This paper investigates how nonlinear hydraulic elasticity and LuGre-type friction affect the position tracking of a valve-controlled electro-hydraulic cylinder. An integrated electro-hydraulic position-servo model is established, including the power amplifier, servo valve, hydraulic actuator, displacement sensor, and feedback loop. Valve dead zone and hysteresis are treated as input-side nonlinearities, whereas the analysis focuses on cylinder-side oil compressibility and friction. The effects of bulk modulus, the nominal air-content parameter, Coulomb friction, viscous damping, and bristle stiffness are examined through MATLAB/Simulink-AMESIM co-simulation. A discrete MPC controller is then implemented in the MATLAB/Simulink MPC Toolbox using a transfer-function prediction model derived from the reduced-order electro-hydraulic position model. The internal prediction model is linear; hydraulic elasticity and LuGre friction remain in the nonlinear plant model and are not implemented as separate real-time feedforward compensation channels. The controller uses a sampling time of 0.02 s, a prediction horizon of 20 steps, a control horizon of 5 steps, and an output-related tuning weight of 0.06. Hardware experiments show that, compared with the tested proportional-integral-derivative (PID) controller, MPC reduces step-response overshoot from 0.22 mm to 0.04 mm for a 10 mm command and from 0.25 mm to 0.04 mm for a 20 mm command; the corresponding steady-state errors decrease from 0.35 mm to 0.20 mm and from 0.25 mm to 0.10 mm. For sinusoidal tracking at 3.14 rad/s and 6.28 rad/s, MPC reduces amplitude attenuation from 0.30 mm to 0.10 mm and from 0.55 mm to 0.15 mm, and phase lag from 18.0 degrees to 6.5 degrees and from 22.3 degrees to 10.1 degrees, respectively. The results demonstrate improved transient positioning and periodic tracking relative to PID under the investigated nonlinear plant conditions.
Modern intelligent transportation and heavy-duty transport technologies increasingly operate under complex road conditions and high-load environments, making vehicle lateral vibration a critical concern for driving safety and ride comfort. Traditional passive suspension systems have inherent limitations in effectively addressing complex and time-varying disturbances. To overcome these challenges, this paper introduces centrifugal harmonic force generator technology and proposes a lateral active vibration reduction control scheme for information-constrained conditions. In practical engineering environments, lateral disturbance forces are difficult to measure directly, while often only the lateral acceleration of the vehicle body is available. A single-degree-of-freedom (SDOF) lateral vibration dynamics model and an equivalent lateral control force model are established. A vehicle motion state reconstruction method relying on a single acceleration sensor is then designed using numerical integration and digital low-pass filtering. Building upon this, a PD feedback closed-loop control framework is constructed, and the Particle Swarm Optimization (PSO) algorithm is introduced to optimize the control gains under the physical constraints of the generator. Numerical simulation results demonstrate that the PSO-optimized active control strategy reduces the lateral vibration index by 96.67
Abstract Sustainable materials and effective strengthening methods can contribute to the significant improvement of the structural performance and durability of reinforced concrete (RC) beams. 40% cement-fly ash replacement offers better sustainability and workability, whereas fibre reinforcement and external fibre reinforced polymer (FRP) wrapping are able to offer better strength and ductility. This paper is an experimental study where thirteen RC beams were studied, one of them being a conventional control beam, basalt fibre reinforced beam (BFRB) fibre contents of 0.5%, 1%, 1.5%, 2% and beams strengthened by applying E-glass fibre reinforced polymer (EGFRP) wraps, and glass fibre reinforced polymer (GFRP) wraps at 50, 75, 100, and 150 mm. In all mixes of concrete, fly ash was used as a partial replacement of cement (12.5%). The beams were tested to assess the load–deflection response, stiffness, ductility, cracking and failure modes. The mid-span deflections were determined analytically and the deflection was determined by the double integration method, which was compared to the experimental results. Basalt fibre content was found to result in much higher load-carrying capacity, stiffness, ductility and crack control than in the control beam, and a 1.5% of basalt fibre gave the best balance between strength and deformation capacity. A higher fibre dosage of 2.0% resulted in reduced performance due to fibre agglomeration. External FRP wrapping further improved beam performance, with closer wrapping spacings (50 and 75 mm) offering superior confinement and crack-bridging effects. GFRP wrapping consistently outperformed EGFRP, and the GFRP-wrapped beam with 1.5% basalt fibre and 100 mm spacing achieved the highest load capacity and deflection. As a whole, the obtained results showed that the concrete beams can be reinforced with the observed load conditions by applying the combination of fly ash, basalt fibres and external FRP wrapping.
Abstract Quantitative detection of SF₆ decomposition gases in gas-insulated switchgear is affected by nonlinear multicomponent spectral responses and environmental variations. A fiber-coupled whispering-gallery-mode (WGM) gas-sensing system integrated with a multi-scale residual one-dimensional convolutional neural network (MSR-1D-CNN) was developed for simultaneous concentration inversion of SO₂, H₂S, and CF₄. Parallel convolution branches with kernel sizes of 3, 5, and 7 provide complementary spectral receptive fields, while depthwise-separable convolution controls computational complexity and residual connections preserve lower-level spectral information. Temperature, relative humidity, and pressure are further fused with the spectral representation to compensate for environmental variations. The concentration-grid dataset was partitioned according to independent gas-preparation batches, and additional environmental conditions were independently prepared for calibration and held-out generalization evaluation. The proposed model achieved an RMSE of 1.82 ± 0.05 ppm, an MAE of 1.52 ± 0.04 ppm, and an R 2 of 0.995 ± 0.001 on the internal test set. The independent-date test yielded an average RMSE of 2.14 ppm. Under the held-out conditions of 55 °C, 70% relative humidity, and 0.13 MPa, RMSE values of 2.26, 2.14, and 2.31 ppm were obtained, respectively. The deployed FP32 model contained 0.186 M parameters and required 37.8 ms per model inference on a Cortex-M7 processor. These results demonstrate the feasibility of WGM-based multicomponent gas concentration inversion and embedded inference under the investigated laboratory conditions.
Abstract While Large Language Models (LLMs) exhibit remarkable zero-shot learning capabilities, their deployment in real-time interactive systems is constrained by sparse terminal feedback in long-sequence generation, unstable policy updates, and the computational overhead associated with verbose outputs. To address these challenges, this paper formalizes autoregressive text generation as a discrete-time Markov decision process (MDP) and proposes a latency-aware composite reward optimization framework based on proximal policy optimization (PPO). The core contribution of the proposed method is a multidimensional sequence-level terminal reward that jointly evaluates semantic coherence, lexical diversity, and computational cost. Specifically, a Natural Language Inference (NLI) cross-encoder evaluates the semantic consistency between the input context and the complete generated response, while Shannon entropy and Distinct-N metrics encourage lexical diversity. In addition, a hardware-aware penalty based on generated sequence length and measured inference latency discourages unnecessarily verbose outputs and promotes computationally efficient generation. During policy optimization, the terminal composite reward is propagated through generalized advantage estimation, while a token-wise Kullback–Leibler divergence term regularizes the updated policy against the reference policy. This formulation does not introduce token-level semantic rewards; instead, it combines informative sequence-level evaluation with step-wise policy regularization to improve empirical training stability. We evaluated the proposed framework on a large-scale hybrid corpus containing approximately 320,000 dialogue turns, including real-world English interaction data. Compared with the standard PPO baseline, the proposed method exhibited faster empirical convergence and lower variability in the observed training trajectories. It achieved a BLEU-4 score of 22.87 and a Distinct-2 score of 21.59%, while reducing the average inference latency to 42.8 milliseconds per token under the reported hardware configuration. These results demonstrate the potential of latency-aware composite reward optimization to balance response quality, lexical diversity, and inference efficiency in real-time domain-specific LLM applications.
