Optimal Input Research Articles

Towards 250-m gigabits-per-second underwater wireless optical communication using a low-complexity ANN equalizer

The breakthroughs in communication distance and data rate have been eagerly anticipated by scientists in the area of underwater wireless optical communication (UWOC), which is seriously limited by the obvious aquatic attenuation in underwater channels. The high-power laser source and ultra-sensitive photodetector are straightforward in extending the UWOC distance. However, nonlinear impairments caused by bandwidth-limited high-power transmitters and sensitive receivers severely degrade the data rate of long-distance UWOC. In this paper, we develop a UWOC system using a high-power transmitter by beam combining of 8-channel cascaded laser diodes (LD) and a sensitive receiver by a silicon photomultiplier (SiPM). The combined linear equalizer and low-complexity artificial neural network (ANN) equalizer are used to robustly achieve 1-Gbps data transmission over a 250-m UWOC system. To the best of our knowledge, this is the first Gbps-level UWOC experimental demonstration in >250-meter underwater transmission that has ever been reported. To lower the complexity of the ANN equalizer, a linear equalizer is applied first in order to prune the input size of the ANN equalizer. The optimal input size of the ANN equalizer is identified as 9. The ANN architecture consists of two hidden layers, with 10 neurons in the first layer and a single neuron in the second layer. The performance of the proposed ANN-based system is compared with that of systems employing Volterra and linear equalizers. The bit error rate (BER) at a data rate of 1 Gbps over a 250-m UWOC is reduced to 3.4 × 10−3 with the combined linear and ANN equalizer, which is below the hard-decision forward error correction (HD-FEC) limit. In contrast, the linear and Volterra equalizer-based systems achieve data rates of 500 Mbps and 750 Mbps, respectively.

A Study on the Differences in Optimized Inputs of Various Data-Driven Methods for Battery Capacity Prediction

As lithium-ion batteries become increasingly popular worldwide, accurately determining their capacity is crucial for various devices that rely on them. Numerous data-driven methods have been applied to evaluate battery-related parameters. In the application of these methods, input features play a critical role. Most researchers often use the same input features to compare the performance of various neural network models. However, because most models are regarded as black-box models, different methods may show different dependencies on specific features given the inherent differences in their internal structures. And the corresponding optimal inputs of different neural network models should be different. Therefore, comparing the differences in optimized input features for different neural networks is essential. This paper extracts 11 types of lithium battery-related health features, and experiments are conducted on two traditional machine learning networks and three advanced deep learning networks in three aspects of input differences. The experiment aims to systematically evaluate how changes in health feature types, dimensions, and data volume affect the performance of different methods and find the optimal input for each method. The results demonstrate that each network has its own optimal input in the aspects of health feature types, dimensions, and data volume. Moreover, under the premise of obtaining more accurate prediction accuracy, different networks have different requirements for input data. Therefore, in the process of using different types of neural networks for battery capacity prediction, it is very important to determine the type, dimension, and number of input health features according to the structure, category, and actual application requirements of the network. Different inputs will lead to larger differences in results. The optimization degree of mean absolute error (MAE) can be improved by 10–50%, and other indicators can also be optimized to varying degrees. Therefore, it is very important to optimize the network in a targeted manner.

Gaussian Process-Based Robust Optimization with Symmetry Modeling and Variable Selection

Gaussian process (GP)-based robust optimization is an effective tool in product quality improvement. However, most existing variable selection methods are designed for parametric models and are unsuitable for nonparametric GP models. Additionally, improving the prediction accuracy of GP models with limited design points remains a significant challenge in robust optimization. To address these issues, this article proposes a GP-based multi-stage robust parameter optimization method that integrates symmetry modeling, sensitivity analysis (SA), and Markov Chain Monte Carlo (MCMC) techniques. First, a modified expected improvement (EI) criterion is introduced to enhance the utilization efficiency of design points. Second, a nonparametric variable selection technique based on SA is developed for GP models to identify significant variables. This method considers both independent variables and their interactions, improving the interpretability of GP models. Finally, the selected variables are used to construct the robust optimization model, and the genetic algorithm (GA) is employed to search for the optimal solution within the feasible domain. Numerical simulations and real-world experiments demonstrate the effectiveness of the proposed method. Comparative results indicate that the proposed method obtains more robust optimal input parameter settings compared to existing approaches.

Influence of treatment and fly ash fillers on the mechanical and tribological properties of banana fiber epoxy composites: experimental and ANN-RSM modeling

ABSTRACT This research examines the impact of chemical treatment and banana fly ash fillers on the mechanical, tribological, and water absorption characteristics of banana fiber-reinforced epoxy composites. Alkaline treatment enhanced fiber-matrix adhesion, markedly improving mechanical characteristics. The optimal performance occurred at 10% fly ash content, yielding a tensile strength of 40.25 MPa, a flexural strength of 77.23 MPa, and an impact strength of 44.82 kJ/m2. Water absorption studies indicated a decline in moisture uptake, reducing from 40% in untreated composites to 25% in composites containing 15% fly ash, due to enhanced bonding and fewer voids. Tribological experiments demonstrated a decrease in Specific Wear Rate (SWR) and Coefficient of Friction (COF) with elevated fly ash concentration, signifying improved wear resistance. Predictive modeling with Artificial Neural Networks (ANN) showed enhanced accuracy (mean error: 0.9584% for SWR, 0.50265% for COF). RSM optimization identified the optimal input parameters for minimizing SWR and COF: a sliding velocity of 5.14491 m/s, a sliding distance of 652.05 m, and a fly ash content of 12.6236%, yielding minimum SWR and COF values of 15.63 × 10− 5 mm3/Nm and 0.242241, respectively. SEM analysis confirmed that treated fibers and fly ash fillers minimized wear and crack propagation while improving fracture toughness. The results underscore the promise of banana fly ash-filled composites for automotive, aerospace, and structural applications that necessitate improved mechanical, tribological, and moisture-resistant characteristics.

Dynamic event‐triggered‐based adaptive frequency control of microgrids under cyber‐attacks via adaptive dynamic programming

AbstractThe increasing penetration of renewable energy sources (RES) and the development of the cyber‐physical microgrids (CPMs) make greater demands for frequency control of microgrids. The common approach for frequency control is controlling the micro‐turbine to compensate for frequency deviations, with energy storage systems serving as an auxiliary approach. This article proposes an online adaptive frequency control method to control the governor and energy storage to realize the frequency recovery of microgrid, subject to the external unknown disturbances caused by wind turbine, power load and false data injection (FDI) attacks. First, the non‐zero sum (NZS) games of the considered system are modeled in this work, where the unknown disturbances are also taken into account. For the sake of estimating the unknown disturbances, a disturbance observer (DOB) for the microgrid system is introduced. Then, on the basis of the estimated results of the applied DOB, the disturbance compensation input is derived to offset the interference of the unknown disturbance. Meanwhile, the adaptive dynamic programming (ADP) approach is employed to derive the adaptive optimal control input for the NZS games of microgrid system. Besides, the dynamic event‐triggered (DET) control is introduced, reducing the occupation of computing resources. By utilizing the Lyapunov's method, the stability of the closed‐loop system, the convergence of the estimation weight, the estimation disturbances and the system state are guaranteed. The effectiveness of the proposed method is ultimately verified by the simulation results.

Dynamic Modeling and Robust Inverse Dynamics Control Enhanced by Optimal Equivalent Input Disturbance for a Shipborne Wave Compensation Hexapod Platform

Dynamic Modeling and Robust Inverse Dynamics Control Enhanced by Optimal Equivalent Input Disturbance for a Shipborne Wave Compensation Hexapod Platform

Development of a novel modeling framework based on weighted kernel extreme learning machine and ridge regression for streamflow forecasting

A precise streamflow forecast is crucial in hydrology for flood alerts, water quantity and quality management, and disaster preparedness. Machine learning (ML) techniques are commonly employed for hydrological prediction; however, they still face certain drawbacks, such as the need to optimize the appropriate predictors, the ability of the models to generalize across different time horizons, and the analysis of high-dimensional time series. This research aims to address these specific drawbacks by developing a novel framework for streamflow forecasting. Specifically, a hybrid ML model, WKELM-R, is developed to predict streamflow based on daily discharge and precipitation. The model combines ridge regression (RR), locally weighted linear regression (LWLR), and kernel extreme learning machine (KELM) to enhance multi-step-ahead predictions by accounting for both linear and nonlinear characteristics. In data preprocessing, this study applies multivariate variational mode decomposition (MVMD) for decomposition to handle non-stationarity and complexity, Boruta-XGBoost for feature selection to select the optimal inputs and decrease the dimension, and gradient-based optimizer (GBO) for adjustment of model parameters to overcome the need to optimize the appropriate predictors. To demonstrate the ability to handle real-world conditions and different time horizons, WKELM-R was applied to a watershed in North Dakota, USA to forecast discharge for three different time horizons. The results were compared with those from the existing standalone and hybrid models by multi-criteria decision-making (MCDM), demonstrating the efficacy and unique capabilities of the new hybrid model in streamflow forecasting (for the testing level at t + 3: R = 0.992, RMSE = 0.426, NSE = 0.983; at t + 7: R = 0.997, RMSE = 0.249, NSE = 0.994; at t + 14: R = 0.996, RMSE = 0.304, NSE = 0.991).

Arctic Sea Ice Prediction Based on Multi‐Scale Graph Modeling With Conservation Laws

AbstractArctic sea ice prediction is critical for exploring climate change, resource extraction, and shipping route planning. This paper introduces a novel neural network model, Ice Graph Attention neTwork (IceGAT), that is trained to predict sea ice concentration (SIC) from a number of atmospheric, oceanic, and land surface measurements. It is based on two design principles: (a) the complex spatial interactions in weather dynamics are captured via a series of graphs corresponding to different spatial resolutions and (b) the incorporation of the physical conservation laws for moisture and potential vorticity. We devise two main variants with 1 hr and 24 hr temporal resolution and determine the optimal input horizon to be 5 days. IceGAT features leading accuracy (96.7%; +2.4% over the current state‐of‐the‐art) and low inference time (1/4 s, on a single GPU). An online implementation (based on data from ERA5) alongside supplementary videos and our shared code are accessible at: https://lannwei.github.io/IceGAT/.

Strategy Deduction for Improving Aviation Emergency Rescue Capability from the Perspective of Public Safety

Aviation emergency rescue plays a crucial role in the emergency management of various emergency events. Consequently, the processes of identifying relevant constraints of aviation emergency rescue ability, analyzing the interactive relationship among factors, and putting forward an economical and effective improvement strategy for aviation emergency rescue capabilities have become widely popular among scholars. To greatly improve the economic input efficiency of aviation emergency rescue capabilities, 18 relevant influencing factors were screened and identified from four perspectives: system institutionalization, technological advancement, rescuer professionalism, and operation safety. Then, the causal relationship among the influencing factors and the feedback mechanism were analyzed via the system dynamics (SD) method, and the influencing factor weights were calculated via the entropy weight method. Then, the equations of state variables and constants among the influencing factors were determined, and an SD model of aviation emergency rescue was built. Finally, a case study based on a practical construction condition of an aviation emergency rescue in Henan Province, China was analyzed. The variation trend of rescue ability was predicted from the temporal dimension, and the optimal economic input strategy for effectively improving the rescue ability was determined. Results demonstrate that the efficacy level of aviation emergency rescue ability in Henan Province is 44.97. The aviation rescue ability level on the 49th month can be improved to 90 if the monthly input into the aviation emergency rescue is raised to 3 million RMB. Additionally, the aviation rescue project can be significantly improved if the input proportions of operation safety, rescuer professionalism, system institutionalization, and technological advancement are 0.4, 0.2, 0.2, and 0.2, respectively. This work can provide an economic decision reference for building and perfecting public safety systems.

Enhancing secrecy rates in visible-light communications with signal-dependent noise: exploring optical jamming technique

Abstract This research investigates the security of a visible light communication (VLC) system in an indoor setting. The system consists of multiple transmitters (referred to as Alice), a legitimate receiver (referred to as Bob), and a passive adversary (referred to as Eve). A scheme is proposed to enhance secrecy by using optical jamming signals alongside the confidential signal, under constraints of nonnegativity and average optical intensity. The optical jamming signals follow a truncated Gaussian distribution to adhere to predefined constraints. The analysis takes into account both noise independent of the signal and noise dependent on the signal. The study obtains lower and upper bounds on secrecy rate through the application of the variational method, the dual expression of secrecy rate, and the notion of ‘the optimal input distribution that tends towards infinity’. An asymptotic analysis reveals a minimal performance difference between the upper and lower bounds under large optical intensity. Introducing a peak optical intensity constraint enables further scrutiny of both the exact and asymptotic secrecy-rate bounds. Numerical results are provided to validate the obtained secrecy-rate bounds.

Dominant Tree Species Mapping Using Machine Learning Based on Multi-Temporal and Multi-Source Data

Timely and accurate tree species mapping is crucial for forest resource inventory, supporting management, conservation biology, and ecological restoration. This study utilized Sentinel-1 and Sentinel-2 data to classify five dominant tree species in Chengde and Beijing. To effectively capture the influence of multi-temporal data, data were acquired in March, June, September, and December 2020, extracting various features, including bands, spectral indices, texture features, and topographic variables. The optimal input variable combination was explored using 1519 field survey samples for training and testing datasets. Classification employed Random Forest, XGBoost, and deep learning models, with performance evaluated through out-of-bag estimation and cross-validation. The XGBoost model achieved the highest accuracy of 81.25% (kappa = 0.74) when using Sentinel-1 and Sentinel-2 bands, indices, texture features, and DEM data. Results demonstrate the effectiveness of using Sentinel data for tree species classification and emphasize the value of machine learning algorithms. This study underscores the potential of combining synthetic aperture radar (SAR) and optical data for large-scale tree species classification, with significant implications for forest monitoring and management.

Optimal input excitations for suppressing nonlinear instabilities in multimode fibers

Wavefront shaping has become a powerful tool for manipulating light propagation in various complex media undergoing linear scattering. Controlling nonlinear optical interactions with spatial degrees of freedom is a relatively recent but fast growing area of research. A wavefront-shaping-based approach can be used to suppress nonlinear stimulated Brillouin scattering (SBS) and transverse mode instability (TMI), which are the two main limitations to power scaling in high-power narrowband fiber amplifiers. Here we formulate both SBS and TMI suppression as optimization problems with respect to coherent multimode input excitation in a given multimode fiber. We develop an efficient method using linear programming for finding the globally optimal input excitation for minimizing SBS and TMI individually or jointly. The theory shows that optimally exciting a standard multimode fiber leads to roughly an order of magnitude enhancement in instability-free output power compared to fundamental-mode-only excitation. We find that the optimal mode content is robust to small perturbations and our approach works even in the presence of mode-dependent loss and gain. When such optimal mode content is excited in real experiments using spatial light modulators, the stable range of ultrahigh-power fiber lasers can be substantially increased, enabling applications in gravitation wave detection, advanced manufacturing, and defense.

Optimization of high speed bone drilling with automated osteonecrosis avoidance using an ANN-GA hybrid method

ABSTRACT Efficient high-speed bone drilling with an automated osteonecrosis avoidance strategy is a very fascinating task to the orthopaedic surgeons. Several drilling parameters such as drill diameter, speed, and feed rate can affect the performance of drilling in terms of generated temperature and thrust force. This research is conducted using a high-speed mini CNC machine to investigate bone drilling parameters’ impact on the outputs. Here, a piece of swine bone has been used for testing material and an automated cooling system based on proportional controller (P) was developed to prevent osteonecrosis. Further, a hybrid Artificial Neural Network-Genetic Algorithm (ANN-GA) model has been proposed based on the different experimental findings. ANN predicts output variables, and GA determines the optimal input drilling parameters by minimizing the ANN model of temperature and force. The simulation results revealed that drill diameter of 1.838 mm, speed of 17,880 rpm, and feed rate of 9.973 mm/min minimized the output temperature generated on swine bone. Similarly, diameter of 1.5 mm, speed of 16,582 rpm, and 10 mm/min feed rate helped to minimize the generated force during drilling. The proposed model has been validated by conducting new experiments based on these obtained optimal parameters.

Optimization of Inputs for the Application of ANN to Rail Track Granular Materials

Machine Learning (ML) models such as Artificial Neural Networks (ANN) have gained increasing popularity in geotechnical engineering applications as an alternative to conventional empirical and computational models. At present, very few ML models exist for predicting the mechanical responses of track granular materials such as ballast and subballast which may even comprise of composite mixtures of blended granular materials. Moreover, the performance of any ML model depends not only on the quality and quantity of available data but also on the selection process for input parameters, which often lacks adequate justification in the past literature. In this context, the current study introduces ANN models for track granular materials based on published laboratory data with special emphasis on the selection of an optimal set of input parameters. Two applications of ANN are considered to: (i) predict the peak friction angle (ϕpeak') of a variety of granular mixtures under static loading and (ii) predict ballast breakage under cyclic loading. The selection process involves prudent analysis of key influential parameters in a geotechnical perspective, while also ensuring that they are conveniently measurable. Performance evaluation of these models with various input combinations is carried out, while proposing optimal input parameters for both applications.

Seeing Beyond Noise: Improving Cryptocurrency Forecasting with Linear Bias Correction

Cryptocurrency is recognized as a leading digital currency by its peer-to-peer transfer capabilities and secure features. Accurately forecasting cryptocurrency price trends holds substantial significance for investors and traders, as they inform critical decisions regarding the acquisition, divestment, or retention of cryptocurrencies, guided by expectations of value, risk assessment, and potential returns. This study also aims to identify a resourceful technique to efficiently forecast prices of cryptocurrencies such as Bitcoin (BTC), Binance (BNB), Ripple (XRP), and Tether (USDT) using optimal data-driven models (LSTM, GRU, and BiLSTM models) using bias correction. The proposed methodology includes collecting cryptocurrency data and precious metal data from Coindesk and BullionVault, respectively, and then finding the optimal model input combination for each cryptocurrency by lag adjustment and correlating feature selection. Hyperparameter tuning was performed by trial-and-error technique, and an early stopping function was applied to minimize time and space complexity. Bias correction (BC) is applied to model-forecasted price trends to reduce errors in evaluation and to enhance accuracy by adjusting model outputs to reduce prediction bias, providing a refined alternative to traditional unadjusted deep learning methods. GRU-BC outperformed other models in forecasting Bitcoin (with MAE 25.291, RMSE 31.266, MAPE 2.999) and USDT (with MAE 0.0006, RMSE 0.0012, MAPE 0.0622) price trends, while BiLSTM-BC was superior in predicting XRP (with MAE 0.0129, RMSE 0.0171, MAPE 2.9013) and BNB (with MAE 2.2759, RMSE 2.8357, MAPE 1.9785) market price flow.

Determining best dressing parameters for internal grinding SKD11 steel using EAMR technique

The article conducts a research study on the use of Multi-Criteria Decision-Making (MCDM) in the internal grinding process of SKD11 tool steel. The aim is to identify the optimal input process parameters for the dressing process in order to minimize surface roughness (SR) and maximize wheel life (Lw). The EAMR strategy was employed to address the MCDM task, whereas the Entropy method was utilized to determine the weights of the criterion. The experiment also included the analysis of six input process parameters: coarse dressing depth, coarse dressing passes, fine dressing depth, fine dressing passes, non-feeding dressing, and dressing feed rate. The experiment was performed using an L16 orthogonal array and the Taguchi method. The wheel's durability and the roughness of the sample's surface were quantified and recorded for analysis in the MCDM problem. The research findings have identified the optimal dressing solutions for internal cylindrical grinding.).

Application of the Multi-Criteria Decision Method to Find the Best Input Factors for Electrical Discharge Machining 90CrSi Tool Steel using Graphite Electrodes

This paper examines the optimization of the Electrical Discharge Machining (EDM) process when machining cylindrical parts of 90CrSi tool steel using various graphite electrodes. A Multi-Criteria Decision Making (MCDM) approach, including the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS), Simple Additive Weighting (SAW), and Multi-Attributive Border Approximation Area Comparison (MABAC) was utilized to identify the optimal input factors that would achieve three machining objectives: minimizing Surface Roughness (SR) and Electrode Wear Rate (EWR) and maximizing Material Removal Rate (MRR). Criteria weights were calculated using the Method based on the Removal Effects of Criteria (MEREC). Additionally, three types of graphite electrodes (HK0, HK15, and HK20) and five process factors, such as Servo Voltage (SV), Input Current (IP), pulse on time (Ton), pulse off time (Toff), and Types of Graphite (TOG) were tested with experiments structured using a Taguchi L18 design and Minitab R19 software. The results indicate that the optimal EDM input parameters are as follows: IP = 9.5 A, SV = 5 V, Ton = 8 µs, Toff = 8 µs, with the HK20 electrode balancing SR, EWR and MRR for enhanced machining performance.

Backward induction-based multi-layer approach for watershed flood management in arid regions

Managing floods in interconnected nonurban and urban areas of arid regions prone to flash floods requires a more dynamic and integrated approach than traditional linear methods. In this regard, this study proposes a novel multi-stage decision-making framework which reshapes the traditional cascade approach to flood management by addressing the interdependencies between upstream and downstream regions. The proposed cyclic decision-making process involves five main steps: First, the Hydraulic and Hydrological (H/H) conditions in urban and nonurban areas were modeled using the SWMM. Then, a multi-stage optimization framework was constructed, in the first step of which the detention dams' location and dimensions were optimized in the nonurban area. Next, to link the first and second stages of the optimization, COPRAS method was employed for selecting the critical upstream management strategies. These strategies were determined as the input for the flood network optimization in the second stage. Finally, the Backward Induction Game Theory (BIGT) method was employed to find the best management strategy, under which not only the upstream conditions were considered in downstream flood protection but also stakeholders' conflicts were totally resolved by choosing the Nash Equilibrium. Our findings revealed that the implementation of the selected scenario could reduce flood peaks at the entrance of urban areas by 60.7 % and enhance shock absorption by 81.8 %. In fact, we guarantee that it is possible to propose a scenario encompassing drainage and detention dams' characteristics that, in addition to resolving stakeholders' conflicts of interest, can reduce the risk of flooding in complex watersheds. Moreover, the framework transforms the sequential, one-way cascade decision-making process into a continuous cycle, where upstream and downstream's H/H conditions and stakeholders' preferences are evaluated in coordination to achieve an agreement. Also, this framework helps decision-makers access the most sustainable solutions considering hydrological and socio-economic points of view.

Prediction of shut-off head for centrifugal pumps based on grey theory and GA-BP neural network

Calculating the shut-off head for centrifugal pumps poses significant challenges due to inaccuracies in existing empirical methods. This paper presents a predictive model based on extensive experimental data, employing a back propagation (BP) neural network optimized via grey theory and genetic algorithms (GAs). Data were collected from 141 single-stage volute centrifugal pumps, and grey theory was used to analyze nine critical parameters of the impeller and volute, yielding five optimal input schemes with correlation coefficients exceeding 0.6. The GA was utilized to optimize the weights and thresholds of the BP model. The training involved 121 samples, while 20 additional samples were used to evaluate the models against three established methods (throne, modified throne, and regression fitting). The results indicate that the optimal input scheme consists of four parameters (impeller diameter, blade wrap angle, inlet diameter, and rotational speed) with correlation coefficients greater than 0.7. Both the BP and GA-BP models achieved training regression coefficients approaching 0.999. Within the specific speed range of 22–215, the GA-BP model demonstrated superior performance to the BP model and the three established methods, with maximum testing errors of 10.0%, 20.6%, 20.7%, 19.9%, and 23.3%, and average relative errors of 3.9%, 5.0%, 8.4%, 8.1%, and 5.8%, respectively. This paper introduces a novel prediction model for the shut-off head with an accuracy of 96%, providing a valuable reference for hydraulic design and performance prediction in centrifugal pumps.

Optimizing carbon source addition to control surplus sludge yield via machine learning-based interpretable ensemble model.

Appropriate carbon source addition can save operational costs and reduce surplus sludge yield in the wastewater treatment plant (WWTP). However, the link between carbon source and surplus sludge yield remains neglected although machine learning (ML) has become a powerful tool for WWTP, and is a challenge due to more complex multidimensional pattern recognition. Herein, weighted average ensemble strategy was conducted to assemble multiple diverse basic models to obtain better prediction capability to optimize carbon source addition (Model-1) and further control surplus sludge yield (Model-2). The ensemble models significantly outperformed all single models with MAE of 5.82g/m3, MSE of 60.59 and R2 value of 0.98 in Model-1 and MAE of 15.09g/m3, MSE of 449.01 and R2 value of 0.93 in Model-2. The optimal input feature subset was explored to reduce model complexity, indicating that the final ensemble models can predict with high precision using relatively few features with MAE of 6.41g/m3, MSE of 78.49 and R2 value of 0.97 in Model-1 and MAE of 12.82g/m3, MSE of 232.71 and R2 value of 0.95 in Model-2. Furthermore, the final models were deployed into an offline web application to facilitate their utility in real-world settings, demonstrating 47.25 % savings in carbon source addition and 15.89 % reductions in surplus sludge yield for an extra month of running. This work offers an efficient approach for the WWTP to optimize carbon source addition and provides new insights into controlling surplus sludge yield.