Research and experiments on the performance optimization of hydraulic impact hammers Authors: Hu Chen, Boqiang Shi, Hui Guo, Zirui Liu, Bingbing Liu Published: 2026 ABSTRACT AbstractIn order to enhance the impact performance and stability of hydraulic impact hammers, this paper first conducts a Design of Experiments (DOE) study on hydraulic impact hammers, identifying ten key parameters that influence their impact performance and stability. Consequently, multi-objective optimization is employed to enhance the impact performance and stability of hydraulic impact hammers. In comparison with the baseline configuration, the optimized configuration attains a 4.6% increase in impact power and an 11.3% reduction in working pressure. Furthermore, a performance analysis surrogate model for the hydraulic impact hammer was developed to enhance the computational efficiency of multi-objective optimization. Research indicates that the errors in impact power and working pressure between the optimized configuration of the surrogate model trained using Elliptic Basis Functions (EBF) and that of the traditional model are minimal, at 0.50% and 0.30%, respectively. In comparison with the baseline configuration, the EBF-trained surrogate model optimization configuration increases impact power by 4.1% and reduces working pressure by 11.0%. Finally, experimental validation confirms the accuracy of both the traditional model and surrogate model optimization configurations. In summary, this study provides a reliable theoretical foundation and experimental support for the multi-objective optimization of hydraulic impact hammers, offering a clear paradigm for their multi-objective optimization design.Subject terms: Energy science and technology, Engineering, Mathematics and computing Introduction The hydraulic impact hammer is composed of a piston, reversing valve, an accumulator, cylinders, and a drill rod. It is a highly versatile tool that is employed in a variety of industrial settings, including mining and infrastructure construction. The piston reciprocates under the control of the reversing valve, thereby converting the fluid’s pressure energy into the piston’s impact energy. This, in turn, drives the piston to strike the drill rod and subsequently break the rock. Impact performance and stability are critical metrics for assessing the efficacy of hydraulic impact hammers. These two characteristics exhibit contrasting characteristics. Hydraulic impact hammers are characterized by intricate designs with pronounced nonlinear characteristics. This characteristic complicates the development of performance optimization solutions and renders the validation process arduous. Consequently, research focused on the optimization of their performance is of critical importance. A number of scholars have developed mathematical models of hydraulic impact systems for the purpose of studying their performance. Oh et al.1 established a mathematical model of a hydraulic impact system as a tool for analyzing impact performance. Ye et al.2 developed a mathematical model of a hydraulic impact hammer and investigated methods of increasing both impact frequency and impact power. Yang et al.3 constructed a nonlinear mathematical model of a hydraulic impact hammer and identified the optimal values for the critical structural parameters of the piston and the reversing valve. Li et al.4 constructed a simulation model of a hydraulic impact hammer and examined the impact of system oil pressure and nitrogen chamber pre-charge pressure on performance. Hu et al.5 developed a mathematical model of a hydraulic impact system based on quasi-steady theory and finite difference theory. Guo et al.6 developed a nonlinear mathematical model of a hydraulic impact system and investigated the influences of inlet oil flow, pre-charge pressure for an accumulator with high pressure, and system setting pressure on impact system performance. Franco et al.7 developed a simplified calculation model of a hydraulic impact hammer to support product design. Redelin et al.8 constructed a mathematical model for the hydraulic impact hammer and studied the effect of accumulator initial filling pressure on performance. Galdin et al.9 developed a mathematical model of a hydropneumatic impact device and studied the relationship between impact energy and pre-charge pressure of the accumulator. Giuffrida et al.10 developed a nonlinear mathematical model of the hydraulic impact system and completed experimental validation, thereby establishing the theoretical foundation for simulation optimization. Yang et al.11 investigated the effects of system flow rate, accumulator pressure, and relief valve set pressure on impact energy and efficiency, providing a reference for multi-parameter coupled optimization. Furthermore, a number of researchers have directed their efforts toward the optimization of hydraulic impact systems. Shin et al.12 enhanced both impact frequency and energy through parameter optimization. Xu et al.13 investigated key parameters for increasing impact energy in hydraulic systems and optimized piston diameter. Kim et al.14 employed a stepwise quadratic response surface method to optimize the design parameter set and proposed design modifications to enhance the impact performance of hydraulic impact systems. Noh et al.15 addressed the trade-off between impact energy and frequency by proposing a multi-objective optimization strategy based on sensitivity analysis. The construction of a weighted objective function was the key to achieving simultaneous improvement in both parameters. Rechkin16 applied CFD modeling to the design and research of pneumatic and hydraulic impact devices, analyzing complex flow field characteristics during impactor operation through numerical simulation to provide theoretical foundations for structural optimization. Baek et al.17 introduced Taguchi methods for orthogonal experimental design of hydraulic impact systems, determining optimal combinations of design variables by analyzing signal-to-noise ratios. This approach effectively resolves matching issues in multi-parameter optimization. In recent years, there has been a progressive application of several advanced algorithms to the optimization calculations of hydraulic impact systems. Hojjati et al.18 proposed a hybrid evolutionary algorithm based on genetic expression programming and particle swarm optimization to analyze data, providing an accurate performance prediction model for hydraulic impact systems. Cao et al.19 proposed a multi-objective optimization method for air hammer structure design based on the Microhabitat Quantum Genetic Algorithm (MNQGA), enabling simultaneous or separate optimization to maximize impact energy and minimize air consumption. Frid et al.20 conducted a multi-domain joint optimization study for hydraulic impact systems, comparing composite shape methods, particle swarm optimization, and differential evolution. The findings of the study demonstrated that differential evolution exhibited optimal convergence efficiency in the resolution of high-dimensional nonlinear optimization problems. Furthermore, research on optimizing complex problems through surrogate modeling is becoming increasingly prevalent, enhancing computational efficiency in hydraulic impact system performance optimization. Yang et al.21 developed a flow area parameter model based on CFD data to replace time-consuming numerical simulations. When employed in conjunction with the NSGA-II algorithm, this approach led to the optimization of spool valve structures across multiple objectives, thereby markedly enhancing computational efficiency. In summary, previous studies have developed mathematical models for hydraulic impact hammers and examined the parameters that influence their performance. Nevertheless, several issues persist. Firstly, extant research has not examined the contribution of structural and working parameters to performance responses. The prevailing focus has been on the impact of single-parameter changes, with a paucity of multi-parameter synergistic analyses. Furthermore, studies on multi-objective optimization for hydraulic impact hammer performance have been relatively scarce, and there has been no investigation into enhancing the computational efficiency of multi-objective optimization solutions. Presently, there is a paucity of systematic multi-parameter collaborative quantitative analysis frameworks and efficient optimization strategies for typical high-dimensional nonlinear systems, such as hydraulic impact hammers. Consequently, the efficacy of global optimization within strongly coupled design spaces remains limited, and the identification of engineering-feasible solutions is challenging to precise. The objective of this study is to enhance the performance of hydraulic impact hammers. This study integrates optimized Latin hypercube sampling (OLHS), the EBF surrogate model, and entropy-weight-based TOPSIS decision-making into an integrated “sampling–modeling–optimization–decision” framework to address the optimization problem of strong multi-parameter coupling and conflicting objectives in hydraulic impact hammers. Through a meticulous examination of the contributions of 18 distinct parameters to impact performance and stability, this research enhances both domains through the implementation of multi-objective optimization. Furthermore, the efficiency of addressing multi-objective optimization problems is enhanced by the establishment of a surrogate model, and the precision of the optimized configuration is validated through experimental verification. This study verifies the superiority of the EBF surrogate model in the multi-objective optimization of hydraulic impact hammers and provides a reusable technical paradigm for the intelligent design of similar complex hydraulic systems. In summary, this work provides a reliable theoretical foundation and experimental support for optimizing the performance of hydraulic impact hammers. The technical framework of this paper is illustrated in Fig. 1. Technical framework of the proposed methodology. Numerical model The schematic cross-section of a hydraulic impact hammer is presented in Fig. 2. The movement of the piston is facilitated by the combined forces of hydraulic oil and nitrogen gas, while the motion of the reversing valve is driven by hydraulic oil. The front chamber is invariably connected to the inlet pipe and maintains elevated pressure. The back chamber is sometimes connected to the inlet pipe for high pressure and sometimes connected to the outlet pipe for low pressure. The pressure within the nitrogen chamber is subject to variation in accordance with the piston’s position. The accumulator is invariably connected to the inlet pipe and functions as a high-pressure accumulator. Schematic cross-section of a hydraulic impact hammer. When the front chamber is subjected to high pressure and the back chamber is subjected to low pressure, the resultant force on the piston directs the piston upward, causing the return stroke, as illustrated in Fig. 2(a). Upon attaining a specific position during the return stroke, an increase in pressure within the signal chamber occurs, thereby inducing a positional change in the reversing valve. This establishes a connection between the back chamber and high pressure, thereby directing the resultant force in a downward direction and initiating the impact stroke, as illustrated in Fig. 2(b). Upon attaining a particular position during the stroke, the piston triggers a decrease in pressure within the signal chamber, thereby initiating the reset of the reversing valve. Consequently, the back chamber is subjected to low pressure, prompting the piston to execute the return stroke once more, thus perpetuating the cycle. (1) Dynamic equations of the piston and the reversing valve. where, Where, MP is the mass of the piston, kg; Fvp is the viscous friction of the piston, N; Fcp is the hydraulic clamping force of the piston, N; Fsp is the seal friction of the piston, N; Fap is the hydraulic axial force of the piston, N; MV is the mass of the reversing valve, kg; Fvv is the viscous friction of the reversing valve, N; Fcv is the hydraulic clamping force of the reversing valve, N; Fav is the hydraulic axial force of the reversing valve, N; yp is the displacement of the piston, mm; Pin is the inlet pipe pressure, Pa; Pout is the outlet pipe pressure, Pa; Pf is the front chamber pressure, Pa; Pb is the back chamber pressure, Pa; Ps is the signal chamber pressure, Pa; PVi is the Vi (i = 1, 2, 3, 4) chamber pressure, Pa; Pn is the nitrogen chamber pressure, Pa. (2) Gas state equations for the accumulator chamber and the nitrogen chamber. Where, Pac is the accumulator pressure, Pa; Pac0 is the accumulator precharge pressure, Pa; Vac0 is the accumulator initial filling volume, m3; Pn0 is the nitrogen chamber precharge pressure, Pa; Vn0 is the initial filling volume of nitrogen chamber, m3. (3) Flow equations for the inlet and outlet pipes. Where, QXtY is the X-chamber to Y-chamber flow rate, L/min; QXlY is the X-chamber to Y-chamber leakage flow rate, L/min. (4) Flow equations for the front chamber, back chamber and signal chamber. (5) Flow equations for the V1, V2, V3 and V4 chambers of the reversing valve. where, (6) Impact power. Research on design of experiments Working pressure () and impact power () are the most critical indicators reflecting the stability and impact performance of a hydraulic impact hammer. Working pressure refers to the maximum pressure in the front chamber during the piston return stroke. Lower working pressure indicates higher stability. Greater impact power signifies better impact performance. To analyze the influence of typical structural parameters and operating parameters on working pressure and impact power, a Design of Experiments (DOE) study is conducted based on an established numerical calculation model for hydraulic impact hammers. Latin hypercube sampling (LHS) ensures uniform distribution in the one-dimensional projection of each parameter dimension through a stratified random sampling strategy. However, conventional LHS may suffer from sample clustering. Optimized Latin hypercube sampling (OLHS) improves the space-filling performance based on the Maximin criterion, and the optimization process is implemented using the simulated annealing algorithm. To more accurately capture the overall characteristics of the hydraulic impact hammer, an OLHS method is employed, with the number of experiments set to 20 times the number of study factors. Taking a specific model of hydraulic impact hammer as an example, the range of values for the study factors is determined based on engineering practice, as presented in Table 1. Table 1Range of values for study factors.Study FactorsUnitDefaultLower BoundUpper BoundAccumulator precharge pressure bar75.0050.00100.00Piston back diameter mm208.50207.00210.00Piston middle diameter mm234.50233.00236.00Reversing valve bottom outer diameter mm114.00114.00114.50Reversing valve bottom inner diameter mm98.0098.0098.50Reversing valve top outer diameter mm116.40116.00116.50Reversing valve top inner diameter mm99.4099.1099.50Low-pressure oil groove depth mm1.501.001.50Low-pressure oil groove length mm5.002.0010.00Low-pressure oil groove width mm5.002.005.00High-pressure oil groove depth mm1.201.001.50High-pressure oil groove length mm7.002.0010.00High-pressure oil groove width mm5.002.005.00Piston front-mid section length mm160.50155.00195.00Nitrogen chamber precharge pressure bar33.5030.0035.00Piston stroke mm210.50200.00220.00Reversing valve stroke mm28.0025.0030.00Reversing valve low-pressure port length mm8.005.0010.00 Range of values for study factors. Using the OLHS method, 360 samples were taken for 18 study factors. After calculation based on the hydraulic impact hammer numerical model, the coefficient of determination (R2) for working pressure is 0.99902, and the R2 for impact power is 0.85862. The difference between the R2 values of working pressure and impact power arises because the working pressure response is mainly dominated by parameters such as the piston front diameter, piston middle diameter, piston back diameter, and nitrogen chamber precharge pressure, making it easier for the surrogate model to fit. In contrast, the impact power response involves strongly nonlinear coupling effects among the front chamber pressure, back chamber pressure, and nitrogen chamber pressure. This indicates that the aforementioned DOE study results accurately fit the influence of these study factors on working pressure and impact power. The main effects and response contributions of the study factors on working pressure and impact power are presented in Figs. 3 and 4, respectively. Main effects of the study factors on working pressure and impact power. Response contributions of the study factors on working pressure and impact power. As illustrated in Figs. 3 and 4, the most significant negative correlation is observed between the piston middle diameter and the working pressure, at 35.34%. The most significant positive correlation is between the nitrogen chamber precharge pressure and the working pressure, at 26.61%. Furthermore, the piston back diameter, accumulator precharge pressure, piston stroke, and low-pressure oil groove width significantly influence the working pressure, with response contributions of 13.17%, 6.78%, 6.07%, and 3.64%, respectively. The lower inner diameter of the reversing valve demonstrates a substantial positive correlation with impact power at 14.26%. The most significant negative correlation is observed between the upper inner diameter of the reversing valve and impact power, at 12.38%. Furthermore, the middle diameter of the piston, the length of the middle-lower section of the piston, the precharge pressure of the nitrogen chamber, and the stroke of the reversing valve significantly influenced impact power, with response contributions of 11.88%, 10.31%, 9.95%, and 6.16%, respectively. A thorough examination of the response contributions of each study factor to both impact power and working pressure indicates that the piston middle diameter and nitrogen chamber precharge pressure exert a substantial influence on both parameters. Furthermore, the piston back diameter, accumulator precharge pressure, piston stroke, and low-pressure oil groove width have been demonstrated to significantly influence working pressure. The lower inner diameter of the reversing valve, the upper inner diameter of the reversing valve, the piston middle-lower section length, and the reversing valve stroke have been found to significantly affect impact power. It is evident that these ten parameters play a pivotal role in determining the performance of hydraulic impact hammers. Multi-objective optimization Design variables In order to reduce the design space dimensions and decrease computational demands for optimization, 10 key parameters from the 18 parameters studied in the DOE are selected as design variables, that is, . According to the prevailing engineering practices, the design variables and their respective ranges for multi-objective optimization are presented in Table 2. Table 2Range of values for design variables.Study FactorsUnitDefaultLower BoundUpper Bound bar75.0050.00100.00 mm208.50207.00210.00 mm234.50233.00236.00 mm98.0098.0098.50 mm99.4099.1099.50 mm5.002.005.00 mm160.50155.00195.00 bar33.5030.0035.00 mm210.50200.00220.00 mm28.0025.0030.00 Range of values for design variables. Constraints In the field of engineering, it has been observed that when the working pressure (), back chamber peak pressure (), and outlet pipe peak pressure () reach levels that are disproportionately high, there is the potential for these factors to exert a deleterious effect on the stability of hydraulic impact hammers and excavator hydraulic systems. In severe cases, this phenomenon can result in piston scuffing and excavator vibration issues. Consequently, it is imperative to constrain the working pressure to a maximum of 320 bar, the back chamber peak pressure to a maximum of 450 bar, and the outlet pipe peak pressure to a maximum of 50 bar. The following constraint conditions are established in this study: Optimization objectives The objective of this study is to develop a refined design scheme for hydraulic impact hammers, with the goal of achieving both superior impact performance and enhanced stability. Impact power is the most critical indicator of hydraulic impact hammer impact performance, and higher impact power is associated with better impact performance. The working pressure of a hydraulic impact hammer is a critical indicator of its stability. A lower working pressure is indicative of higher stability. Consequently, impact power and working pressure are adopted as evaluation metrics for the impact performance and stability of the hydraulic impact hammer. The objective of this study is to optimize the hydraulic impact hammer to achieve greater impact power at lower working pressure. In order to ensure that dimensional and scale differences between impact power and working pressure are eliminated, thus facilitating the attainment of fair, reasonable, and interpretable optimization results, it is imperative that the optimization target values for impact power and working pressure be normalized while adhering to the maximization principle. According to the results of the previous DOE study, the value ranges for impact power and working pressure can be determined. This determination establishes their optimization target values. The optimization target value for impact power is , and that for working pressure is . Therefore, for the purposes of research and analysis, the optimization target is hereby established as follows: Optimization Results The parameters are set as follows when employing the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) with multi-objective optimization: The population size has been set to 100, the number of generations to 100, the crossover probability to 0.9, and the crossover distribution index to 20. Following the implementation of the aforementioned computations, the Pareto frontier for and is obtained, as presented in Fig. 5. Pareto frontier for and . As illustrated in Fig. 5, the Pareto frontier for and is continuous, indicating sufficient computational iterations. Additionally, a substantial negative correlation has been identified between and , indicating that elevated work pressure is associated with increased impact power. Employing the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) with entropy weighting, the points with the highest relative proximity on the Pareto frontier for and are calculated. This calculation derives the optimal solution for the multi-objective optimization of working pressure and impact power in hydraulic impact hammers. Preliminary calculations indicate that ‘s weight is 0.5494, ‘s weight is 0.4506, and the maximum relative proximity is 0.6058. (0.7929, 0.4818) is the point on the Pareto frontier with the maximum relative proximity. Consequently, the optimized configuration attains a working pressure of 278.64 bar and an impact power of 81.68 kW. The parameters for the baseline and optimized configurations of the hydraulic impact hammer are presented in Table 3. Table 3Parameters for the baseline and optimized configurations.ParametersUnitBaseline ConfigurationOptimized Configuration bar75.0071.46 mm208.50210.00 mm234.50235.99 mm98.0098.23 mm99.4099.21 mm5.002.02 mm160.50182.89 bar33.5030.09 mm210.50200.02 mm28.0029.99 kW78.1081.68 bar314.00278.64 Parameters for the baseline and optimized configurations. As illustrated in Table 3, in comparison with the baseline configuration, the optimized configuration attains a 4.6% increase in impact power and an 11.3% decrease in working pressure. This finding indicates that the optimized configuration enhances the impact performance and stability of the hydraulic impact hammer. Multi-objective optimization of surrogate models Establishing surrogate models The complexity of the numerical calculation model for hydraulic impact hammers, in conjunction with both the significant computational expense and the inefficient solution efficiency of multi-objective optimization, necessitates the establishment of a surrogate model of hydraulic impact hammer performance analysis of multi-objective optimization. The hydraulic impact hammer performance analysis surrogate model is capable of extracting the mapping relationship between input and output parameters from data. This facilitates the acquisition of simulation results with a high degree of accuracy at a low computational cost. This approach offers an efficient and feasible solution for multi-objective optimization calculations of hydraulic impact hammers. First, a global exploration of the 10 design variables for multi-objective optimization is conducted through DOE research, employing the OLHS method with a sample size 100 times the number of design variables. It is noteworthy that surrogate models typically exhibit maximum prediction errors at the boundaries, and the incorporation of corner points can significantly mitigate boundary errors. Consequently, the global exploration must incorporate the corner points of the design variables, which totals 2n points, where n denotes the number of design variables. The full-factor sampling method is employed to sample 1,024 corner points across the 10 study factors. The results from full-factor sampling and OLHS are combined, yielding a total of 2,024 sampling points. A surrogate model is established based on these sampling points, and then multi-objective optimization is performed using the NSGA-II algorithm. There are numerous methods for establishing surrogate models, with response surface modeling (RSM), radial basis functions (RBF), elliptic basis functions (EBF), and Kriging being common approaches. It is important to note that RSM is only suitable for simple quadratic nonlinear models. RBF demonstrates particular aptitude in the management of high-dimensional nonlinear models and the interpolation of sparse data. EBF exhibits remarkable proficiency in accommodating intricate surfaces. Kriging is well-suited for deterministic analysis and has proven to be highly effective in capturing local nonlinearities. In light of the aforementioned global exploration sampling points, performance analysis surrogate models for the hydraulic impact hammer are trained using RSM, RBF, EBF, and Kriging, respectively. The error analysis is conducted via cross-validation, with the prediction errors of the output parameters for the four surrogate models presented in Table 4. Table 4Prediction errors of the output parameters for the surrogate models.MAEMaxAERMSER2 RSM 0.043490.160050.059430.96671 0.006690.022060.008490.99894 0.058750.180040.075240.95627 0.118570.269160.418750.81977 RBF 0.023220.088500.030890.98119 0.005130.023420.007640.99896 0.025460.060650.030230.98963 0.048680.201800.070580.93691 EBF 0.006000.020860.008130.99885 0.006450.020100.008450.99926 0.018550.099320.030010.98605 0.025820.149670.044240.97302 Kriging 0.140660.435800.173660.56282 0.166770.414870.208050.47030 0.138700.720030.207560.37888 0.149580.514610.209620.26739 Prediction errors of the output parameters for the surrogate models. As illustrated in Table 4, the surrogate model trained using EBF attains the highest prediction accuracy for output parameters. Specifically, the R2 value for impact power is 0.99885, while that for working pressure is 0.99926. Conversely, the surrogate model that was trained using Kriging demonstrates the most significant prediction error for output parameters. Multi-objective optimization analysis of surrogate models Multi-objective optimizations of the impact performance of a hydraulic impact hammer are conducted based on four surrogate models. The design variables, boundary conditions, and optimization objectives are consistent with those delineated in Sect. 4. Furthermore, the multi-objective optimization algorithm NSGA-II is employed with the following parameters: the population size set to 100, the number of generations set to 100, the crossover probability set to 0.9, and the crossover distribution index set to 20. The Pareto frontiers for and of the traditional model and the four surrogate models are presented in Fig. 6. Pareto frontiers of traditional and surrogate models. The proximity between the Pareto frontier of the surrogate model and that of the traditional model is evaluated by means of Generation Distance, Inverse Generation Distance, and Hypervolume Ratio. Generation Distance is defined as the extent to which points on the surrogate model’s Pareto frontier are displaced from those of the traditional model’s Pareto frontier. This is calculated as the mean Euclidean distance between each point on the surrogate model’s Pareto frontier and the nearest point on the traditional model’s Pareto frontier. Inverse generation distance is a metric that quantifies the distance between points on the traditional model’s Pareto frontier and the closest points on the surrogate model’s Pareto frontier. This is calculated as the mean Euclidean distance between each point on the traditional model’s Pareto frontier and the nearest point on the surrogate model’s Pareto frontier. Hypervolume ratio is defined as the ratio of the area enclosed by the Pareto frontier of the surrogate model and the reference point, to the area enclosed by the Pareto frontier of the traditional model and the reference point. A hypervolume ratio greater than 100% is indicative of an optimistic prediction, while a ratio less than 100% is indicative of a negative prediction. The proximities of the Pareto frontiers of the four surrogate models to the Pareto frontier of the traditional model are presented in Table 5. Table 5Proximity of the Pareto frontiers of the surrogate and traditional models.RSMRBFEBFKriging Generation Distance 0.0016020.0003070.0004420.002933 Inverse Generation Distance 0.0448640.0067110.0120950.058609 Hypervolume Ratio 106.12%98.66%98.59%106.25% Proximity of the Pareto frontiers of the surrogate and traditional models. As illustrated in Table 5, the generation distance between the Pareto frontier of the RBF-trained surrogate model and that of the traditional model is 0.000307, with an inverse generation distance of 0.006711 and a hypervolume ratio of 98.66%. The generation distance between the Pareto frontier of the EBF-trained surrogate model and that of the traditional model is 0.000442, with an inverse generation distance of 0.012095 and a hypervolume ratio of 98.59%. It is evident that both the generation distance and reverse generation distance between the Pareto frontiers of the RBF-trained and EBF-trained surrogate models and the traditional model are less than 0.02, while the difference between their hypervolume ratios and 100% are less than 2%, respectively. This finding suggests that the Pareto frontiers of the RBF-trained and EBF-trained surrogate models are highly proximate to those of the traditional model. Employing the TOPSIS method optimized with the entropy weight approach, the points with the highest relative proximity on the Pareto frontier for the four surrogate models are calculated. The parameters of the surrogate models optimization configuration are obtained, as presented in Table 6. Table 6Parameters for the optimized configurations of the surrogate models.ParametersUnitRSMRBFEBFKriging bar99.9866.0770.1975.27 mm209.92209.95210.00209.18 mm236.00235.97235.99235.29 mm98.3898.3498.2398.26 mm99.5099.2799.2499.24 mm2.312.042.002.92 mm185.08179.57189.80181.36 bar30.0232.2530.1930.85 mm200.04201.86200.11208.89 mm29.7727.7829.1728.02 kW82.2887.5581.2885.58 bar274.34292.68279.46281.06 Parameters for the optimized configurations of the surrogate models. Errors of the optimized configurations of the surrogate models The errors in impact power and working pressure between the optimized configurations of the four surrogate models and the optimized configuration of the traditional model are presented in Table 7. Furthermore, the corresponding impact power and working pressure of the optimized configurations of the four surrogate models were calculated using the traditional model. The errors in impact power and working pressure between these results and the optimized configuration of the traditional model are presented in Table 8. Table 7Errors in impact power and working pressure between the optimized configurations of the surrogate models and the optimized configuration of the traditional model.RSMRBFEBFKriging- Error0.73%7.19%0.50%4.77%- Error1.54%5.04%0.30%0.87% Errors in impact power and working pressure between the optimized configurations of the surrogate models and the optimized configuration of the traditional model. Table 8Errors in impact power and working pressure between the optimized configuration of the surrogate models calculated by the traditional model and the optimized configuration of the traditional model.RSMRBFEBFKriging 78.01 kW86.66 kW82.00 kW84.28 kW- Error4.50%6.09%0.39%3.18% 272.99 bar291.83 bar280.24 bar293.74 bar- Error2.03%4.73%0.57%5.42% Errors in impact power and working pressure between the optimized configuration of the surrogate models calculated by the traditional model and the optimized configuration of the traditional model. As illustrated in Table 7, the errors in impact power and working pressure between the optimized configuration of the EBF-trained surrogate model and that of the traditional model are at most minimal, at 0.50% and 0.30%, respectively. As illustrated in Table 8, after validation using the traditional model, the errors in impact power and working pressure between the optimized configuration of the EBF-trained surrogate model and that of the traditional model are at most minimal, at 0.39% and 0.57%, respectively. Furthermore, the EBF-trained surrogate model optimization configuration, in comparison with the baseline configuration, increases impact power by 4.1% and reduces working pressure by 11.0%. As a result, the impact performance and stability of the hydraulic impact hammer are enhanced. This finding indicates that the EBF-trained surrogate model optimization configuration demonstrates a high degree of reliability. Furthermore, the hybrid basis function structure of the EBF model can simultaneously capture the quasi-linear trend of working pressure and the transient nonlinear characteristics of impact power in the hydraulic impact hammer system, thereby avoiding the representational limitations of a single basis function. In addition, the lightweight prediction mechanism of the EBF model significantly reduces the computational cost of multi-iteration optimization, providing a practical tool for rapid engineering design. In summary, employing the EBF-trained performance analysis surrogate model for multi-objective optimization of impact power and working pressure achieves exceptional accuracy while significantly outperforming traditional models in computational speed. This approach offers an efficient and viable solution for multi-objective optimization calculations of hydraulic impact hammers. Experimental validation In order to ascertain the actual impact performance optimization of hydraulic impact hammers and to verify the accuracy of both the optimized configuration of the traditional model and the optimized configuration of the surrogate model, this study conducted working pressure experiments and impact power experiments on hydraulic impact hammers. The experimental system primarily consisted of an excavator, a hydraulic impact hammer, pressure sensors, a high-speed camera, and a data acquisition instrument, as presented in Fig. 7. The experiment utilized hydraulic oil No. 46, with an oil temperature set at 313.15 K and a system overflow pressure set at 340 bar. The pressure sensors utilized in this study are the HT-PD PR140 model manufactured by HYDROTECHNIK. The high-speed camera utilized in this study is the VEO 410 model manufactured by PHANTOM. A pressure sensor with a range of 600 bar was installed on the high-pressure port of the reversing valve. The data acquisition instrument was set to a sampling frequency of 10,000 Hz to measure the working pressure. The piston and cylinder surfaces were marked with tracking points. The high-speed camera lens was adjusted to face the piston horizontally. The camera’s sampling frequency was set to 10,000 fps to measure piston displacement, enabling calculation of impact power. A series of experiments were conducted to assess the baseline configuration, the optimized configuration of the traditional model, and the optimized configuration of the surrogate model. Experimental system. The front chamber pressure curves of the hydraulic impact hammer under the baseline configuration, the optimized configuration of the traditional model, and the optimized configuration of the surrogate model are presented in Fig. 8. As illustrated in Fig. 8, the working pressures for the baseline configuration, the optimized configuration of the traditional model, and the optimized configuration of the surrogate model are 303 bar, 270.34 bar, and 270.77 bar, respectively. Preliminary experimental measurements indicate that, in comparison with the baseline configuration, the working pressure of the optimized configuration of the traditional model decreased by 10.78%, while that of the optimized configuration of the surrogate model decreased by 10.64%. Front chamber pressure curves. The contact position between the piston and drill rod is defined as 0, with upward indicating positive direction. Export the coordinate data of the piston and cylinder tracking points to calculate the positional change of the piston relative to the cylinder. The piston displacement curves for the baseline configuration, the optimized configuration of the traditional model, and the optimized configuration of the surrogate model are presented in Fig. 9. The piston velocity curves for the baseline configuration, the optimized configuration of the traditional model, and the optimized configuration of the surrogate model are presented in Fig. 10. Piston displacement curves. Piston velocity curves. As illustrated in Figs. 9 and 10, the impact velocities for the baseline configuration, the optimized configuration of the traditional model, and the optimized configuration of the surrogate model are 12.65 m/s, 11.20 m/s, and 11.22 m/s, respectively. The corresponding impact frequencies are 2.36 Hz, 3.12 Hz, and 3.11 Hz. Consequently, the impact power for the baseline configuration, the optimized configuration of the traditional model, and the optimized configuration of the surrogate model are 76.36 kW, 79.58 kW, and 79.70 kW, respectively. Preliminary experimental measurements indicate that, in comparison with the baseline configuration, the impact power of the optimized configuration of the traditional model increases by 4.21%, while that of the optimized configuration of the surrogate model increases by 4.37%. The simulation and experimental comparison of working pressure and impact power is presented in Table 9. Table 9Comparison of simulation and experimental results for working pressure and impact power.Baseline ConfigurationTraditional ModelOptimized ConfigurationSurrogate ModelOptimized ConfigurationWorking pressure simulation314.00 bar278.64 bar279.46 barWorking pressure experiment303.00 bar270.34 bar270.77 barWorking pressure error3.63%3.07%3.21%Impact power simulation78.11 kW81.68 kW81.28 kWImpact power experiment76.36 kW79.58 kW79.70 kWImpact power error2.28%2.64%1.98% Comparison of simulation and experimental results for working pressure and impact power. As illustrated in Table 9, the simulation and experimental errors for both working pressure and impact power in the baseline configuration, the optimized configuration of the traditional model, and the optimized configuration of the surrogate model are all below 5%. This finding indicates that both the traditional and surrogate models exhibit a high degree of accuracy in the multi-objective optimization of the impact performance of hydraulic impact hammers. In order to reduce computational costs and enhance solution efficiency, the EBF-trained hydraulic impact hammer performance analysis surrogate model can be employed for multi-objective optimization of impact power and working pressure. Conclusion In this study, a DOE research was conducted on hydraulic impact hammers, with the objective of analyzing the effects of 18 structural and operating parameters on working pressure and impact power. The study indicates that the piston middle diameter demonstrates a substantial negative correlation with working pressure, with an observed value of 35.34%. The nitrogen chamber pre-charge pressure exhibits the most substantial positive correlation with working pressure, with an observed percentage of 26.61%. The lower inner diameter of the reversing valve demonstrates a substantial positive correlation with impact power, with a correlation coefficient of 14.26%. The upper inner diameter of the reversing valve exhibited a substantial negative correlation with impact power, with a percentage of 12.38%. The present study identified ten key parameters influencing the impact performance and stability of hydraulic impact hammers: the piston middle diameter, the nitrogen chamber precharge pressure, the piston back diameter, the accumulator precharge pressure, the piston stroke, the low-pressure oil groove width, the lower inner diameter of the reversing valve, the upper inner diameter of the reversing valve, the piston middle-lower section length, and the reversing valve stroke. This paper enhances the impact performance and stability of a hydraulic impact hammer through multi-objective optimization. Ten key parameters are selected as design variables, with constraints imposed on the maximum values of working pressure, back chamber peak pressure, and outlet pipe peak pressure. Multi-objective optimization is conducted with the goals of maximizing impact power and minimizing working pressure, yielding the Pareto frontier for and . Furthermore, the optimal configuration is determined using TOPSIS optimized with the entropy weighting approach. Compared to the baseline configuration, the optimized configuration achieves a 4.6% increase in impact power and an 11.3% reduction in working pressure, thereby enhancing the impact performance and stability of the hydraulic impact hammer. Furthermore, this paper establishes surrogate models for analyzing the performance of hydraulic impact hammers, enhancing the computational efficiency of multi-objective optimization. Multi-objective optimization is achieved through the utilization of surrogate models that are trained using four distinct methods. RSM, RBF, EBF, and Kriging. A substantial body of research has demonstrated that the EBF-trained surrogate model for hydraulic impact hammer performance analysis exhibits a high degree of accuracy in the multi-objective optimization of impact power and working pressure. This model also exhibits a notable increase in computational speed compared to traditional models. This approach offers an efficient and viable solution for multi-objective optimization calculations of hydraulic impact hammers. The errors in impact power and working pressure between the optimized configuration of the EBF-trained surrogate model and that of the traditional model are at most minimal, at 0.50% and 0.30%, respectively. The EBF-trained surrogate model optimization configuration, in comparison with the baseline configuration, increases impact power by 4.1% and reduces working pressure by 11.0%. Consequently, the impact performance and stability of the hydraulic impact hammer are enhanced. This paper verifies the accuracy of the optimized configurations for both the traditional model and the surrogate model through working pressure and impact power experiments on a hydraulic impact hammer. The experimental findings indicate that, in comparison with the baseline configuration, the working pressure of the optimized configuration of the traditional model decreased by 10.78%, while the impact power increased by 4.21%. Similarly, the work pressure of the optimized configuration for the EBF-trained surrogate model decreased by 10.64%, while the impact power increased by 4.37%. The simulation and experimental errors for both working pressure and impact power in the baseline configuration, the optimized configuration of the traditional model, and the optimized configuration of the surrogate model are all below 5%. In summary, this study provides a reliable theoretical foundation and experimental support for multi-objective optimization research on hydraulic impact hammers, offering an accurate and efficient computational tool for their multi-objective optimization design. Acknowledgements This work was supported by the National Key Research and Development Program of China (Grant No. 2021YFB3401502) and the National Natural Science Foundation of China (Grant No. 52275108). Author contributions H.C., B.S. and H.G. conceptualized the study. Z.L. and B. L. curated the data. H.C. and B.S. developed the methodology. H.C. and B.S. supervised the work. H.C., H. G. and Z.L. validated the results. H.C. and B. L. created visualizations. H.C. wrote the original draft. H.C., B.S. and H.G. reviewed and edited the manuscript. All authors reviewed the manuscript. All authors read and approved the final manuscript. Funding This work was supported by the National Key Research and Development Program of China (Grant No. 2021YFB3401502) and the National Natural Science Foundation of China (Grant No. 52275108). Data availability The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.