik dimana telah diketahui komponen penyusun pekerjaan serta harga pembentuknya. Namun dari data tahun 2016-2019 terdapat realisasi atas biaya yang belum pernah dialokasikan seperti pekerjaan tidak terencana (unplaned) dan penanganan kondisi gawat darurat (emergency) dan mengambil alokasi anggaran yang telah direncanakan untuk kegiatan O & M. Realisasi biaya tersebut dapat diartikan sebagai Biaya Kontingensi O & M dimana biaya tersebut tidak dapat diprediksi dan dihitung sebelumnya namun terjadi. Dampak jangka Panjang dari diambilnya alokasi anggaran O & M yang sudah direncanakan dimana perusahaan belum mengalokasikan khusus anggaran kontingensi, dapat mengganggu Integrity dan keselamatan penyaluran gas.Untuk menjawab masalah tersebut diperlukan Pemodelan Estimasi Biaya Kontingensi Operasi dan Pemeliharaan yang digunakan untuk menentukan formula beban biaya kontingensi Operasi dan Pemelihaaan Jalur Pipa Gas Distribusi terhadap perencanaannya. Dalam Tesis ini dilakukan penelitian mengenai pengujian komponen variabel pembentuk biaya kontingensi Pengoperasian dan Pemeliharaan pipa gas seperti variabel Perbaikan Pipa, Relokasi Pipa, Sewa Lahan agar didapatkan suatu Pemodelan Estimasi Biaya kontingensi Operasi dan Pemeliharaan yang sesuai menggunakan Artificial Neural Network dengan bantuan Matlab Program. Dari hasil Pemodelan tersebut didapatkan formula model estimasi biaya kontingensi Operasi dan Pemeliharaan. Dari hasil estimasi biaya yang diperoleh, perusahaan dapat melakukan perencanaan anggaran yang lebih aktual dengan mengalokasikan biaya kontingensi secara khusus untuk dapat tetap memastikan kegiatan preventive, predictive maintenance dan corrective action dapat berjalan sesuai dengan siklus life time asset serta integrity dan keselamatan kegiatan operasional dapat dilaksanakan guna keberlangsung bisnis perusahaan serta memberikan pelayanan terbaik kepada pelanggan dan pemangku kepentingan.

---

Operation and maintenance (O & M) activities are carried out to guarantee that the Gas Pipeline Infrastructure can operate reliably and safely. The O&M Cost Estimation utilized to prepare Routine, Non-Routine, and O&M Supporter budgets employs a deterministic methodology in which the components of the job and their costs are known. However, the 2016-2019 data shows that costs that were never allocated, such as unanticipated work (unplanned) and dealing with emergency situations (emergency), as well as taking planned budget allocations for O&M operations, were realized. These costs can be regarded as costs when they are realized. O&M contingency expenditures cannot be forecast or anticipated in advance, yet they do occur. The long-term consequences of taking the scheduled O&M budget allocation and not allocating a special contingency budget can jeopardize the integrity and safety of gas distribution.

Model the Estimation of Operations and Maintenance Contingency Costs, which are used to establish the contingency cost load formula for Operation and Maintenance of Distribution Gas Pipelines against its plans, is required to solve this problem.

In order to obtain a suitable contingency cost estimation modeling using Artificial Neural Network with the help of the Matlab Program, a research was conducted on testing the variable components that form contingency costs for operation and maintenance of gas pipes, such as variable pipe repair, pipe relocation, and land rent.

A model formula for estimating Operations and Maintenance contingency costs is derived from the modeling findings. The company can plan a more actual budget based on the estimated costs obtained by allocating contingency costs specifically to ensure that preventive, predictive maintenance, and corrective action activities can run in accordance with the life time asset cycle, and the integrity and safety of operational "

"Kualitas berkas yang dinyatakan dalam besaran Half Value Layer (HVL) perlu diukur secara berkala pada saat proses uji jaminan mutu pesawat sinar-X. Penelitian ini mengembangkan sebuah model komputasi yang mampu memprediksi nilai HVL dari sebuah citra kosong radiografi umum untuk mengatasi permasalahan instalasi radiologi dalam pengadaan detektor radiasi. Model dibuat menggunakan teknik regresi Artificial Neural Network (ANN) dengan memanfaatkan nilai-nilai fitur yang dapat diekstrak dari sebuah citra medis sebagai masukan model. Tiga jenis model dibuat dengan memvariasikan jenis fitur masukan yang digunakan, yaitu fitur langsung, fitur tidak langsung atau properti Gray-Level Co-occurrence Matrix (GLCM), dan gabungan keduanya. Model dilatih menggunakan arsitektur dan hyper-parameters yang telah ditentukan hingga menghasilkan nilai galat terendah. Nilai HVL sebenarnya diukur menggunakan Solid-State Detector (SSD) dan digunakan untuk mengevaluasi performa model yang telah dilatih. Model Gabungan atau model dengan fitur masukan berupa gabungan antara jenis fitur langsung dan tidak langsung menghasilkan nilai performa terbaik. Pengujian performa Model Gabungan menggunakan data uji menghasilkan nilai mean absolute error, root mean squared error, dan mean absolute percentage error masing-masing sebesar 0,006, 0,009, dan 0,248%. Model harus diperlakukan sebagaimana detektor radasi pada umumnya sehingga proses akuisisi citra berulang perlu dilakukan. Perbedaan model pesawat sinar-X dan reseptor citra menghasilkan nilai dan pola fitur yang berbeda

---

The beam quality stated in Half Value Layer (HVL) value needs to be measured periodically during Quality Assurance (QA) X-rays device. This study develops a computational model that can predict the HVL value from a general radiography empty image to solve the problems of radiology installations in the procurement of radiation detectors. The model was created using Artificial Neural Network (ANN) regression technique by utilizing feature values that can be extracted from a medical image as model input. Three types of models were created by varying the type of used input features, those were direct features, indirect features or Gray-Level Co-occurrence Matrix (GLCM) properties, and combination of both. The model was trained using the predefined architecture and hyper-parameters until producing the lowest error value. The real HVL value was measured using a Solid-State Detector (SSD) and used to evaluate the performance of the trained model. Combined Model or a model with an input feature in the form of a combination of the types of direct features and indirect features produced the best performance value. The performance test of the Combined Model using the test data produced the mean absolute error, root mean squared error, and mean absolute percentage error value of 0,006, 0,009, and 0,248%, respectively. The model must be treated as a radiation detector in general so that the repeated image acquisition is necessary. Differences in the X-ray device and the image receptor model produce different feature values and patterns."