Book summary “Transactions on Engineering Technologies”: summary

Introduction

You face a rapidly evolving landscape where mechanical design, computing science, and industrial applications intersect. Modern engineering projects demand advanced mathematical modeling, rigorous safety standards, and efficient algorithms to handle complex multi-physics systems. When you attempt to bridge theoretical research and practical industrial implementation, you often encounter fragmented methods and steep optimization hurdles. You need a unified reference that captures peer-reviewed breakthroughs across multiple technical disciplines.

This volume compiles selected extended papers from the International MultiConference of Engineers and Computer Scientists. The central idea is that cross-disciplinary engineering technologies share common underlying mathematical and computational principles. By examining rigorous solutions in fluid dynamics, circuit design, computer vision, and supply chain management, you can apply proven frameworks to your own technical challenges.

The book is structured into twelve comprehensive sections that address distinct engineering domains. You will explore advanced control systems, safety standards, security frameworks, and computational algorithms. The chapters guide you through specific industrial applications and theoretical models without unnecessary abstraction.

• How do you stabilize underwater vehicles and accelerate multi-physics simulations?
• What methods ensure machinery safety and manage distributed energy storage?
• How can you optimize asynchronous circuits and deploy automotive soft sensors?
• What techniques improve tourist information retrieval and privacy-preserving learning?
• How do you enhance transportation safety and segment color images?
• What algorithms improve image thresholding and automated quality inspection?
• What text mining models and aspect-based dictionaries power recommendation systems?
• What tools translate sign language and support historical thinking?
• How do you achieve cross-language entity linking and secure Industrie 4.0 systems?
• What factors influence dual-task performance and biodiesel reactor control?
• How can you manage scheduling disruptions and optimize perishable supply chains?
• What forces shape online shopping attitudes, airport public-private partnerships, and asphalt adhesion?

1. Underwater Vehicle Stabilization and Multi-Physics Acceleration

Descent underwater vehicles connected to carrier ships by elastic ropes require specialized stabilization systems to prevent resonance, rope breakage, and bottom collisions caused by sea disturbances. You must manage complex parameters when equipment operates at extreme ocean depths.

Specialized stabilization systems protect descent underwater vehicles from sea disturbances, while Quasi-Newton methods accelerate partitioned multi-physics simulations.

The stabilization systems utilize a shock-absorbing hoist controlled by different input signals.

• A measuring converter of the rope length measures baseline depth.
• A sensor of rope tension deviation detects dynamic forces.
• A combination of both inputs improves overall feedback.

Incorporating a rope tension deviation sensor with a PI-controller increases the overall transfer coefficient and improves stabilization accuracy compared to using a measuring converter alone. Researchers carried out simulations of the descent underwater vehicle location stabilization at a depth of 6000 m. Robust control approaches based on parametric synthesis and interval parameters can handle affine uncertainty and maximize the robust degree of stability in these systems.

Partitioned multi-physics codes couple separate mathematical models without modifying underlying black-box solvers. Non-linear Gauss-Seidel iteration loops often converge slowly unless you apply acceleration techniques.

• Type I methods approximate the Jacobian directly.
• Type II methods approximate its inverse.
• Least Change Secant Update methods use past iteration states.

Re-using the final approximate Jacobian from the previous time step as the initial guess for the next time step significantly reduces the number of coupling iterations required for convergence in transient problems.

Simulations coupling OpenFOAM for fluid flow and Calculix for structural deformation were conducted across multiple transient fluid-structure interaction benchmarks. The fluid domain in the dynamic piston-channel test problem measured 10 m in length. The dam break simulation utilized 3670 linear fluid elements and 14 quadratic finite-elements. The 3D flexible tube problem employed 600 quadratic solid elements and 6000 linear fluid flow elements.

Simulations were run using simplified models representing wave equations, Fokker-Planck equations, and diffusion equations for tokamak plasma heating. Systems that reached 100 as the maximum number of iterations after which non-convergence or divergence was recorded benefited directly from these advanced acceleration methods.

2. Machinery Safety Standards and Energy Management

The transition from the deterministic standard EN 954-1 to the statistical standard EN ISO 13849-1 introduced complex parameters, causing confusion among machine designers who often rely on automated tools without understanding underlying calculations. You must navigate metrics like MTTFd, DC, and CCF when replacing deterministic categories with this combined deterministic and statistical approach.

Implementing EN ISO 13849 safeguards high-risk industrial machinery, while bidirectional DC/DC converters manage distributed energy storage systems.

Machinery Safety Standards

Tyre manufacturing machinery presents severe safety risks around pinch and rotation points, particularly during the tyre building process where automation requires stringent risk reduction measures. You evaluate risk graph parameters S2, F2, and P1 using safety reviews conducted every five years.

Machine designers in the tyre industry struggled to interpret statistical values like PFHDavg when the standard replaced EN 954-1 in December 2011.

Upgrading an emergency stop system from Category 1 to Category 3 with redundant contactors and a monitoring safety relay significantly improves safety performance to meet higher required levels. You integrate components such as:

• emergency stop device E1;
• redundant contactors K1 and K2;
• a monitoring safety relay MSR1.

Integrating safety light curtains as a complementary protective measure in front of automatic building units prevents unexpected machine start-ups when operators or maintenance staff access hazardous zones. The light curtain device LC1 achieves a PFHD of $7.93 * 10^{-9}$ per hour, ensuring safety-related parts of control systems meet PLe performance.

Energy Management Fundamentals

The inherent power fluctuations of renewable energy generation systems necessitate the integration of battery energy storage systems with real-time charging and discharging capabilities to maintain grid stability. You offset the phase-out of nuclear energy and reduce greenhouse gas emissions by utilizing distributed power generation across regions like Taiwan.

Dual active full bridge phase shift DC/DC converters provide electrical isolation, high voltage gain, and fast real-time current regulation for battery energy storage systems.

Implementing a fully digital control scheme on PSIM software and a hardware setup using a Texas Instruments digital signal processor as the core controller enables this power flow management. The DAFBFS hardware test system operates with a $1.2\text{ kW}$ power rating, a $200\text{ V}$ DC bus voltage, and a $48\text{ V}$ rated voltage for the vanadium redox flow battery.

3. Asynchronous Circuits and Automotive Soft Sensors

Asynchronous control circuits under bounded gate and wire delay assumptions require explicit acknowledgment of transitions on wire forks. A distributed lock relation ensures that candidate transitions of a multiple-cycle signal are fully locked with two other signals. This allows the identification of a volunteer signal that acts as a tardy acknowledgement of the completion path.

Distributed lock relations optimize asynchronous control circuits, and radial basis function neural networks provide analytical redundancy for spark-ignition engines.

Asynchronous Control Circuits

Scalable-Delay-Insensitive optimization based on distributed lock relations significantly improves the area cost of asynchronous control circuit implementations compared to Quasi-Delay-Insensitive models. By eliminating the volunteer signal in partial transition graphs, circuits can be synthesized with fewer logic gates and literals. Experimental benchmarks demonstrate that the method yields smaller circuit implementations and reduces literal counts and gate overhead.

In experimental comparisons for the ebergen benchmark, the presented method achieved a literal count of 4 with 3 gates and 2 c-elements, whereas the Quasi-Delay-Insensitive implementation required 6 literals and 5 gates.

Automotive Soft Sensors

Autonomous vehicle control systems require high reliability and redundancy. Soft sensors provide analytical redundancy for hardware sensors by mapping measurable variables such as throttle angle, fuel injection rate, and manifold pressure and temperature to unobserved engine states.

The Mean Value Engine Model simulates spark-ignition engine dynamics by dividing them into intake manifold filling, crankshaft speed, and fuel injection subsystems. The model incorporates manifold pressure and temperature dynamics based on air mass conservation, crankshaft rotational energy conservation, and wall-wetting fuel flow dynamics characterized by evaporation time constants and fuel deposition proportions. The mean value engine model validated using real-time data from a Ford 1.6 L engine test bed includes three distinct subsystems modeled independently.

Radial basis function neural network-based soft sensors can accurately monitor air-fuel ratio and engine speed in spark-ignition engines under moderate computational loads. Using a dataset of 5000 samples where the first 4000 were used for training and 1000 for validation, radial basis function neural networks achieved mean absolute errors of 0.008 for air-fuel ratio and 0.036 for engine speed.

4. Tourist Information Retrieval and Privacy-Preserving Learning

You can search for tourist spots based on functional features rather than metadata tags by utilizing paragraph vectors derived from user reviews. Users often search for spots with specific functions such as a meeting point or night view spot regardless of location or category tags.

Paragraph vectors enable functional tourist spot searches, and secure multiparty computation facilitates privacy-preserving fuzzy modeling on cloud systems.

By combining distributed representations of tourist spots, locations, and categories using vector arithmetic, you can generate a query vector representing a functional feature in a new location. The Jalan dataset contains 43,759 spots, 1,481,831 reviews, and 30 dimensions used in the paragraph vector model.

For instance, finding a meeting point in New York similar to the Hachiko statue in Tokyo is achieved by evaluating vector expressions like Hachiko minus Tokyo special wards plus Osaka city.

Using the average vector of paragraph vectors for all spots sharing a specific location or category tag yields more accurate query vectors than treating all reviews of a tagged location as a single document. Directly aggregating all reviews under a single tag as one document biases the vector toward dominant spots within that category or location.

Cloud Computing and Privacy-Preserving Machine Learning

Traditional data encryption requires expensive continuous client-side encryption and decryption. Secure multiparty computation combined with fuzzy modeling enables privacy-preserving machine learning and data mining on cloud systems without heavy encryption complexity.

Secure multiparty computation divides data across multiple independent servers using addition and multiplication forms so that individual servers cannot see the complete data or parameters, while a client integrates the final outputs. Random real numbers are selected from a 0 to 1 range.

Simulations of two-category classification on spheres and an obstacle avoidance control problem were performed to demonstrate that divided learning data and parallel algorithms yield results comparable to non-divided conventional methods. The classification experiments utilized 512 learning data points and 6400 test data points.

5. Transportation Safety and Color Image Segmentation

Computer vision techniques enhance railway obstacle detection, and rho-theta space transformations streamline color blindness image segmentation.

Railway Safety and Obstacle Detection

You can detect railway obstacles in real time using a computer vision approach with prior knowledge and bottom-up adaptive windows. The method initializes the rail gap and starting position, applies a Sobel operator for gradient images instead of Canny to avoid threshold issues, and creates adaptive sliding windows that adjust size and position based on rail deflection and local gradients.

Tested on an open-source video dataset captured from a train traveling between Malmo and Angelholm in Sweden, with digital obstacles added via Adobe Premiere Pro CC 2015, the system ran with a 640 x 360 frame size, 16:9 aspect ratio, and 25 fps frame rate during a 2 min recording test duration, achieving 10 out of 10 obstacles successfully identified.

Color Image Segmentation

Converting three-dimensional RGB color space into a two-dimensional rho-theta space transforms color image segmentation into detecting cloud-like peaks. By treating the white color point as a reference anchor, any color pixel vector can be parameterized by two angles representing its included angles with the B-axis and R-axis, causing grouped color line distributions to gather tightly around 255, 255, 255 values for the white anchor point reference.

Color plane segmentation can be achieved by performing local maxima peak detection and equivalent relationship grouping within the rho-theta parameter space. A sliding window detects local maxima, small peaks are removed via a threshold filter, and an equivalent relationship table based on pixel value continuity along line segments between peaks determines which peaks belong to the same cloud.

Illustrated using Ishihara color blindness test plates containing size-varied color dots:

• eleven peaks were initially detected from the transformed image of an Ishihara plate;
• peaks were subsequently grouped into six final peaks and resulting color planes.

Background and pattern in color blindness images can be segregated by computing spatial distances among color planes followed by K-means clustering. Euclidean spatial distances are calculated between non-zero pixels across different binary color planes to build a symmetrical distance matrix, which is then inputted into the K-means algorithm with two clusters, where K equals 2 for background and pattern clustering.

6. Image Thresholding, Quality Inspection, and Forensic Identification

Global image thresholding can be accurately formulated as a statistical change-point detection problem using Kullback-Leibler divergence without assuming prior probability distributions.

Statistical change-point detection improves global image thresholding, computer vision automates beverage quality checks, and advanced feature extraction identifies bodies.

Image Processing and Computer Vision

By interpreting the estimation of an optimal global threshold as a change-point detection (CPD) problem, you avoid traditional assumptions like normal distributions for background and foreground classes. The algorithm is robust against outliers and heavily unbalanced class areas through the application of a symmetric Kullback-Leibler (KL) divergence measure.

Tested on various images including Dice, Rice, Object, and Number plate, the proposed change-point detection method outperformed classical algorithms like Otsu's and Kittler-Illingworth, yielding higher Structural Similarity Index scores. A 0.9907 SSIM score was achieved by the CPD algorithm on the Dice test image.

A new performance evaluation criterion for image thresholding algorithms can measure dissimilarity without requiring any ground truth image. The proposed threshold evaluation criterion calculates the dissimilarity between foreground and background binary patterns using normalized Manhattan distance applied to binary morphological gradients. A well-segmented image yields a normalized Manhattan distance close to one.

The normalized Manhattan distance was used to evaluate multiple thresholding algorithms across images such as Dice, Rice, Object, Denise, Train, and Lena without ground truth comparison.

Industrial Automation and Quality Inspection

An automated computer vision system using an IP camera and quadratic distance classifier can inspect beverage product color concentration and water level with high accuracy. The system acquires images via a wireless IP camera, converts RGB color models to HSV color space, uses Otsu's method on the saturation component for segmentation, and applies a quadratic distance classifier on combined RGB histograms to evaluate color concentration.

Soft drink beverages including tropical, strawberry, orange, root beer, and grape flavors were tested using 246 total samples on a conveyor setup. The system achieved 98 percent accuracy using 246 samples, with the IP camera positioned at a 30 centimeters distance from the front of the bottle, utilizing 30 bins combined in the RGB histogram and a 0.2 threshold value for color pass and reject.

Human bodies after mass disasters can be accurately identified by matching antemortem chest X-ray radiographs with postmortem CT scans using advanced feature extraction. The proposed identification method addresses cases where traditional biometrics like fingerprints or dental records fail. It extracts features from the stable parts of ribs using Histograms of Oriented Gradients, Bag of Words, and Markov Feature Vectors, followed by Euclidean distance ranking.

During the Great East Japan Earthquake, 15,824 dead bodies were examined and 15,749 were identified. The method was tested using a database of 27 subjects from the Miyazaki University School of Medicine, evaluating performance against postmortem CT scan query images, utilizing 2,470 by 2,470 resolution for chest radiographs, 256 by 256 resolution for CT scanned images, and achieving 63.0 percent accuracy using steady MFV and BoW features.

7. Text Mining Path Models and Cosmetics Recommender Systems

Text corpora often present low term frequencies that cause insignificant paths in topic and structural equation models. Pre-processing text corpora with word2vec before applying topic modeling and structural equation modeling helps solve the problem of low term frequencies and insignificant paths.

You can calculate cosine similarities between words using a two-layer neural network. Words exceeding a similarity threshold are equalized to artificially increase term frequencies. This prevents zero-frequency problems and stabilizes factor analysis models.

Researchers tested this approach on 754 live project data from Kickstarter collected on April 28, 2015, across 11 categories such as 3D printer, apps, and hardware.

The optimal similarity threshold for word2vec word equalization balances abstraction and keyword precision. The optimal similarity threshold for word2vec word equalization in structural equation models balances abstraction and keyword precision.

When similarity thresholds are set too low, unrelated words group together, causing observed variables to lose touch with the model. Empirical evaluations using GFI, AGFI, and RMSEA indices show that a threshold score around 0.8 frequently yields the best model evaluations.

In the hardware category analysis, models with a threshold value of 0.8 achieved superior GFI and AGFI indices and lower RMSEA scores compared to conventional models without word2vec.

Cosmetics Recommender System

Review recommender systems for personal products require automatic scoring of fine-grained aspects because user compatibilities differ significantly based on skin type and concerns. Review recommender systems for personal products like cosmetics require automatic scoring of fine-grained aspects because user compatibilities differ significantly based on skin type and concerns.

Because unsuitable cosmetics frequently cause skin irritations, general review scores are insufficient. The method utilizes an evaluation expression dictionary built from a corpus of online reviews to automatically score multiple aspects of product reviews.

The study analyzed data from '@cosme', a popular portal for beauty and cosmetics among young Japanese women, which recorded 280 million monthly page views, 3.9 million members, and 1,300 million total reviews in October 2016.

Automatic scoring of review texts can be achieved by extracting phrasal evaluation expressions, dividing them into aspect keywords, feature words, and degree words, and assembling co-occurrence relations. Evaluators manually scored face lotion reviews on a scale from 0 to 7 stars across 10 predetermined aspects to compare against the automated dictionary-based scoring method.

Tuning the evaluation expression dictionary through targeted user experiments significantly improves the accuracy and achievement rate of automated review scoring. Research participants consisted of 37 females aged 19 to 25 years who tested 15 kinds of face lotions, wrote reviews, and scored all aspects to help tune the dictionary.

Calculating user similarity for collaborative filtering in a cosmetics recommender system is more effective when grouped by clusters of similar product ratings across multiple product aspects.

8. Sign Language Translation and Historical Thinking Support

Leap Motion Controllers and finite automata translate sign language, while interactive software support systems foster historical thinking skills.

Thai Sign Language Translation

A computer program can translate Thai sign language into Thai alphabets using a Leap Motion Controller and Quadtree spatial structures combined with finite automata. The Leap Motion Controller detects the positions of five fingertips and the palm center on X and Y axes. A decision tree is created via the Quadtree technique to continuously divide a two-dimensional square into equal parts, and finite automata models process the finger codes to recognize alphabets.

Data normalization and position shifting are necessary to adjust for different hand sizes between a master user and an examiner during sign language recognition. The system calculates the Euclidean distance from the middle fingertip to the palm center to establish a ratio that corrects for varying hand dimensions. Additionally, the palm center position is shifted to the midpoint of the Quadtree table as a fixed reference point.

Researchers tested the system on 42 Thai sign language alphabets, handling 1-step, 2-step, and 3-step fingerspelling movements. The test resulted in 78.70% accuracy for overall performance.

Historical Thinking and Story Creation

Historical thinking is a reasoning skill used to analyze past events, understand human intentions, and infer future outcomes by drawing similarities to modern situations. Students typically memorize historical facts rather than acquiring reasoning skills. To improve this, a learning method involving story creation requires learners to derive problem situations by applying historical solutions to modern contexts.

Applying a historical solution to a modern problem requires abstracting the historical event, identifying similarities, and specializing the abstraction. You must first abstract the problem and solution from the historical event, discover economic or structural similarities with a modern situation, and then specialize the generalized solution to fit the modern problem.

The Kamakura period Tokusei Edict canceled samurai debts. This historical action can be abstracted into voiding lending contracts and applied to a modern friend suffering from a student loan.

Stories representing historical and simulated events are composed of characters, character properties, relationships between characters, and events that alter states. Characters possess properties such as money or power, forming a specific state. Relationships define how state changes in one character affect another, while events act as triggers that modify these states.

A specialized software support system provides history and story databases alongside an interface for creating characters, events, relationships, and scenes. The support system allows you to define character types and initial states, build events that alter properties, establish relational rules, and sequence them in a scene tab to visualize state transitions.

9. Cross-Language Entity Linking and IT Security in Industrie 4.0

Cross-language entity linking adapts to user language proficiency, and dedicated security frameworks protect industrial control systems in Industrie 4.0.

Cross-Language Entity Linking

Cross-language entity linking automatically discovers valuable keyphrases in Japanese documents. You can link these keyphrases to related Chinese Wikipedia pages based on your language proficiency.

The method extracts noun keyphrase candidates using top consecutive nouns cohesion. It judges their difficulty level via the Japanese-Language Proficiency Test vocabulary. Word2vec calculates word vector relevancies when a keyphrase does not exist in the JLPT vocabulary list.

Extracting the word Trump from a news article finds its most similar words in the JLPT vocabulary list as President and head of state with scores of 0.624 and 0.596 respectively.

Multiple translation methods prevent missing appropriate translations:

• Google Translate;
• Bing Translator;
• character code conversion method.

Researchers evaluated the approach using 10 randomly selected news articles from Mainichi Shimbun in 2014. 18 Chinese students participated in the questionnaire survey, possessing N2 or N1 proficiency. The system extracted 27 out of 38 correct keyphrases and achieved 0.81 accuracy in acquiring relevant Chinese candidate articles.

IT-Security Framework

The convergence of cyber-physical systems and internet technologies in Industrie 4.0 introduces severe IT security vulnerabilities. Traditional industrial control systems were never designed for connected operation. Companies must replace security by obscurity with production IT security strategies.

Platform Industrie 4.0 announced 17 technology development areas for cyber-physical production systems. VDMA represents over 3,000 mostly medium-sized companies facing these integration challenges. You must adopt a Defense in Depth strategy alongside Security by Design principles.

Defense in Depth deploys layered protective measures across separated system areas. Security by Design integrates threat analysis and risk assessment into early development phases.

A developed process and data model algorithm generates the Industrie 4.0 Toolbox IT-Security catalog. A Generic Procedure Model to introduce Industrie 4.0 in Small and Medium-sized Enterprises guides implementation. A sensor network application level protecting confidentiality against eavesdropping returns development stages such as AES, Public-Key Encryption, and Random-Key Predistribution.

10. Human Factors in Dual-Tasks and Biodiesel Reactor Control

Spatial compatibility and tactile precues affect dual-task performance, and nonlinear model-based control optimizes batch transesterification reactors.

Human-Machine Interface and Precuing Effects

Complex dual-task environments require operators to manage a primary continuous task alongside a secondary discrete task. Precuing spatial locations and matching spatial stimulus-response orientations significantly affects human reaction times and performance.

Tactile and auditory precues do not overload visual perception resources. They leave more mental capacity available for secondary tasks.

Thirty-six Chinese students from City University of Hong Kong participated in an experiment using a 17-inch LCD monitor to perform a primary pursuit manual tracking task paired with a secondary spatial compatibility task. The participants ranged in age from 20 to 24 years with a median of 20, comprising 21 males and 15 females.

Tactile precues outperform visual precues in reducing operator reaction times when responding to visual stimuli across various spatial mappings:

• tactile precues under BC S-R mapping produced the fastest reaction time of 501 ms;
• visual precues under TC S-R mapping resulted in the slowest reaction time of 645 ms;
• tactile precues yielded 10.978%, 14.572%, 11.723%, and 16.216% faster reaction times for BC, BI, LC, and TC mappings respectively.

The tracking error in primary continuous tasks increases substantially with higher task difficulty levels. Participants completed manual tracking tasks at three distinct difficulty levels, yielding an average root mean square tracking error of 24.74 pixels for low difficulty, 32.76 pixels for medium difficulty, and 45.95 pixels for high difficulty.

Biodiesel Production and Reactor Control

Maximizing biodiesel production in batch transesterification reactors requires off-line optimization of temperature set points and robust nonlinear model-based control.

Transesterification proceeds through three sequential reversible steps converting triglycerides to diglycerides, monoglycerides, and glycerol. Researchers at Naresuan University modeled a batch reactor operating under atmospheric pressure using vegetable oil, methanol, and homogeneous alkaline catalysts. The reaction system features a 6:1 molar ratio between alcohol and triglycerides, a 0.3226 mol/L initial concentration of triglyceride, a 1.9356 mol/L initial concentration of methanol, a 450 kJ/min K overall heat transfer coefficient UA, and a 1 m3 reactor volume V.

A sequential optimization approach utilizing control vector parameterization and Runge-Kutta Fehlberg integration determines optimal operating profiles. Because transesterification kinetics exhibit nonlinear, time-varying behavior, an extended Kalman filter estimates uncertain parameters like the overall heat transfer coefficient and heat of reaction. This allows the model predictive controller to outperform traditional PID controllers under plant-model mismatch. Simulations testing a 30% reduction in the overall heat transfer coefficient with a 4 seconds sampling time and a 10 prediction horizon demonstrated that model predictive control paired with an extended Kalman filter successfully tracked the optimal trajectory.

11. Project Scheduling Disruptions and Perishable Supply Chains

Project scheduling is heavily exposed to uncertainty that causes disruptions during execution. While mass production effectively manages disruptions due to their high frequency, project management literature on reactive scheduling for the multi-mode resource constrained project scheduling problem is very scant.

Proactive-reactive scheduling frameworks handle uncertainty in resource-constrained projects, and stochastic models optimize perishable supply chains.

Angela Hsiang-Ling Chen, Yun-Chia Liang, and Jose David Padilla proposed a scheduling framework for the multi-mode resource constrained project scheduling problem based on production scheduling methods.

Production environments utilize various structured methods to handle real-time uncertainty and disruptions:

• stochastic scheduling treats scheduling as a multi-stage decision process at stochastic decision points to minimize makespan;
• rolling horizons simultaneously consider short-term and long-term planning periods;
• decomposition splits the planning horizon into hierarchical levels to solve temporal and technological constraints.

Proactive baseline scheduling relies on fuzzy set theory, robust optimization, and entropy functions to build schedules capable of absorbing variations. Fuzzy programming treats random parameters as fuzzy numbers and constraints as fuzzy sets to handle imprecision. Robust optimization finds solutions that remain close to the optimum or feasible across various scenarios.

Reactive scheduling responds to disruptions through specific strategies:

• schedule repairing actions quickly push affected activities forward until resources become available;
• rescheduling involves re-sequencing affected activities using minimum perturbation or match-up points.

Supply Chain Design for Perishable Products

Supply chain design for perishable products requires a stochastic mathematical model that accounts for demand uncertainty and route disruptions. Perishable product supply chains are complex because product value deteriorates over time and transportation routes face potential disruptions.

According to a World Economic Forum report, supply chain disruptions cause a 7 percent reduction in the share prices of affected companies. Furthermore, 80 percent of global companies consider supply chain protection as their top priority.

Himanshu Shrivastava, Pankaj Dutta, Mohan Krishnamoorthy, and Pravin Suryawanshi developed a stochastic mathematical model for perishable products under route disruptions and demand uncertainty. The model optimizes facility location and distribution strategies using a two-stage mixed-integer programming approach.

Decision-making under uncertainty follows different operational policies:

• a risk-neutral policy minimizes the expected total cost of the supply chain across scenarios;
• a risk-averse policy utilizes the conditional value-at-risk approach to minimize the worst-case total cost.

Simulation via @Risk executed 1000 iterations to evaluate these policies. The resilient supply chain yielded a total expected cost of INR 132,591.2, compared to INR 116,888.82 for the disruption-free supply chain.

12. Online Shopping Attitudes, Airport PPPs, and Asphalt Adhesion

Income and experience shape student online shopping attitudes, airport PPPs require risk mitigation, and particle probe SFM quantifies mineral-asphalt adhesion.

Online Shopping Attitudes in Higher Education

Academic advancement does not make college students significantly more positive about internet shopping. A survey of Hong Kong higher education students yielded 160 fully completed questionnaires out of 240 distributed, showing a Cronbach's alpha value for attitude of 0.795.

Students earning between $8,000 and $11,999 showed a higher mean attitude score of 3.8125 compared to a mean score of 3.3540 for the less than $4,000 income group. Prior experience also significantly fosters a positive attitude and reduces perceived risks. Comparative t-tests between a group of 80 less experienced participants and 80 more experienced participants demonstrated a statistically significant difference in attitude with p = 0.000.

Risk Mitigation for Airport Infrastructure

Public-Private Partnerships in Indonesian airport infrastructure face major financial deficits and high risks requiring structured mitigation. National budgets are insufficient to cover massive archipelago infrastructure demands, with an estimated investment need of IDR1.450 quintillion between 2010 and 2014.

Data was collected from 16 public agencies and 16 ongoing private airport projects across Indonesia, such as Kertajati International Airport and Komodo Airport. Survey results on 24 risk variables indicate that respondents lean toward specific operational responses:

• a 52% overall risk retention preference;
• a very low 7% risk transfer response.

Adhesion Between Minerals and Asphalt Binders

Particle probe scanning force microscopy directly quantifies adhesion between microscale primary aggregate minerals and asphalt binders. Microspheres made of calcium carbonate, silica, and alumina with an approximate diameter of 6 micrometers were attached to tipless cantilevers to test against binders with a 30 nanometers root mean square surface roughness of the substrate.

Granite composition comprises 72.04 wt% silica and 14.42 wt% alumina. Alumina-binder pairs exhibit significantly higher adhesion than silica-binder and calcium carbonate-binder pairs due to higher polarity, backed by 180 total measurements collected per mineral-binder pair data replica.

Conclusion

Engineering technologies and computer science methods require a unified approach to solve complex industrial problems. You must integrate specialized hardware design, advanced mathematical modeling, and robust software algorithms to achieve system reliability and efficiency. The diverse research presented in this volume demonstrates that modern engineering relies on cross-disciplinary collaboration between mechanical systems, computational intelligence, and data analytics.

You can apply these rigorous methodologies to your own projects by following a systematic implementation process.

1. Assess your system requirements by identifying mechanical safety standards, computational constraints, and environmental disturbances before designing control architectures or simulation models.
2. Implement data-driven algorithms and machine learning techniques to process sensor inputs, optimize resource scheduling, and automate quality inspection processes across your industrial applications.
3. Secure your digital and physical infrastructure by adopting cryptographic frameworks, privacy-preserving protocols, and risk mitigation strategies to protect sensitive data in distributed environments.
4. Validate your theoretical models through rigorous experimentation, iterative prototyping, and real-world testing to ensure that your engineering solutions perform reliably under operational uncertainty.

By bridging the gap between theoretical innovations and practical implementations, you position yourself to address the evolving demands of modern engineering technologies. You must continuously adapt your technical skills and design philosophies to leverage emerging computational tools and cross-disciplinary methodologies effectively.

10 Key Ideas

1. Robust control approaches based on parametric synthesis and interval parameters can handle affine uncertainty and maximize the robust degree of stability in descent underwater vehicle location stabilization at a depth of 6000 m.
2. Integrating safety light curtains as a complementary protective measure in front of automatic building units prevents unexpected machine start-ups when operators or maintenance staff access hazardous zones.
3. Scalable-Delay-Insensitive optimization based on distributed lock relations significantly improves the area cost of asynchronous control circuit implementations compared to Quasi-Delay-Insensitive models.
4. Secure multiparty computation divides data across multiple independent servers using addition and multiplication forms so that individual servers cannot see the complete data or parameters, while a client integrates the final outputs.
5. Detected railway obstacles can be processed in real time using a computer vision system that runs with a 640 x 360 frame size, 16:9 aspect ratio, and 25 fps frame rate during a 2 min recording test duration.
6. A symmetric Kullback-Leibler divergence measure allows change-point detection algorithms to outperform classical methods like Otsu's and Kittler-Illingworth by yielding higher Structural Similarity Index scores.
7. Equalizing words that exceed a similarity threshold calculated by a two-layer neural network artificially increases term frequencies and prevents zero-frequency problems in structural equation models.
8. Extracting noun keyphrase candidates using top consecutive nouns cohesion and judging their difficulty level via the Japanese-Language Proficiency Test vocabulary allows automatic discovery of valuable keyphrases in Japanese documents.
9. Tactile precues outperform visual precues in reducing operator reaction times when responding to visual stimuli across various spatial mappings in human-machine interface experiments.
10. Microspheres made of calcium carbonate, silica, and alumina with an approximate diameter of 6 micrometers can be attached to tipless cantilevers to directly quantify adhesion between microscale primary aggregate minerals and asphalt binders using particle probe scanning force microscopy.