Army Institute of Technology

Massive Multiple Input Multiple Output (MIMO) communication system is created for the fifth-generation (5G) network to improve spectral efficiency and transmission reliability. However, accurate channel estimation remains a critical challenge in massive MIMO systems, particularly due to issues such as pilot contamination and the high-dimensional beamforming complexity associated with large antenna arrays. Despite the recent development of effective estimating approaches, estimation accuracy still has to be increased. In order to estimate channels, this study presents an optimized LSTM system. The Least Square (LS) channel estimation approach is first used to gather historical data on the channel state information (CSI) derived from pilot sequences. These gathered channel answers are used to train the proposed LSTM, in which the squirrel search algorithm (SSA) is used for optimal weight initialization in the most effective manner. By optimizing the initial weights, SSA enhances convergence behavior and avoids poor local minima during training. The trained SSA-LSTM model is subsequently used to predict the current channel response. The recommended transmit-predict strategy's functionality is assessed by changing the pilot sequence. Bit Error Rate (BER) and Mean Square Error (MSE) as functions of Signal-to-Noise Ratio (SNR) were used to assess the performance of SSA-LSTM model. In order to compare various channel estimate methods, the studies altered the SNR (usually between 5 and 50 dB), QAM modulation order (128, 256, 512, and 1024 QAM), and number of transmit and receive antennas (Nt = Nr = 100, 200, 300, and 400) under a Rayleigh fading channel model. The suggested SSA-LSTM model obtained a lower BER, which is a 27

Friction Stir Welding (FSW) of aluminium alloy dissimilar welding is essential to lightweight constructions in the automotive, aerospace and energy sector, yet stabilizing the quality of the joints in this context continues to be a challenge because of extreme thermal gradients, Intermetallic compound (IMC) formation and the dynamics of the process. Traditional offline or rule-based optimization methods are not flexible to material flow and heat production changes in real-time, which usually leads to defects and poor mechanical behavior. The paper was dedicated to the optimization of the main FSW process parameters, such as the tool shoulder-to-pin diameter ratio (SP), tool rotational speed (TRS), and tool traverse speed (TTS) with the help of a reinforcement learning (RL)-based real-time optimization algorithm to maximize the weld integrity and performance. The suggested methodology incorporated the in-situ sensor feedback, such as torque, axial force, interface temperature, and tool vibration in a digital process environment to allow the continuous state monitoring during welding. A model-free RL model based on Deep Q-Network (DQN) and Proximal Policy Optimization (PPO) to dynamically adjust SP, RS and TF in order to regulate heat input and material flow. An objective rewarding multi-objective was developed to reduce IMC thickness, thermal variability, and maximise tensile strength and joint effectiveness. The dissimilar aluminium alloy butt joints were experimentally validated to converge rapidly on policy and to be able to stay stable amidst process disturbances. The RL-optimized process demonstrated a 19.7% increase in tensile strength (142 f 5.8 MPa to 170 f 4.2 MPa), a 23.3% reduction in IMC thickness (8.6 f 0.7 & micro;m to 6.6 f 0.5 & micro;m), and a 17.2% improvement in electrical conductivity compared to the optimized static baseline. The study demonstrates that reinforcement learning enables intelligent and self-optimizing FSW through realtime adjustment of SP, TRS, and TTS, thereby offering a scalable route toward autonomous manufacturing of high-performance Cu-Al dissimilar joints.

Deep learning has advanced at a staggering pace, facilitating the production of near-original deepfakes—video and audio files that impersonate real people convincingly. These advancements can be beneficial in entertainment, accessibility, and content creation, however, the misuse of such technologies can cause significant harm relating to misinformation, digital trickery, Impersonation in cybercrimes, fraud and identity theft. This paper describes a dual-modality deepfake detection system that targets both video and audio forgeries through custom deep learning techniques. For the video detection, we suggest a hybrid CNN-LSTM model which retrieves spatial information of cropped faces and also tracks temporal changes across sequences. For audio deepfake detection, a lightweight LSTM architecture with Mel Frequency Cepstral Coefficients (MFCCs) is suggested. Both models are embedded within a web-based detection platform with independent upload and processing pipelines for audio and video content implemented through a Flask-based backend to enable practical application in the real world, allowing users to independently upload and examine media content for credible authenticity. For video deepfake detection, the results are evaluated using a composite dataset comprising FaceForensics++, Celeb-DF (v2), and DFDC facial frame samples, and attained an accuracy of $94\%$ whereas for audio detection, the SceneFake dataset with different noise and enhancement levels is used with detection accuracy of $98 \%$. The suggested methodology works in real-time and can be deployed in non-controlled scenarios using standard CPU-based systems during inference, while model training may benefit from GPU acceleration. The experimental results show that the framework remains reliable across different types of manipulations, making it an effective instrument for multimedia forensics and digital content authentication.

In contrast to human intelligence, which is based on innate knowledge, artificial intelligence (AI) and big data refer to the mental capacity displayed by robots. AI has had a profound impact on human labour and other stakeholders in the banking and finance industry. The idea of "industry 4.0" has fundamentally changed how businesses function in the modern world. Initial, controlled, defined, established, and digitally orientated are the five maturity levels in the model. There is a growing need to protect, improve, and serve the interests of customers and financial institutions. The financial institution now prioritises technology. The many technology paradigms used in the banking and financial industries are examined in this article. In order to better comprehend and pinpoint possible new areas, the study investigates the function of digital technology in the banking and finance industries. The study investigates the possible technological obstacles that the banking industry may face in the future.

The paper presents a scalable, low-cost Automatic Print Machine Extender with the goal of improving traditional printing shops. The system combines a Raspberry Pi module, cloud-based database and two mobile applications, one for users and the other for shop owners. Users can now upload files remotely, select print options, receive location of print centres using Google Maps API, and securely pay for items. The shop will only start printing after the user made a payment and the file will be automatically removed from the system after printing, thus protecting the user from potential misuse. The shop owners will be used to manage queue and assign jobs based on solved printers’ availability. Compared to antiquated vending systems this solution offers increased flexibility; dynamic pricing; increased security; and is best when used in a universal public and institutional environment. This approach significantly advances digital printing workflows by demonstrating a working simulation validated through Raspberry Pi-based testing.