Process safety consequence modeling using artificial neural networks for approximating heat exchanger overpressure severity

Industry incidents happened more frequently when various processes safety and Risk Management activities associated with various sub-processes are not followed, tracked and address on a timely basis. Having a robust tracking system of mitigations plan help to follow up to closure identified corrective actions and consequently this data can be of the valuable support for any incident investigation. Cabinda Gulf Oil Company Limited (CABGOC) has developed over the years several Process Safety (PS) metrics involving Management of Change (MOC) Process, Risk Management (RM) studies recommendations, Loss of Containment (LOC), Safety System Function Test, Control System Data, and open Non-Conformance Report (NCR) related to discrepant material conditions. This manuscript will describe how these metrics are put together, the process and how the teams involved are functioning around that. It also describes the process of closure and the functional authority of each metrics governance. The sufficient information on the data analysis, inference and the utilization of the historical trend for the continual improvement of the targets on the risk mitigation and RM planning are discussed. The key challenges of building the Organizational Capability (OC) to sustain the processes and Process Safety Information (PSI) validation for an effective RM studies will be highlighted in the article. These PS metrics created a multi-disciplinary teams working on different tasks and tracking the actions leading to work culture improvement. CABGOC Leadership Teams does field verification and validation through General Manager of Operations (GMO), audits, Facility Engineering (FE)/ Heath Environment Safety (HES) field audits, operations top 10 critical safeguard verification list to authenticate the importance of verification and action tracking. Some of the key risk reduction benefits of the PS metrics tracking and monitoring will be deliberated in the article. The article will provide key lesson learned for future industry practices. PS Metrics help to track, trend, and have dialog about our high risk areas.

Industry over the last two to three years has put a lot of effort into implementing Process Safety Metrics. This is a good practice but, now that we have metrics data, how are we utilizing them to improve process safety performance? Air Products has been monitoring process safety performance for the last five years using combined Center for Chemical Process Safety and American Petroleum Institute RP 754 metrics. This article will highlight the approach we have applied, both through our lagging and leading indicators, to drive our process safety performance improvement programs. It will review the programs we initiated, leading indicators we identified, and how it also helped to improve our process safety culture. This article will also look at the importance of measuring severity and its impact in driving leadership behaviors. We will also discuss future opportunities to improve process safety performance via metrics analysis and follow‐up initiatives. © 2013 American Institute of Chemical Engineers Process Saf Prog 33: 172–178, 2014

Companies use metrics to monitor their process safety performance. Process safety metrics may relate to any aspect of a process safety program but many metrics are developed for individual process safety elements, such as process hazard analysis (PHA). Traditional metrics for PHA mostly address aspects of managing PHA studies, such as the timely completion of corrective actions. However, there are other types of metrics that are of interest, specifically those that characterize the level of safety achieved for a process and the level of quality achieved in a study. This article describes the use of such metrics in PHA and provides examples of their use.

The process-induced deformation (PID) during the manufacturing of thermosetting composite materials can significantly compromise manufacturing precision. This paper introduces an innovative method that combines a finite element analysis (FEA), feature classification algorithms, and an Artificial Neural Network (ANN) framework to rapidly predict the PID of a typical L-shaped structure. Initially, a comprehensive range of parameters that influence PID are compiled in this research, followed by the generation of a dataset through FEA considering viscoelastic constitutive models, validated by experimental results. Influential parameters are classified using Random Forest and LASSO regression methods, with each parameter rated according to its impact on PID, delineating their varying degrees of importance. Subsequently, through a hyperparameter analysis, an ANN framework is developed to rapidly predict the PID, while also refining the assessment of the parameters’ significance. This innovative approach achieves a computational time reduction of 98% with less than a 5% loss in accuracy, and highlights that under limited computational conditions, considering only a subset or all of the parameters—the peak temperature, corner angle, coefficient of chemical shrinkage, coefficient of thermal expansion, curing pressure, and E1—minimizes accuracy loss. The study demonstrates that machine learning algorithms can effectively address the challenge of predicting composite material PID, providing valuable insights for practical manufacturing applications.

Monitoring heart conditions through electrocardiography (ECG) has been the cornerstone of identifying cardiac irregularities. Cardiologists often rely on a detailed analysis of ECG recordings to pinpoint deviations that are indicative of heart anomalies. This traditional method, while effective, demands significant expertise and is susceptible to inaccuracies due to its manual nature. In the realm of computational analysis, Artificial Neural Networks (ANNs) have gained prominence across various domains, which can be attributed to their superior analytical capabilities. Conversely, Spiking Neural Networks (SNNs), which mimic the neural activity of the brain more closely through impulse-based processing, have not seen widespread adoption. The challenge lies primarily in the complexity of their training methodologies. Despite this, SNNs offer a promising avenue for energy-efficient computational models capable of displaying a high-level performance. This paper introduces an innovative approach employing SNNs augmented with an attention mechanism to enhance feature recognition in ECG signals. By leveraging the inherent efficiency of SNNs, coupled with the precision of attention modules, this model aims to refine the analysis of cardiac signals. The novel aspect of our methodology involves adapting the learned parameters from ANNs to SNNs using leaky integrate-and-fire (LIF) neurons. This transfer learning strategy not only capitalizes on the strengths of both neural network models but also addresses the training challenges associated with SNNs. The proposed method is evaluated through extensive experiments on two publicly available benchmark ECG datasets. The results show that our model achieves an overall accuracy of 93.8% on the MIT-BIH Arrhythmia dataset and 85.8% on the 2017 PhysioNet Challenge dataset. This advancement underscores the potential of SNNs in the field of medical diagnostics, offering a path towards more accurate, efficient, and less resource-intensive analyses of heart diseases.

本研究設計一個分散式類神經網路框架以處理巨量資料之即時分析並能在極短的時間內得到不錯的結果。我們的實驗結果顯示在24 核心叢集平台上訓練分散式類神經網路模型可於17 秒收斂，進行預測時在0.7 投票閥值（voting threshold）設定下採用分層多重模型（multi-model with stratification）可獲得最多的真陽性結果且準確率達70%左右。在我們所建構的系統裡，類神經網路是用在資料採礦階段來發掘金融時間序列資料之模式。我們將訓練類神經網路的框架建置在分散式運算平台上，該平台我們採用具高效能記憶體內運算（in-memory computing）的Apache Spark 來建造底層基礎的運算叢集環境。我們評估了一些特別適用於預測金融時間序列資料的分散式後向傳導演算法，加以調整並整合進我們所設計的框架。同時，我們也提供了許多細部的選項，讓使用者在進行類神經網路建模時能有很高的客製化彈性。

Risk assessment and risk reduction of an acrylonitrile production plant

This work provides an advanced prototype for pipeline and wellbore tube inspection using electromagnetic (EM) resonance coupling, electromagnetic (EM) coupling, and machine learning. Utilizing only two transmitters and eight sensor coils, the described device can detect and characterize inner, outer, and total metal loss in a pipe's body. A defect in the pipe body alters the impedance of the transmitter and receiver coils and the mutual coupling between them, resulting in a relative change in one or more Rx outputs. A framework for artificial neural networks (ANN) is designed to assess the eight outputs and generate a two-dimensional map of the pipe cross-section. The ANN is trained using a finite-difference time-domain electromagnetic forward solver. The designed prototype is tested and validated through simulations and an experimental setup. The results demonstrate that the tool, with the assistance of the ANN, could not only detect single and double flaws with either complete or partial metal loss, but also define the defect's size, location, and depth.

We review the structure of cerebral cortex to find out the number of neurons and synapses and its modular structure. The organization of these neurons is then studied and mapped onto the framework of an artificial neural network (ANN). The computational requirements to run this ANN model are then estimated. The conclusion is that it is possible to simulate the mouse cortex today on a cluster computer but not in real-time.

Environmental sustainability in the city of Shanghai municipal solid waste treatment system: an integrated framework of artificial neural network (ANN) and LCA-emergy methodology

Protein interactions are essential in discovery of protein functions and fundamental biological processes. In this paper, we aim to create a reliable computational model for protein interaction prediction by integrating information from complementary data sources. An integrative Artificial Neural Network (ANN) framework is developed to predict protein-protein interactions (PPIs) from heterogeneous data in Human. Performance of our proposed framework is empirically investigated by combining protein domain data, molecular function and biological process annotations in Gene Ontology. Experimental results demonstrate that our approach can predict PPIs with high sensitivity of 82.43% and specificity of 78.67%. The results suggest that combining multiple data sources can result in a 7% increase in sensitivity compared to using only domain information. We are able to construct a protein interaction network with proteins around mitotic spindle checkpoint of the human interactome map. Novel predictions are made and some are supported by evidences in literature.KeywordsProtein-Protein InteractionInteraction Network PredictionHeterogeneous Data IntegrationIntegrative Neural Network

This article briefly summarises the author's research on automatic performance, started at CSC (Centro di Sonologia Computazionale, University of Padua) and continued at TMH‐KTH (Speech, Music Hearing Department at the Royal Institute of Technology, Stockholm). The focus is on the evolution of the architecture of an artificial neural networks (ANNs) framework, from the first simple model, able to learn the KTH performance rules, to the final one, that accurately simulates the style of a real pianist performer, including time and loudness deviations. The task was to analyse and synthesise the performance process of a professional pianist, playing on a Disklavier. An automatic analysis extracts all performance parameters of the pianist, starting from the KTH rule system. The system possesses good generalisation properties: applying the same ANN, it is possible to perform different scores in the performing style used for the training of the networks. Brief descriptions of the program Melodia and of the two Java applets Japer and Jalisper are given in the Appendix. In Melodia, developed at the CSC, the user can run either rules or ANNs, and study their different effects. Japer and Jalisper, developed at TMH, implement in real time on the web the performance rules developed at TMH plus new features achieved by using ANNs.

Fouling of feed/effluent heat exchanger in Hydrotreating unit have a significant impact on the operation efficiency. Hydrotreater fouling presents loss of heat transfer, enlargement of energy consumption, increasing in maintenance cost and throughput limitation. Thus, to solve these problems, heat exchangers are cleaned during shutdown. In this research, an artificial neural network has been used to develop the fouling model of feed/effluent heat exchanger in Hydrotreating unit. Developed algorithm is employed to determine the optimal operation and cleaning interval of the feed/effluent heat exchanger surfaces. The artificial neural network is widely applied in modeling and optimization of chemical process especially nonlinear systems. The main advantage of the use of artificial neural network is obtaining a highly accurate mathematical model of the system without the detail of the system. The process modeling applications use the artificial neural network to approximate the relationship between the input and output variables.

The work aim is to predict the thermal state of the air-ground heat exchanger based on an artificial neural network. Training, testing and validation of the proposed model were made on experimental data obtained in the thermophysical laboratory of the Institute of Engineering Thermophysics of the National Academy of Sciences of Ukraine. A simple neural network is used in this work. The air temperature at the inlet to the heat exchanger, and its relative humidity are selected as input parameters for the neural network. The MATLAB (R2016a) and Levenberg-Markwatt model were used in this article's calculations. One hidden layer and 10 neurons were presented in the model. The array of analysed data was divided into ratios of 70%, 15%, 15% for neural network training, validation and testing, respectively. As a result, it is obtained that the forecasting takes place with acceptable accuracy in all models. The root mean square error for the whole data set for different models varies from 0.105 to 2.323°С. The maximum mean absolute percentage error was the largest for CFD model and was 11.2%. The minimum mean bias error of the predicted data from the experimentally measured ones was found in the model using temperature, humidity, and air temperature at the outlet of the air-ground heat exchanger for the previous hour and was 0.02%. The training and testing of the proposed models based on an artificial neural network are satisfactory enough to predict the temperature taking into account the influence of weather conditions. Artificial neural networks can be used to predict the thermal state of the air-ground heat exchanger. Data representing the description of a real system are required for forecasting the parameters based on the ANN.

The objective of this work is the development of a highly effective Wiener-Physically-Informed-Neural-Network (W-PINN) modeling methodology for systems and processes with highly nonlinear dynamic and highly nonlinear static behavior. This approach has two basic stages. The first stage estimates all the unknown linear dynamic modeling coefficients. With these estimates fixed, the second stage estimates all the unknown static coefficients in an artificial neural network (ANN) framework. When the ANN is a nonlinear function, as in this work, the fitted model is a nonlinear dynamic and nonlinear static structure. In a previous one stage, 60-minute forecast modeling case, the dynamic outputs (vij, where i = the subject number and j = the sample number) were estimated in Excel® for eleven (11) type 1, freely-living, diabetes data sets (cases) of approximately two weeks of data, yielding an average input-only sensor glucose concentration (SGC) validation fit () of 0.68. With these vij’s fixed, this new approach obtains final, two-stage (TS) W-PINN, 60-minute forecast fitted models, for each of the 11 cases, in two ways. One two-stage approach uses the JMP® ANN toolbox for second stage modeling and achieves an average input-only SGC of 0.74 and maximum of 0.84. The second two-stage approach uses a novel ANN methodology coded in Python® and achieves an average input-only SGC of 0.82 and maximum of 0.93. Incorporating bias correction, using current and past SGC residuals, the Python® estimator improved the average from 0.82 to 0.87 with the maximum still 0.93.