Chenbo Yin

The return oil filter in an excavator hydraulic system plays a critical role in maintaining oil cleanliness, and its blockage fault directly affects the thermal state and operational reliability of the machine. Owing to the progressive evolution and concealed characteristics of this fault, accurate identification remains challenging. To address this issue, this study proposes a return oil filter blockage fault identification model combining the whale optimization algorithm (WOA), multilayer perceptron (MLP), and gated recurrent unit (GRU). First, the hydraulic system configuration, operating process, and blockage mechanism of the return oil filter are analyzed. Then, multi-source operational variables are selected according to the fault propagation mechanism, while ACF and FFT analyses are employed to support signal denoising. In addition, variational mode decomposition (VMD) is introduced for representative pressure signals to provide supplementary evidence for fault evolution and the rationality of variable selection. On this basis, WOA is used to optimize the key hyperparameters of the hybrid model. Finally, the proposed scheme is systematically validated. Experimental results show that WOA-MLP-GRU achieves the best overall performance among the compared models, with an identification accuracy of 98.3%. Moreover, the proposed model exhibits lower validation loss, better feature separability, and a more concentrated error distribution, demonstrating superior training stability, robustness, and generalization capability.

Distributed electro-hydraulic pump-controlled excavators offer significant potential for improving energy efficiency by eliminating throttling losses and decoupling multi-actuator interactions. However, the highly nonlinear dynamics, time-varying loads, and parameter uncertainties inherent in excavator operations pose substantial challenges to achieving high-precision trajectory control in such systems. This paper proposes a distributed independent electro-hydraulic servo pump-controlled excavator and develops a unified adaptive robust trajectory control framework to address these challenges. An adaptive robust backstepping controller is constructed by explicitly incorporating hydraulic driving force dynamics into the trajectory tracking loop, enabling coordinated regulation of displacement, velocity, and force. To further enhance robustness against time-varying operating conditions, a novel feedforward compensation strategy is introduced by experimentally identifying the mapping relationships between hydraulic cylinder displacement and both load force and inertial mass. The feedforward mechanism is integrated with a projection-based adaptive law, forming a unified parameter adaptation scheme capable of handling strong nonlinearities and uncertain dynamics. In addition, a quadrant-identification mechanism is incorporated to ensure consistent control performance across different operating modes of the pump-controlled excavator. Comprehensive experimental studies, including step response tests, trajectory tracking of boom and arm cylinders, and practical leveling operations, are conducted to validate the proposed method. The results demonstrate that the proposed strategy significantly improves tracking accuracy, dynamic response, and disturbance rejection capability compared with conventional PID and non-adaptive robust control approaches.

Distributed independent pump-controlled systems enhance excavator energy efficiency by reducing throttling losses, but sustained high-load and high-speed operation often forces a single pump into low-efficiency regions. To overcome this limitation, this study proposes a dual-pump load-sharing architecture that enables load sharing and dynamic cooperative operation of two pumps within the high-efficiency region. An adaptive closed-loop GOA-SQP optimization strategy is also developed for real-time pump speed distribution. The proposed strategy combines the global exploration capability of the Grasshopper Optimization Algorithm (GOA) with the rapid local convergence characteristics of Sequential Quadratic Programming (SQP). Unlike conventional serial hybrid optimization methods, a bidirectional efficiency-feedback mechanism is introduced to dynamically coordinate global exploration and local refinement processes in real time. Furthermore, a local perturbation and re-explosion mechanism is incorporated to suppress premature convergence, enhance population diversity, and reduce redundant iterations. Experimental validation on an excavator test platform under four-quadrant conditions shows that the proposed system improves mechanical, volumetric, and overall pump efficiencies by 14.22, 4.57, and 18.79 percentage points, respectively, and reduces total energy consumption by 8.83%. Compared with GOA, SQP, GA-SQP, and PSO-SQP, the proposed GOA-SQP algorithm reduces solution time by 39.15% and improves optimization accuracy by 0.5%. The proposed architecture and optimization strategy offer a novel and effective solution for further improving the energy efficiency of distributed independent pump-controlled excavator systems.

With the continuous development of engineering machinery technology, excavators, as a key engineering equipment, have played an important role in the fields of construction and mining. Accurate identification of excavator working objects is crucial to realize their automated operation. However, traditional working object recognition methods have obvious limitations in practical applications due to problems such as complex computational models and susceptibility to lighting conditions. To meet the above challenges, this paper proposes a CNN-LSTM-Adaboost neural network model based on Bayesian optimization (BO) for excavator working object recognition. The model integrates the feature extraction capability of convolutional neural networks (CNN), the modeling advantage of long short-term memory networks (LSTM) for time series data, and the ability of Adaboost algorithm to build strong classifiers by integrating weak classifiers. At the same time, BO is used to automatically obtain the optimal hyperparameter combination. Experimental results show that compared with other comparison models, the BO-CNN-LSTM-Adaboost model has significant advantages in comprehensive performance.

Due to the low efficiency of internal combustion engine (ICE) and throttling losses of load-sensing (LS) system, conventional excavators which employed ICE with LS system as power source have high energy waste. To avoid this problem, a distributed independent electro-hydraulic actuator (iEHA) that directly driven by a electric motor is designed for excavators. And a speed control strategy based on mode switching and pump map is proposed. The results of proposed control strategy are carried out to verify the velocity stability. Besides, different signal delays for switching valve are explored to obtain better control characteristics. Furthermore, the influence of different precharged pressure of accumulator is explored when boom and arm share an accumulator. Results demonstrate that the energy efficiency of iEHA system improves from 19.19% to 59.56% compared to load-sensing system, the comprehensive response performance is better when signal delay time is 0.1s and response speed of system increases by 16.6% compared to no delay, the system can achieve good results when the precharged pressure is set to 20 bar.

With the development of hardware and software, the pressure control of construction machinery embraced new characteristics of sensors failure. In recent years, soft sensors have been widely used in construction machinery for fault-tolerant controlling. To cope with failure of some pressure sensors which cause great harm to the electro-hydraulic proportional control of excavators and accurately measure hydraulic pressure without actual pressure sensors, a pressure soft sensor based on Long Short-Term Memory (LSTM) neural network is proposed in this work. And the Bayesian Optimization (BO) is used to tune the hyperparameters of LSTM. The hydraulic system of a SY375IDS excavator is analyzed and the Mutual Information (MI) strategy is adopted to determine features which are given to LSTM. For case of multiple sensors fail at the same time, a (multiple input multiple output) MIMO prediction mechanism based on BO-LSTM and a MIMO cyclic prediction mechanism based on two (multiple input single output) MISO models are proposed. Finally, the performance of the proposed pressure soft sensor is evaluated in the compared experiments with BP, CNN and RnN. Results shows that the proposed method is superior to other methods in terms of pressure prediction.

When a robotic excavator is controlled by a human operator, it must comply with the intention of the operator. Besides, it must meet the safety and operational quality of construction indicators. Coordinating the relationship between the active decision of the operator and the accurate control of the machine system is the key issue of this paper. Flexible virtual fixtures are proposed to control trajectory of the robotic excavator, the input signal of the restricted motion direction is weakened by introducing the flexibility coefficient kτ. In this way, a human-robot shared control system for the robotic excavator is built. Trajectories are designed to validate the performances of the flexible virtual fixtures, and experiments are performed on the trajectory control and motion visualization system which is built based on virtual instrument. The results show that the flexible virtual fixtures reduce the tracking errors and the operator can avoid the obstacle on the trajectory.