A study of calibrating seismic design response spectrum based on high-intensity ground motion records Authors: Qian Liang, Bai Yunpeng, Xie Quancai, Zhang Xiaonan, Li Jilong, Wang Yuhuan Published: 2026 ABSTRACT AbstractThe seismic design response spectrum is a vital parameter for determining the potential seismic load of the engineering structure. Differential evolution algorithm (DE) with a novel hybrid mutation operator is utilized to calibrate the spectral parameters in order to enhance iteration efficiency. This study calibrates the seismic design response spectrum for China based on the Chinese seismic intensity scale and compares the results with the Code for Seismic Design of Buildings (CSDB2010).The calibration spectra are based on strong ground motion records with destructive power exceeding the Chinese seismic intensity 7 and above. The characteristics of the calibration spectral parameters are analyzed and subsequently compared with the design spectra of the Code for Seismic Design of Buildings (CSDB2010). It is found that: Increasing the number of iterations can enhance the fitting goodness of DE between the calibration spectrum and the record response spectrum. The average site characteristic period (Tg) for rock and hard soil site conditions (Class I and II) is greater than the Tg specified in CSDB2010. The average Tg for intensity 10 + is greater than the average Tg for intensity 7, 8, and 9. Tg increases with the seismic intensity. The average spectra platform value (βmax) from this study gradually approaches the βmax in CSDB2010 as the intensity increases. The average attenuation index (γ) at different intensities of this study is greater than γ in CSDB2010.Keywords: Calibration of the design response spectrum, Chinese seismic intensity, Adaptive guided differential evolution algorithm, Spectral parameters, Site conditionsSubject terms: Engineering, Natural hazards, Solid Earth sciences Introduction The seismic design response spectrum serves as a fundamental parameter in seismic engineering design for determining potential seismic loads on structures. Newmark and Hall1 first proposed three parameters—peak ground acceleration (PGA), peak ground velocity (PGV) and peak ground displacement (PGD)—to smooth design seismic response spectra, forming the well-known three-parameter calibration method. This method has been widely adopted in international building codes such as EC82 and ASCE20023. Hayashi et al.4 selected numerous accelerogram records under different soil conditions in Japan and classified response spectra into three groups, representing one of the earliest studies on the influence of site conditions on response spectral characteristics. Malhotra5,6 introduced the concept of the central period of ground motion as a critical indicator for smoothing design response spectra, enabling smoothed spectra to exhibit reasonable asymptotic characteristics in both short- and long-period ranges. Chopra and Choudhury7 investigated the influence of local geological conditions on acceleration response spectra using observational data from 32 sites in Gujarat, India. Bommer et al.8 formulated seismic design response spectra for nuclear power plant engineering in the United Kingdom. Calvi et al.9 established new design spectra based on Italian ground motion records collected from 1972 to 2017, and further revealed the combined effects of local soil conditions, earthquake magnitude and source-to-site distance on empirical design response spectra10,11,41,42. The differential evolution (DE) algorithm is a heuristic optimization approach designed to search for global optimal solutions. It performs random searching within a given parameter range and is suitable for the direct fitting of multi-parameter functional models12,13. This algorithm is well-suited for fitting problems where the functional form is explicitly known but the characteristic parameters remain undetermined. Shi et al.14 verified that the DE algorithm outperforms the simulated annealing algorithm in the calibration of seismic design spectra. For the purpose of analysis and comparison, this study transforms the response spectra into standardized amplification factor spectra, which is the ratio of response spectra values to peak ground acceleration. The transformed design spectra are as follows15: where β(T) represents the amplification factor spectral value, T is the spectral period, T0 is the first corner period, βmax is the spectra platform value, Tg is the site characteristic period, and γ is the attenuation index. Figure 1 shows the shape of β(T). The Code for Seismic Design of Buildings (CSDB2010) only divide site characteristic periods into different fixed intervals based on site categories, without considering the potential variation of spectral parameters under different seismic intensity levels, which may lead to deviations in seismic load calculation for structures in high-intensity seismic regions. Therefore, this paper collects a large number of actual strong-motion observation records from regions that experienced seismic intensity exceeding 7 in China, and adopts the adaptive guided differential evolution algorithm to calibrate the four parameters of the standardized amplification factor design spectrum for different site conditions and seismic intensity levels. The statistical distribution characteristics of the calibrated parameters are analyzed, and the results are compared. The curved shape of the normalized response spectrum. In the standard differential evolution (DE) algorithm, the two parameters F (scaling factor) and CR (crossover rate) are fixed values, which reduces the optimization efficiency of the algorithm. To address this limitation, many scholars have developed improved DE strategies with adaptive parameter adjustment to enhance algorithm performance. Liu et al.16 proposed the fuzzy adaptive DE (FADE) algorithm, which uses the adaptive adjustment mechanism of a fuzzy logic controller to control parameters. Santana-Quintero and Coello17 proposed the ε-MyDE algorithm, which was the first to incorporate the concept of ε-dominance to achieve a uniform distribution of retained solutions. On this basis, a series of improved DE variants have been proposed18; these include Trigonometric Differential Evolution (TDE)19, Opposition Based Differential Evolution (OBDE)20, Self-adapting Parameter Setting in Differential Evolution21, Self-adaptive Hybrid Differential Evolution (SaDE)22, and others23–30. In the response spectra calibration process of this study, the root mean square error is used as the objective function of DE, so that the calibrated response spectra can better fit the recorded spectra. However, the shape of response spectra is related to site conditions and ground motion characteristics. This study still uses the root mean square error as the objective function of DE during response spectra calibration, which allows the calibrated response spectra to better fit the recorded spectra. In this study, high-quality strong-motion records from near-fault stations are selected to calculate the Chinese seismic intensity. Records with high Chinese seismic intensities (7, 8, 9, 10, and above) are selected to calibrate the record response spectra using the DE algorithm. The fitting effects are then compared with records’ response spectra. The distribution of calibration spectra parameters is analyzed. The variation trend of βmax, Tg, and γ parameters of the calibrated design spectra with respect to the site average shear wave velocity of the top 30 m (referred to as Vs30) is analyzed. This study provides a reference for the improvement of seismic design response spectra parameters. Methodology of Chinese seismic intensity In order to study the relationship between parameters of strong ground motions and instrument seismic intensity, Trifunac et al.31 calculated the correlation between the intensities of different stations with strong ground motions and PGA and PGV. United States Geological Survey (USGS) established the ShakeMap system to release disaster information, in which the calculation method of the seismic intensity of the system’s instruments mainly referred to PGA and PGV32,33. Similar to the calculation method of seismic intensity used in the ShakeMap system of USGS, the PGA and PGV are also used as the physical scale of seismic intensity in China. The main reference for instrument seismic intensity in the ShakeMap system is the modified Mercalli intensity scale. The Chinese Seismic Intensity Scale (GB/T 17742 − 2020)34 was compiled on the basis of revising the Mercalli intensity scale and combining the actual situation of earthquake damage in China. The Chinese Seismic Intensity Scale (GB/T 17742 − 2020) provides detailed instructions on the calculation method for instrumental seismic intensity. The steps are as follows: (1) Baseline correction and Digital filtering. The method for baseline correction of seismic motion records involves subtracting the arithmetic average of the pre-seismic background noise from the seismic record. In general, the duration of the background noise is selected as 10 s. Digital filtering. Acceleration and velocity records of the east-west (EW), north-south (NS), and vertical (UD) directions components all require bandpass filtering within the range of 0.1 to 10 Hz. The filter passband ripple should not exceed 0.5 dB, and the filter out-of-band attenuation should be greater than 12 dB/octave. (2) Records synthesis. The following formulas are used to calculate the three-component synthesized acceleration and velocity time histories: In which, a(ti)EW, a(ti)NS, and a(ti)UD represent acceleration values in the east-west, north-south, and vertical direction components at time ti, respectively. v(ti)EW, v(ti)NS, and v(ti)UD represent the filtered velocity values in the east-west, north-south, and vertical directions at time ti. (3) Calculate the three-component synthesized seismic parameters PGA and PGV. (4) Calculate the Chinese instrumental seismic intensity. The following formulas are used to calculate IA and IV: The instrumental seismic intensity (abbreviated as Ichina) is calculated using the formula (6), and the result can be two significant digits after the decimal point. If Ichina is less than 1.0, it should be rounded up to 1.0. If Ichina is greater than 12.0, Ichina equals 12.0. Dataset of strong-motion In this study, a total of 452 strong motion records were selected from six major earthquake events. Detailed information about the earthquakes is listed in Table 1. The records were sourced from the Pacific Earthquake Engineering Research Center (PEER), the Japanese KiK‑net and K‑net networks, and the Institute of Engineering Mechanics, China Earthquake Administration (IEM). The moment magnitude (Mw) and hypocentral depth of each event were obtained from the official websites of the Japanese F‑net network and the U.S. Geological Survey (USGS) to ensure data authenticity and reliability. The Mw of the selected events ranges from 6.7 to 9.1. The hypocentral depth ranges from 8 to 24 km. The selected events are well balanced in terms of basic information, and the number of records per event is relatively consistent, which helps reduce potential biases caused by uneven data distribution. Table 1Details of the earthquakes in the database.Event nameTimeDepth(km)Mwrecord numberNorthridge1994/1/17186.754Kobe1995/1/16106.924Chichi1999/9/20117.6104Wenchuan2008/5/12147.922Tohoku2011/3/11249.1184Turkey2023/2/68.67.764 Details of the earthquakes in the database. The following criteria were applied to select the 452 records: (1) Fault distance: records were taken from stations with a rupture distance of less than 150 km to focus on near-field strong motions. (2) Waveform completeness: only three‑component records with no significant gaps or missing data were included. (3) Signal-to-noise ratio (SNR): each record is required to have SNR > 3. (4) Shear-wave velocity information: the time-averaged shear-wave velocity in the upper 30 m (Vs30) at the station must be available, either from direct measurement or through reliable empirical estimates. (5) Chinese seismic intensity: only records from stations with Chinese seismic intensity of VII or above were retained. Before analysis, all raw acceleration records were processed as follows. Baseline correction was performed by subtracting the mean value of the first 10 s of the acceleration record from the original data to remove baseline drift. Subsequently, a non-causal fourth-order Butterworth band-pass filter was applied. The low-cut frequency was set between 0.02 and 0.1 Hz depending on the quality of each record, while the high‑cut frequency was fixed at 30 Hz. Figure 2(a) shows most of the records are above 100 gal, while a few are above 1000 gal. fault distances of most records are between 20 and 100 km, including 44 records with fault distance less than 5 km. The stations are classified according to the conversion method between site average shear wave velocity of the top 30 m (Vs30) and Chinese site classification proposed by Xie et al.35. It can be seen from Fig. 2(b) that Vs30 of most stations is between 300 and 600 m/s, Vs30 of 9 stations is above 760 m/s, and Vs30 of few stations was less than 150 m/s. Most stations belong to site classes II and III, while a few stations belong to site classes I. The selected ground motions for this study are shown in Table 1. The Mw range is 6.7–9.1 and the depth range is between 10 and 20 km. (a) Distribution of records with respect to PGA (gal) and fault distance (km); (b) Distribution of records with respect to fault distance (km) and Vs30 (m/s). Figure 3. shows the distribution of the number of records of intensity 7, intensity 8, intensity 9, intensity 10+. The records of strong ground motions selected in this study are mostly intensity 8 or 9. The number of intensity 7 records is 51. The number of intensity 8 records is 80. The number of intensity 9 records is 61. The number of intensity ≥ 10 (referred to as 10+) records is 34. Number of records at different Chinese seismic intensity levels. Differential evolution algorithm (DE) The general steps of the differential evolution algorithm are as follows: (1) Initialization; (2) Mutation operation; (3) Crossover operation; (4) Selection operation. The initial parameters are population size NP (number of candidate solutions) and max evolution generations Gmax (max number of iterations). Storn and Price13 recommended that NP usually take a value between 5D and 10D (D is the dimension of solutions). Gmperle et al.17 summarized the test results and concluded that the performance of DE algorithm depended heavily on the setting of control parameters, and the control parameters were set as follows: the number of populations NP and the ideal interval was 3D-8D. Ronkkonen et al.27 believed that the ideal interval for the population number NP was 2D~40D. As computer performance improves, more data and calculations become possible. Based on the suggestion of Ronkkonen et al.25,27 D is considered the proper population number for this study. The target vectors of gth generation (Xg), are as follows: In which, g represents the evolution generations, g = 1, 2,…, Gmax. D represents dimension of solutions. The dimension of solutions in this study is 4, which are respectively the first corner period (T0), the spectra platform value (βmax), the site characteristic period (Tg), and the attenuation index (γ). The iteration steps of the DE are as follows: (1) Initialization: This study randomly initializes initial generation (g = 0) target vectors in the search space. The initialization can be defined as: In which, Xmin and Xmax are the lower bound and upper bound in each dimension. rand (0, 1) is a random number between 0 and 1. (2) Mutation. Yi et al.36 proposed a self-adapting mutation in DE, named the novel hybrid mutation operator. The proposed algorithm can balance exploration and exploitation well. The mutation vectors(Vg) of gth generation, can be follows: A novel hybrid mutation operator is used. For each individual in the population, this study selects five random mutually different indexes (r1, r2, r3, r4, r5). A mutant vector is Generated by combining the information from the parent individuals using the mutation formula: In which, Fi, g represents scaling factor of ith individual of gth generation, f(Xi, g) is objective function value of Xi, g. For a randomly selected individual Xrl and Xr2 the probability that its function value is worse than that of the individual Xi is 0.5, and the same applies for Xr2.Because Xr1 and Xr2 are separately selected, the selection of Xr1 and Xr2 are independent random events; therefore, the possibility of f(Xi, g) > f(Xr1,g) and f(Xi, g) > f(Xr1,g) is 0.25. In practical applications, if F is set as a constant and F is too large, the convergence speed of the algorithm will be slowed down and the accuracy of the global optimal solution obtained will be reduced. If F (scaling factor) is too small, the diversity of the population will be reduced and prematurity will occur. Therefore, parameter F can be set to a value that changes with the number of iterations. With the increase of the number of iterations, F decreases, which can preserve excellent population information and avoid destroying the optimal solution. Self-adaptive parameter control strategy is used in F.According to different objective functions, Liu et al. proposed different methods to definite the value of F. The formula is as follows: In which, Fl represents the lower scaling factor, Fl = 0.1. Fu represents the upper scaling factor, Fl=0.9. (3) Crossover: After mutation, a crossover operator is applied to enhance the diversity of the population. The gth generation mutation vectors (Ug), can be follows: A binomial crossover is applied to combine the mutant vector with the target vector to create a trial vector. mutation vectors(Ug) are obtained using the Equation: In which, CR is crossover factor. The value of CR in the standard DE algorithm is fixed, typically taken as a constant between 0.5 and 1.0. A higher CR value improves the diversity ability of the algorithm, meanwhile reduces the ability of sensitive neighborhood search, and reduces the efficiency of search. A low CR value can facilitate sensitive neighborhood search, but it may reduce the diversity ability of the system and lead to the local optimal situation. This study underscores the importance of effectively adjusting the value of CR. The CR is dynamically adjusted during the search process as follows: (4) Selection: Evaluate the objective function for the trial vector. If the trial vector is better than the current individual, replace the current individual with the trial vector; otherwise, keep the current individual. The essence of response spectra calibration is to fit the design spectra according to ground motion spectra characteristics. Optimization problems should select appropriate optimization objectives to find the optimal parameters. According to the characteristics of response spectra calibration, the objective function for each candidate solution in the population is as follows: In which f is the root-mean-square error function, Tm represents the maximum spectral period.β (T) donates the records normalized response spectra value of a given spectral period T, and β (T) represents the calibrating spectra value of the same spectral period T. Figure 4 presents the comparison between Chinese instrumental seismic intensity (Ichina) and USGS instrumental seismic intensity (IMM) of the stations selected in this study33. It can be observed that Ichina and IMM present a linear relationship. However, the Ichina is significantly higher than IMM, with Ichina approximately one degree larger than IMM. A comparison of Chinese instrumental seismic intensity (Ichina) and USGS instrumental seismic intensity(IMM). Figure 5 illustrates the trend of the average objective function for all records and average values as the number of iterations increases. As depicted in Fig. 5, f initially decreases and then stabilizes with the increase in iterations. As the number of iterations increases, f for most records gradually decreases, falling within the ranges of e-1 to e-2. The first 10 iterations show a more obvious decline. When the number of iterations is between 30 and 50, f is stable in the range of e-0 to e-2, with a minimum value lower than e-2 and a maximum higher than e-1. The goodness of fit of the calibration design spectra can be improved by appropriately increasing the number of iterations. In this study, stable objective function values can be obtained by performing 50 iterations. Trends of the objective function with respect to iterations. It is worth noting that the standard differential evolution (DE) algorithm suffers from premature convergence or stagnation problems, often falling into local optima. Its performance is also highly sensitive to the initial parameter settings, which are largely determined by human factors. To overcome these limitations, this study employs a hybrid mutation differential evolution algorithm with self‑adaptive control parameters for response spectrum calibration. This approach avoids the influence of manually set mutation and crossover parameters on the fitting results, thereby generating stable calculations and achieving fast convergence. Figure 6 compares the proposed algorithm with the standard DE algorithm in terms of the objective function (RMSE) over evolution generations. As can be seen from Fig. 6 and 7 the RMSE of the proposed algorithm decreases rapidly from generation 1 to 5 and remains stable after generation 50. In contrast, the RMSE of the standard DE algorithm declines rapidly in the first 20 generations, but then fluctuates and remains consistently larger than that of the proposed algorithm. These results demonstrate that the proposed algorithm outperforms the standard DE algorithm in both convergence speed and solution stability. Convergence comparison of the proposed algorithm and the standard DE algorithm. Calibration response spectrum of different intensities. Distribution of calibration parameters It should provide a concise and precise description of the experimental results, their interpretation, as well as the experimental conclusions that can be drawn. In order to in-depth investigate the relationship between calibration spectra parameters and the intensity, this study categorizes the calibration spectra parameters according to different intensities. Ground motion records of intensities ranging from 7 to 10 + were selected for analysis, and the trends of high intensity parameters are analyzed as follows. Figure 8 shows the distribution of calibration spectra parameters for different intensities. The mean value and standard deviation of Tg for records with an intensity of 7 are 0.56 and 0.28, respectively, with Tg for most records falling between 0.2 and 1.2. The mean value and standard deviation of Tg for records with an intensity of 8 are similar to those of intensity 7. The mean value of Tg for records with an intensity of 9 is 0.71, which is significantly higher than that of intensity 7 and 8, while the standard deviation, which is 0.26, is similar. The mean value of Tg for records with an intensity of 10 + is approximately 0.81, which exceeds the mean value of Tg for intensities 7, 8, and 9, indicating that Tg significantly increases under the condition of high-intensity ground motions. Distribution of calibration spectra parameters of different intensities. The mean value of βmax for records with an intensity of 7 is 2.54, the standard deviation is 0.35, and most records are in the range of 1.6–3.2. The mean value of βmax for records with an intensity of 8 is 2.51, the standard deviation is 0.42, and the βmax of most records is in the range of 1.6–3.2. The mean value of βmax for records with an intensity of 9 is 2.31, the standard deviation is 0.35, and βmax for most records is in the range of 1.6–3.2. The mean value of βmax for records with an intensity of 10 + is 2.33, the standard deviation is 0.39, and most records are in the range of 1.6-3.0. The mean value of γ for records with an intensity of 7 is 1.39, and the standard deviation is 0.37, with most values in the range 0.8-2.0. The mean value of γ for records with an intensity of 8 is 1.43, the standard deviation is 0.48, and most values are in the range of 0.8-2.0. The mean value of γ for records with an intensity of 10 + is 1.34, the standard deviation is 0.62, and most values are in the range of 0.8-2.0. In general, the distributions of βmax for records with an intensity 7 and 8 are similar, as are the distributions of βmax for records with an intensity 9 and 10+. The βmax for intensity 7 is the largest among all four intensity groups, while the βmax for intensity 9 is the smallest. The average βmax for intensities 7 and 8 are greater than those for intensities 9 and 10+. The standard deviation of γ for intensities 9 and 10 + is greater than that for intensities 7 and 8. The distribution of γ is discrete. This indicates that the difference in the spectra values of the medium-long period (1–5 s) response spectra of different records increases significantly in strong ground motion. Influence of Vs30 on calibration spectra parameters Site conditions have an important impact on the design spectral parameters. According to the Code for Seismic Design of Buildings (GB50011-2010, hereafter CSDB2010), the characteristic period of the response spectra is in the range of 0.2–0.9 s according to different site classifications. This study studied the variation of the calibration spectra parameters with respect to Vs30 and compared them with the response spectra parameters of the code (CSDB2010). It provides a reference for design spectra revision. Figure 9 shows the Tg variation according to records with different intensities calibrated by this study. This study finds the trend of Tg with respect to Vs30 and compares the Tg for records with the Tg of CSDB2010. The Tg of the CSDB2010 is related to the site’s condition classification and the seismic distance group. The majority of Tg for records in this study are significantly higher than the Tg of CSDB2010, while a few are near to or less than the Tg of CSDB2010. Correspondingly, the average Tg for records with intensity 7 is close to Tg of CSDB2010. The average Tg for records of intensity 8 is a little higher than the Tg of CSDB2010. Average Tg for records with intensity 9 and 10 + is obviously greater than the Tg of CSDB2010. For intensity 7, there are a few records that belong to the class III site. However, the average Tg of records of the III class site is similar to the Tg of CSDB2010. For the class II site, Tg for most records of intensity 7 is larger than the Tg of CSDB2010, and the average Tg is significantly higher than Tg of CSDB2010. For class I sites, the average Tg for records of intensity 7 is slightly higher than Tg of CSDB2010, and Tg for most records with intensity 7 are similar to Tg of CSDB2010. Tg for most records with intensity 7 is generally greater than the Tg of CSDB2010. while the Tg of a few records is similar to the Tg of CSDB2010. For different intensities, Tg of Class II and III sites are slightly larger than the Tg of CSDB2010, and Tg of Class I sites are obviously larger than Tg of CSDB2010. The Tg of CSDB2010 in Class I site and Class II site with different intensities site are small on the whole. Therefore, the influence of intensity on Tg should be considered, and the value of Tg should be increased in high intensity. Distribution of site characteristic period (Tg) of different intensities. The red lines are the site characteristic period (Tg) of CSDB2010. Figure 10 shows the βmax variation according to records with different intensities calibrated by this study. The average βmax of with different intensities is shown in the figure and compared with the βmax of CSDB2010. It can be seen from the figure that βmaxfor most records are between 1.5 and 3, and βmax for records with different intensities are about 2.25. When in intensity 7, the average βmax of records calibration spectra in different sites was similar, both greater than the βmax of CSDB2010, and the average βmax of class III sites was the largest. When Vs30 is less than 260 m/s, the average βmax for records with intensity 8 is less than 2.0, and the average βmax is more than the βmax of CSDB2010 when Vs30 is more than 260 m/s. For records with intensity 9, the average value of βmax in class III and II sites is close to βmax of CSDB2010, which is similar to βmax of CSDB2010, while the average βmax in class I sites is significantly greater than βmax of CSDB2010. For records with intensity 10+, the average value of βmax in class III sites is significantly greater than 2.3, while that in class II and I sites is close to the βmax of CSDB2010. On the whole, with the intensity increasing, the average βmax in different class sites is gradually different from βmax of CSDB2010; records of class II site are the closest, and records of class I and III are significantly different from βmax of CSDB2010. Distribution of the spectra platform value (βmax) of different intensities. The red lines are the spectra platform value (βmax) of the code (CSDB2010). βmax (CSDB2010) = 2.25. Figure 11 shows the distribution of γ with respect to Vs30. The average γ for records with different intensities is shown in the figure and compared with the γ of CSDB2010. It can be seen from the figure that most of the calibration records spectra γ are between 0.7 and 2.5. For records of 7 intensity, most of the γ are greater than the γ of CSDB2010, and the average γ of different sites is 1.2. For records of intensity 8, the average γ in Class III sites is the closest to γ of CSDB2010, while that in Class II and I sites is significantly higher than 0.9. When the intensity is 9, the distribution of γ is relatively discrete, and the γ of some records is less than 0.9, while the average value of γ in different sites is close to 0.9. When the intensity is 10+, the average γ of all sites is greater than the γ of CSDB2010. On the whole, the average value of γ with different intensities in this study is greater than γ of CSDB2010. Distribution of the attenuation index (γ) of different intensities. The red lines are the attenuation index (γ) of code (CSDB2010). γ (CSDB2010) = 0.9. Compare with the ground motion prediction equation (GMPE) The ground motion prediction equation is the basis of seismic ground motion parameter zonation, including source parameter, path parameter, and site parameter, which can predict ground motion parameters for potential earthquakes in a region. GMPE is of great significance for the seismic design of a region. In this study, the calibrated response spectra are compared with those of GMPE37–40. Two typical records were selected to be compared with the calibrated response spectra in this study and two GMPEs. In the perspective of the record of the Chi-Chi earthquake, as Fig. 12 (a) showing, In the spectral period range of 0.01–0.1 s, the value from this study is higher than those of the response spectra, while the two GMPEs are close to the record response spectra. In the spectral period range of 0.1–0.5 s, the value of the response spectra of this study is higher than the peak value of the response spectra, and the response spectra of Abrahamson et al.37. (hereafter ASK14) and Zhao et al.38–40(hereafter Zhao2016) are lower than the peak value of the response spectra. All the spectra fit well with the record response spectra. The values from ASK14 and Zhao2016 are significantly higher than the record response spectra in the medium and short period, and are closer to the record response spectra in the long period. However, values of ASK14 and Zhao2016 are slightly lower than the record response spectra in the medium and long spectra period. Another comparison for the typical record of the Kobe earthquake is presented in Fig. 12 (b). In the spectra period of 0–0.1 s, both the studied and GMPEs values are higher than the actual response spectra. The value of Zhao2016 is the largest in the spectral period of 0.1–0.6 s, and the value ofASK14 is the smallest in this study. When the spectra period exceeds 1.0s, the values of this study are similar to those of record spectra, while the values of ASK14 and Zhao2016 are smaller than those of record response spectra. Compared with GMPE, the response spectra calibrated in this study is more consistent with the trend of the record spectra. The characteristic periods (Tg) calculated in this study are generally higher than the standard values specified in current seismic design codes, and this discrepancy is significantly correlated with the dataset filtering criteria adopted in this work. To focus on the attenuation characteristics of high-intensity ground motions, this study only includes ground motion records from earthquakes with magnitude Mw > 6.5, and does not cover small-magnitude earthquake events that may also produce high-intensity ground motions in their near-field regions. Comparison of calibrated response spectra with GMPEs: (a) Chi-Chi earthquake record (Mw 7.6, distance 15 km, Vs30 471 m/s, intensity 8); (b) Kobe earthquake record (Mw 6.9, distance 11 km, Vs30 256 m/s, intensity 9). Mechanistic explanation for this discrepancy can be derived from existing GMPE studies: under the same observation distance, the source rupture width of large earthquakes is significantly larger than that of small earthquakes. A larger source rupture scale can excite more abundant long-period ground motion components, which consequently prolongs the characteristic period of ground motions. This mechanism indicates that direct application of Tg values calculated based on moderate and strong earthquake records to seismic design of engineering structures in the near-field area of small earthquakes may cause a certain degree of overestimation of the actual characteristic period. The Tg value conclusions of this study are more applicable to seismic design checking of engineering structures in areas affected by moderate and strong earthquakes. Future studies can further incorporate high-intensity near-field records from small-magnitude earthquakes to improve the characteristic period value system for different magnitude intervals, and provide more accurate parameter basis for seismic design of engineering structures under different magnitudes and distance conditions. Conclusions In this study, the Differential Evolution (DE) algorithm is used to calibrate the near-fault records of strong ground motions. Three objective functions (root-mean-square error function) are selected and compared with the objective function of the Differential Evolution (DE) algorithm, and the relationship between calibration response spectrum parameters and site Vs30 was analyzed. Conclusions were drawn as follows: The calibrated design spectrum in this study is based on ground motion records. Compared with GMPE, the calibrated response spectrum in this study is more consistent with the trend of the record response spectrum. The calibrated design spectrum in this study is based on ground motion records. Compared with GMPE, the calibrated response spectrum in this study is more consistent with the trend of the record response spectrum. Average Tg values of Class I and Class II sites are greater than the Tg of the code (CSDB2010). This study suggests that the influence of seismic intensity on Tg should be considered, and Tg should be increased for 8, 9, and 10 + intensities. The average Tg of 10 + intensity is greater than the average Tg of 7, 8, and 9 intensities, indicating that the site characteristic period will increase with the increase of intensity. Average Tg values of Class I and Class II sites are greater than the Tg of the code (CSDB2010). This study suggests that the influence of seismic intensity on Tg should be considered, and Tg should be increased for 8, 9, and 10 + intensities. The average Tg of 10 + intensity is greater than the average Tg of 7, 8, and 9 intensities, indicating that the site characteristic period will increase with the increase of intensity. Most βmax values of records range from 1.5 to 3, and βmax of different intensities is similar to 2.25, among which class II is the closest, class I and class III sites are significantly different from the βmax of CSDB2010. when the intensity is 7 degrees, the calibrated βmax value of the record is significantly higher than the βmax of CSDB2010. The average βmax of 7 intensity records and 8 intensity records is close to each other. The average βmax of 9 intensity records and 10 + intensity records is close to each other. The average βmax of 7 intensity records is the largest, and the average βmax of 7 intensity records is the smallest. Most βmax values of records range from 1.5 to 3, and βmax of different intensities is similar to 2.25, among which class II is the closest, class I and class III sites are significantly different from the βmax of CSDB2010. when the intensity is 7 degrees, the calibrated βmax value of the record is significantly higher than the βmax of CSDB2010. The average βmax of 7 intensity records and 8 intensity records is close to each other. The average βmax of 9 intensity records and 10 + intensity records is close to each other. The average βmax of 7 intensity records is the largest, and the average βmax of 7 intensity records is the smallest. Most of the calibration records spectra γ are between 0.7 and 2.5. The average γ of 7, 8 intensity records are greater than that of intensity 9 and intensity 10, and the standard deviation of γ of intensity 9 and intensity 10 is greater than that of intensity 7 and intensity 8, indicating that γ is more discrete in the case of high intensity. Most of the calibration records spectra γ are between 0.7 and 2.5. The average γ of 7, 8 intensity records are greater than that of intensity 9 and intensity 10, and the standard deviation of γ of intensity 9 and intensity 10 is greater than that of intensity 7 and intensity 8, indicating that γ is more discrete in the case of high intensity. Author contributions Qian. L. and Zhang X. finished the conceptualization, methodology and wrote the initial draft. Bai Y., Wang Y., and Li J. provided important suggestions for interpreting the results; Xie Q. determined the research goals. All authors have read and agreed to the published version of the manuscript. Funding This work is supported by Scientific Research Fund of Institute of Engineering Mechanics, China Earthquake Administration (Grant No. 2024B19). Data availability We use data from the Center for Engineering Strong Motion Data (CESMD), providing the Engineering Strong Motion Data from the network publicly available at https: //www.strongmotioncenter.org. We use data from the Pacific Earthquake Engineering Research Center (PEER), publicly available at https://ngawest2.berkeley.edu/site. Data from the Disaster and Emergency Presidency of Turkey (AFAD) for providing the Engineering Strong Motion Data from the network is publicly available at https://tadas.afad.gov.tr/login. Data from the National Research Institute for Earth Science and Disaster Resilience (NIED) is publicly available at [https://www.kyoshin.bosai.go.jp](https:/www.kyoshin.bosai.go.jp) . 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 We use data from the Center for Engineering Strong Motion Data (CESMD), providing the Engineering Strong Motion Data from the network publicly available at https: //www.strongmotioncenter.org. We use data from the Pacific Earthquake Engineering Research Center (PEER), publicly available at https://ngawest2.berkeley.edu/site. Data from the Disaster and Emergency Presidency of Turkey (AFAD) for providing the Engineering Strong Motion Data from the network is publicly available at https://tadas.afad.gov.tr/login. Data from the National Research Institute for Earth Science and Disaster Resilience (NIED) is publicly available at [https://www.kyoshin.bosai.go.jp](https:/www.kyoshin.bosai.go.jp) .