Accurate Dynamic Model Research Articles

Analyzing Burner Performance and Combustion Phenomenon in an Olefin Plant's Industrial Furnace: A CFD Study.

This work presents a comprehensive study of the combustion performance of an industrial furnace in an olefin plant using computational fluid dynamics (CFD) simulations. The focus was on analyzing the heat release pattern of bottom burners to optimize the furnace efficiency in steam-cracking processes. The study developed an accurate computational fluid dynamics (CFD) model for predicting combustion behavior in a cracking furnace. The computational model was validated by comparing the simulation results with industrial data and was used to investigate the impact of burner clogging and the importance of small holes in the body of burners in the furnace. The results also provided insights into the influence of excess air, temperature distribution, fluid behavior, composition of combustion products, and thermal efficiency of the furnace. The presented results contributed to a better understanding of parameters controlling combustion performance in steam-cracking furnaces.

Analysis of longitudinal-vertical coupling vibration of four hub motors driven electric vehicle under unsteady condition

The influence of hub motor unbalanced magnetic force (UMF) on the vibration of electric vehicle under steady state conditions has been known, but under unsteady conditions, the hub motor UMF will change with the vehicle operation condition, and there would exist complex coupling vibration for the hub motors driven electric vehicle. Here, the longitudinal-vertical coupling dynamics of the four hub motors driven electric vehicle under unsteady condition is studied. Integrating the motor electromagnetic excitation and electric vehicle dynamics, a longitudinal-vertical coupling dynamics model of the four hub motors driven electric vehicle is established. Based on the variable switching frequency field-oriented control model, analytical model of the UMFs acting on the motor stator and rotor parts under unsteady condition are developed. For model validation, a four hub motors driven electric vehicle has been tested, the accuracy of the longitudinal-vertical coupling dynamics model established in this paper was verified. Then, longitudinal-vertical coupling vibration characteristics of the four hub motors driven electric vehicle under road excitation and coupling excitation are analyzed. The results show that the longitudinal and vertical movement of the four hub motors driven electric vehicle is coupled by hub motor. In addition, under unsteady condition, the motor UMFs will cause vertical vibration of the electric vehicle body and hub motor stator, the vibration shows order characteristics including low order harmonic hfc and inverter switching frequency sideband harmonic k1 fs±k2 fc ( fc = pn/60, fs is inverter switching frequency, k1 and k2 are positive integers, p is the number of pole pairs and n is motor speed.). The motor electromagnetic torque will cause longitudinal vibration of the electric vehicle body, the vibration shows order characteristics including harmonics fs±3 fc and 2 fs. The main harmonic of vehicle body pitch angle acceleration is 2 fs.

Digital Twin for the SBR Cold Emulsion Copolymerization Process

AbstractDevelopment of accurate and robust dynamic models for the description of emulsion copolymerization processes is fundamental for the implementation of monitoring, advanced control, and optimization strategies. There are several studies on the dynamic modeling of styrene/1,3‐butadiene rubber (SBR) emulsion copolymerization, but most of them focus on hot conditions or only one semi‐batch reactor, as in the case of cold conditions. For this reason, the present study focuses on the dynamic modeling of SBR cold emulsion copolymerization processes considering a train of 15 continuous stirred tank reactors, as in many real industrial sites. The developed dynamic model is implemented by using the digital twin (DT) concept, which involves the online reading of process variables and an adaptive strategy for online tuning of some of the model parameters, being also sensitive to the effect of real‐time changes on the number of reactors in the train, a subject that has been overlooked previously, but which is important at the plant site. The practical application of the DT for monitoring a real industrial process illustrates the robustness and accuracy of the developed tool, making it useful for opportune detection of process anomalies and opening the way for future advanced control strategies.

The implementation of a practice-oriented approach in teaching students of technical universities based on the development of a digital twin of the information system of an industrial enterprise

Introduction. In teaching future engineers and mechanics, difficulties arise when they study such disciplines as Theoretical Mechanics and Strength of Materials. As a rule, this is due to the need to consolidate students’ acquired theoretical knowledge. However, carrying out a sufficient number of strength tests of products is associated with significant financial costs, so it is often economically expensive to obtain results by implementing physical experiments, and in some cases it is simply impossible. In this regard, for teaching it is advisable to use accurate distributed dynamic mathematical and software models of objects, that is, digital twins. Materials and methods. As a prototype of a digital twin, an information system of an industrial enterprise was considered, designed to monitor the stress-strain state of a beam using strain gauge data. 198 students and 8 teachers from the Faculty of Mechanics and Technology and the Faculty of Automation and Computer Engineering of the Novosibirsk State Technical University (Russian Federation) took part in testing the methodology. Methods of mathematical statistics: Pearson's chi-squared test, survey and correlation analysis. Results. As a result of the introduction of teaching methods based on the use of a digital twin, situations were created that required students to critically analyze and include metacognitive skills. The implementation of practice-oriented training when working with a digital twin of an information system of an industrial enterprise has an impact on the academic performance of students (χ2 = 4.08; p < 0.05) and allows students to firmly consolidate the acquired theoretical knowledge and practical skills, which will subsequently allow the formation of sustainable professional competencies. Scientific novelty. The study revealed that learning using a digital twin contributes to a more controlled passage of the curriculum by students, develops qualities such as self-control and self-discipline, and stimulates cognitive interest in a specific academic discipline and the educational process in general. Practical relevance. The main advantage of introducing a digital twin of an industrial enterprise information system into the learning process is the effectiveness of this approach in terms of consolidating and developing practical skills among students. In addition, teaching using digital twins is particularly useful for preparation activities in high-risk industries, where practice at real-life facilities is impossible or difficult.

Development and Implementation of Physics-Informed Neural ODE for Dynamics Modeling of a Fixed-Wing Aircraft Under Icing/Fault

Accurate dynamics modeling is crucial for the safety and control of fixed-wing aircraft under perturbation (e.g. icing/fault). In this work, we propose a physics-informed Neural Ordinary Differential Equation (PI-NODE)-based scheme for aircraft dynamics modeling under icing/fault. First, icing accumulation and control surface faults are considered and injected into the nominal (clean) aircraft dynamics model. Second, the physics knowledge of aircraft dynamics modeling is divided into kinematics and kinetics. The former is universally applicable and borrows directly from the nominal aircraft. The latter kinetics knowledge, which hinges on external forces and moments, is inaccurate and challenging under icing/fault. To address this issue, we employ Neural ODE to compensate for the residual between the aircraft dynamics under icing/fault and the nominal (clean) condition, resulting in a naturally continuous-time modeling approach. In experiments, we benchmark the proposed PI-NODE against three baseline methods in a dedicated flight scenario. Comparative studies validate the higher accuracy and improve the generalization ability of the proposed PI-NODE for aircraft dynamics modeling under icing/fault.

A Multi-Objective Evolutionary Approach to Discover Explainability Tradeoffs when Using Linear Regression to Effectively Model the Dynamic Thermal Behaviour of Electrical Machines

Modelling and controlling heat transfer in rotating electrical machines is very important as it enables the design of assemblies (e.g., motors) that are efficient and durable under multiple operational scenarios. To address the challenge of deriving accurate data-driven estimators of key motor temperatures, we propose a multi-objective strategy for creating Linear Regression (LR) models that integrate optimised synthetic features. The main strength of our approach is that it provides decision makers with a clear overview of the optimal tradeoffs between data collection costs, the expected modelling errors and the overall explainability of the generated thermal models. Moreover, as parsimonious models are required for both microcontroller deployment and domain expert interpretation, our modelling strategy contains a simple but effective step-wise regularisation technique that can be applied to outline domain-relevant mappings between LR variables and thermal profiling capabilities. Results indicate that our approach can generate accurate LR-based dynamic thermal models when training on data associated with a limited set of load points within the safe operating area of the electrical machine under study.

Load model of high power magnet power supply for EAST based on GMPSO-BP neural network

To provide an accurate dynamic load model for the stability analysis of Tokamak, a method of back propagation (BP) neural network employing particle swarm optimization(PSO) algorithm based on mutation of Gaussian white noise disturbance (GMPSO-BP) is recommended. This method performs high-precision fitting using the GMPSO-BP neural network on the measured data of the experimental advanced superconducting Tokamak (EAST) poloidal field magnet power supply, extracts network parameters to build a dynamic load model, and then compares the simulation results with the measured data to calculate the error coefficient. The model is a component of the EAST distribution network's digital simulation model. An analogous load model is employed instead of developing the poloidal field magnet power supply from the mechanism. The revised algorithm has quick convergence times, good initial value adaptability, and error function accuracy of 0.3 to 3 %. The simulation outcomes show that the approach has a greater training effect.

Oscillation-free point-to-point motions of planar differentially flat under-actuated robots: a Laplace transform method

AbstractDifferentially flat under-actuated robots are characterized by more degrees of freedom (DOF) than actuators: this makes possible the design of lightweight cheap robots with high dexterity. The main issue of such robots is the control of the passive joint, which requires accurate dynamic modeling of the robot.Friction is usually discarded to simplify the models, especially in the case of low-speed trajectories. However, this simplification leads to oscillations of the end-effector about the final position, which are incompatible with fast and accurate motions.This paper focuses on planar $n$ -DOF serial robotic arms with $n-1$ actuated rotational joints plus one final passive rotational joint with stiffness and friction properties. These robots, if properly balanced, are differentially flat. When the non-actuated joint can be considered frictionless, differentially flat robots can be controlled in open loop, calculating the motor torques demanded by point-to-point motions. This paper extends the open-loop control to robots with a passive joint with viscous friction adopting a Laplace transform method. This method can be adopted by exploiting the particular structure of the equations of motion of differentially flat under-actuated robots in which the last equations are linear. Analytical expressions of the motor torques are obtained. The work is enriched by an experimental validation of a $2$ -DOF under-actuated robot and by numerical simulations of the $2$ - and $4$ -DOF robots showing the suppression of unwanted oscillations.

Dynamics modeling and simulation of steel cord conveyor belt based on dynamic elastic modulus

Steel cord conveyor belts are widely used in long-distance transportation of various materials. The steel cord belt is composed of rubber and steel cord core, exhibiting excellent viscoelasticity and complex dynamic characteristics during operation. The existing static elastic modulus cannot accurately reflect the real dynamic properties of the conveyor belt. In this paper, based on the catenary theory, the formula for the dynamic elastic modulus of the conveyor belt is derived by introducing the sag and elastic deformation of the steel cord conveyor belt on the basis of the static elastic modulus. The dynamic modeling of the steel cord conveyor belt is carried out using rigid belt block element + flexible beam method. Then, the RecurDyn software is used to establish the dynamic simulation system of the belt conveyor and conduct simulations, which are validated through a test bench. The results show that dynamics simulation based on the dynamic elastic modulus is closer to the real dynamic characteristics of the belt conveyor. Finally, the dynamic characteristics of belt conveyors under different start-up times, different start-up speeds, and idler spacing are analyzed using this modeling and simulation method. The results demonstrate that the dynamic elastic modulus can accurately reflect the real dynamic properties of the conveyor belt, providing a more accurate dynamic modeling and simulation method for the design of belt conveyors.

Iterative learning–based model-free adaptive precise heading following of an autonomous underwater vehicle with unknown disturbances

Due to the nonlinearity, strong coupling, and uncertain parameters of autonomous underwater vehicle (AUV), it is difficult to build an accurate dynamic model, which makes precise control of AUV extremely challenging. To handle the precise heading-following problem of AUV, this paper proposes an iterative learning-based redefine model-free adaptive heading control (IL-RMFAC) method for the underactuated AUV with unknown disturbances based on data driven. The control scheme consists of a learning control algorithm, a parameter iterative update algorithm, and a parameter reset algorithm. It is designed using only the input and output (I/O) data of the controlled system and is a data-driven control method. The pseudo partial derivative (PPD) can be iteratively calculated through the parameter iterative update algorithm and reset algorithm to adjust the learning gain, solving the problem of strictly limited initial position of the traditional fixed learning gain iterative learning control (ILC). A linear combination of angle and angular velocity is introduced in the kinematic layer to avoid overshooting of the expected following target, and an iterative learning method is introduced in the dynamics to improve the accuracy. As the number of iterations increases, the steady-state error is gradually decreased. Finally, by comparing traditional proportional–integral–derivative (PID) simulations, the proposed algorithm’s effectiveness and outstanding performance for the AUV heading tracking are confirmed.

Velocity-aware spatial-temporal attention LSTM model for inverse dynamic model learning of manipulators.

An accurate inverse dynamics model of manipulators can be effectively learned using neural networks. However, further research is required to investigate the impact of spatiotemporal variations in manipulator motion sequences on network learning. In this work, the Velocity Aware Spatial-Temporal Attention Residual LSTM neural network (VA-STA-ResLSTM) is proposed to learn a more accurate inverse dynamics model, which uses a velocity-aware spatial-temporal attention mechanism to extract dynamic spatiotemporal features selectively from the motion sequence of the serial manipulator. The multi-layer perception (MLP) attention mechanism is adopted to capture the correlation between joint position and velocity in the motion sequence, and the state correlation between hidden units in the LSTM network to reduce the weight of invalid features. A velocity-aware state fusion approach of LSTM network hidden units' states is proposed, which utilizes variation in joint velocity to adapt to the temporal characteristics of the manipulator dynamic motion, improving the generalization and accuracy of the neural network. Comparative experiments have been conducted on two open datasets and a self-built dataset. Specifically, the proposed method achieved an average accuracy improvement of 61.88% and 43.93% on the two different open datasets and 71.13% on the self-built dataset compared to the LSTM network. These results demonstrate a significant advancement in accuracy for the proposed method. Compared with the state-of-the-art inverse dynamics model learning methods of manipulators, the modeling accuracy of the proposed method in this paper is higher by an average of 10%. Finally, by visualizing attention weights to explain the training procedure, it was found that dynamic modeling only relies on partial features, which is meaningful for future optimization of inverse dynamic model learning methods.

Complete Kinematics/Dynamics Modeling and Performance Analysis of a Novel SCARA Parallel Manipulator Based on Screw Theory

Abstract In this paper, a novel Selective Compliance Assembly Robot Arm (SCARA) high-speed parallel manipulator that can realize three-translation and one-rotation motion is proposed, and an accurate dynamic modeling methodology is investigated. The mechanism is composed of four limbs with a double parallelogram structure and a single moving platform. The high bearing capacity and high dynamic response of the novel mechanism make it a viable alternative choice for this kind of automation equipment. The degree-of-freedom (DOF) of the mechanism is analyzed by the screw theory. At the same time, the velocity mapping model of the mechanism is established by the twist screw and the actuated Jacobian matrix. Then, the acceleration mapping model of the mechanism, including the generalized kinematic pairs, is established by reduced acceleration state, the modified Lie screw, and the acceleration Hessian matrix. On this basis, the complete dynamic model with a compact form of the mechanism is deduced by the combination of screw theory and virtual work principle, and the correctness of the developed model is verified by multibody simulation software. Finally, considering the inertial characteristics of the mechanism, the dynamic performance distribution in the reachable workspace of the mechanism is analyzed by the Joint-Reflected Inertia (JRI) index and Coefficient of Variation of joint-space Inertia (CVI) index, and some areas are selected as the task workspace using the above index to guarantee good dynamic performance.

Adaptive Compensation Tracking Control for Parallel Robots Actuated by Pneumatic Artificial Muscles With Error Constraints

As pneumatic artificial muscles (PAMs) are similar to biological muscles in structure and movement mechanisms, parallel robots actuated by PAMs have development prospects in rehabilitation and industry, with advantages such as compliance, high safety, strong bearing capacity, and satisfactory dynamic performance. However, the parameter uncertainties and model complexity related to inherent characteristics of parallel robots actuated by PAMs (e.g., time-varying, coupling, hysteresis, creep, and high nonlinearity), bring challenges to accurate dynamic modeling and controller design. Therefore, to achieve satisfactory tracking performance, this paper presents an adaptive compensation tracking controller with error constraints for parallel robots actuated by PAMs. The proposed controller deals with parameter uncertainties by estimating system parameters to ensure accurate tracking, which is indicated as an effective solution for a combination of PAMs and parallel robots. Furthermore, using desired trajectory signals in the complicated regression matrix, the online computational burden is significantly reduced. Moreover, to improve operation safety further, an auxiliary term with a theoretical demonstration guarantees that the tracking errors are maintained within allowable ranges. Then, the closed-loop stability is demonstrated by Lyapunov techniques. As far as we know, it is the first time that the challenges of parameter uncertainties, computational burdens, and error constraints of parallel robots actuated by PAMs are simultaneously addressed, which has both theoretical significance and practical value. Finally, the hardware experiments are implemented under different scenarios, and the results indicate that the proposed method achieves satisfactory tracking performance.

On Accurate Discrete-Time Dynamic Models of an Induction Machine

Induction machines have become the standard for highly demanding industrial applications. This has led to the utilization of modern discrete-time control techniques (such as model predictive control) that require the estimation of internal variables that are not subject to measurement (such as the rotational velocity in sensorless applications). From this point of view, it is fundamental to have accurate discrete-time models of induction machines, particularly given their nonlinear nature, so that control techniques perform according to design requirements. In spite of the above, the modeling of induction machines has not received much attention in the literature, even though more powerful machines and faster microcontrollers are currently being used. To better understand induction machine models for control, in this paper, we develop and compare various discrete-time models of induction machines based on Euler, Taylor, and Runge–Kutta methods. In addition, we compare the Extended Kalman Filter and Unscented Kalman Filter for state estimation in terms of accuracy and computational burden. The models are derived and compared through extensive Monte Carlo simulations and the state estimation techniques are compared in terms of root mean squared error, execution time, and maximum absolute error. Our simulations show that, in general, the Taylor method yields more accurate models than both the Runge–Kutta and Euler methods. In particular, the Taylor method results in a root mean square error that is one order of magnitude smaller than the Euler method for stator current and rotor flux linkages. For rotor angular speed, the Runge–Kutta methods are more accurate than both the Taylor and Euler methods, resulting in a root mean square error that is two orders of magnitude smaller than the Euler method. On the other hand, the Extended Kalman Filter results in smaller execution time than the Unscented Kalman Filter, up to two orders of magnitude. In terms of root mean squared error and maximum absolute error, both filtering algorithms perform similarly.

Modelling and control of electromechanical disc brake for heavy-duty vehicles

This paper proposes a validated procedure that can be used for development stage of model-based controller for heavy-duty electromechanical disc brakes. Firstly, system dynamics model of a single piston electromechanical disc brake is constructed in Matlab/Simulink environment in consideration of nonlinear friction model. To ensure the accuracy of system dynamics model, open loop measurements (clamping force, motor angle and motor current) are conducted on a prototype of single piston electromechanical disc brake. Experimental data is used for the determination of system variables. The predicted outputs are verified by comparison of experimental measurement results. For the control purpose of electromechanical brake, a multi-stage closed loop architecture is introduced. To regulate clamping force, PID and sliding mode controllers are developed in Matlab/Simulink in consideration of braking performance requirements for heavy-duty vehicles. For management of running clearance between brake disc and pad, PID position controller is developed in simulation environment. Those controller parameters obtained in the simulation process are introduced to the control hardware. After that closed loop clamping force and position measurements are conducted. When the rise time values of both clamping force controllers are compared, it is seen that sliding mode controller can reach to the settling point faster than PID controller. As for the position controller, the rise time requirement has been achieved by the designed PID controller. The switching process of force and position controllers are implemented and functional closed loop measurements are conducted for different reference input signals. It is seen from the results that the rise time requirement for position controller has been achieved. The system response of clamping force parameter is considerably stable.

In-plane free vibration analysis of multi-folded beam structures

Folded beam structures have been widely used in mechanical engineering, aerospace engineering, and other fields due to their flexibility and extensibility in applications. Although the components of a folded beam are simple beam structures, building an accurate dynamic model is complex due to its multi-degree-of-freedom, multi-mode, and close-mode features. This study presents a novel method to analyze in-plane free vibration of a multi-folded beam structure with arbitrary numbers of beam segments with different beam lengths and folding angles. The governing equations of the system can be easily obtained based on the beam theory and the work-energy principle. However, the key issue is how to solve the eigenvalue problem of a multi degree of freedom system defined by a 6 n × 6 n matrix. As the coefficient solution vector of the modal function is unknown, solving equations defined by high-order determinants involving variables has considerable complexity. In this work, the intricate matrix is initially turned into a single matrix of block Hessenberg form using numerous "6 × 6" block square matrices. Furthermore, by combining the characteristics of the block Hessenberg matrix with the approach to solve block determinants, a considerable reduction in the determinant's size can be attained, diminishing it from a dimension of 6 n × 6 n to 6 × 6. This procedure markedly simplifies the computational intricacies involved in the analysis. Finally, four kinds of multi-folded beams, namely L-shaped beams, Z-shaped beams, C-shaped beams, and four-segment folded beams with arbitrary lengths, folding angles, and numbers of beam segments are considered and studied, and the mode shapes and the influence of geometric parameters on the first three natural frequencies are discussed in detail. The results are also verified by the finite element method (FEM), and good agreement is found between the two methods. The results show that the proposed method can offer a unified framework for the in-plane vibration analysis of multi-folded beam structures.

Numerical Investigation of Thermal Energy Storage Systems for Collective Heating of Buildings

This study aims to investigate and identify the most effective thermal energy storage (TES) system configuration for the collective heating of buildings. It compares three TES technologies, i.e., sensible, latent, and cascade latent shell and tube storage, and examines their respective performances. A fast and accurate lumped thermal dynamic model to efficiently simulate TES system performances under different operation conditions is developed. The validation of this model’s accuracy is achieved by aligning numerical findings with data from prior experimental studies. Key findings indicated that the latent and cascade latent shell and tube storage systems demonstrate superior thermal energy storage capacities compared to the sensible configuration. Using a single-phase change material (PCM) tank increases the duration of constant thermal power storage by about 50%, and using a cascade PCM tank further enhances this duration by approximately 65% compared to the sensible TES case. Moreover, the study revealed that adjusting the PCM composition within the cascade TES significantly influenced both thermal power storage durations and pumping energy consumption. In summary, the recommended cascade PCM configuration for collective heating of buildings offers a balanced solution, ensuring prolonged stable thermal power production, elevated HTF outlet temperatures, and improved energy efficiency, presenting promising prospects for enhancing TES systems in district heating applications.

Interface stress transfer model and modulus parameter equivalence method for composite materials embedded with tensile pre-strain shape memory alloy fibers.

The constitutive model and modulus parameter equivalence of shape memory alloy composites (SMAC) serve as the foundation for the structural dynamic modeling of composite materials, which has a direct impact on the dynamic characteristics and modeling accuracy of SMAC. This article proposes a homogenization method for SMA composites considering interfacial phases, models the interface stress transfer of three-phase cylinders physically, and derives the axial and shear stresses of SMA fiber phase, interfacial phase, and matrix phase mathematically. The homogenization method and stress expression were then used to determine the macroscopic effective modulus of SMAC as well as the stress characteristics of the fiber phase and interface phase of SMA. The findings demonstrate the significance of volume fraction and tensile pre-strain in stress transfer between the fiber phase and interface phase at high temperatures. The maximum axial stress in the fiber phase is 705.05 MPa when the SMA is fully austenitic and the pre-strain increases to 5%. At 10% volume fraction of SMA, the fiber phase's maximum axial stress can reach 1000 MPa. Ultimately, an experimental verification of the theoretical calculation method's accuracy for the effective modulus of SMAC lays the groundwork for the dynamic modeling of SMAC structures.

Adaptive control of transonic buffet and buffeting flow with deep reinforcement learning

The optimal control of flow and fluid–structure interaction (FSI) systems often requires an accurate model of the controlled system. However, for strongly nonlinear systems, acquiring an accurate dynamic model is a significant challenge. In this study, we employ the deep reinforcement learning (DRL) method, which does not rely on an accurate model of the controlled system, to address the control of transonic buffet (unstable flow) and transonic buffeting (structural vibration). DRL uses a deep neural network to describe the control law and optimizes it based on data obtained from interaction between control law and flow or FSI system. This study analyzes the mechanism of transonic buffet and transonic buffeting to guide the design of control system. Aiming at the control of transonic buffet, which is an unstable flow system, the control law optimized by DRL can quickly suppress fluctuating load of buffet by taking the lift coefficient as feedback signal. For the frequency lock-in phenomenon in transonic buffeting flow, which is an unstable FSI system, we add the moment coefficient and pitching displacement to feedback signal to observe pitching vibration mode. The control law optimized by DRL can also effectively eliminate or reduce pitching vibration displacement of airfoil and buffet load. The simulation results in this study show that DRL can adapt to the control of two different dynamic modes: typical forced response and FSI instability under transonic buffet, so it has a wide application prospect in the design of control laws for complex flow or FSI systems.

Identification of the friction model of a single elastic robot actuator from video

Obtaining a highly accurate dynamic model from video data is a challenging task within system Identification. More specifically, in robotics, there is a constant need for precise models to optimize the performance and safety of these systems. With the developments in machine learning, algorithms such as convolutional neural networks (CNNs) have shown good performances in extracting features from images. Techniques like transfer learning (TL) can lower the computational cost and optimize the training task. This work aims to make the gray-box Identification of friction parameters of an actuator from video. The proposed approach is composed of three steps. Firstly, it compares the transfer learning performance of three pre-trained CNNs to estimate the pose of the motor and the link from video data. Secondly, a long-term short-term (LSTM) network is added to estimate the velocity and acceleration. Thirdly, the parameters of the two friction models are estimated using the estimated states and the dynamic equations of the system. Through these steps, a vision-based gray-box Identification is obtained. The approach is tested using an original elastomer-based series of elastic actuators (eSEA) benchmark. The results show that the VGG19 model combined with the LSTM layer is the optimal pre-trained CNN for the vision module, obtaining a coefficient of determination higher than 0.96 for all three states. From the state estimates, the parameters of the friction models are optimized. The best-performing friction model is the Coulomb-Stribeck, which decreases the mean absolute error (MAE) by 15.50% compared to the viscous linear model.