Skip to main content
Heritage Science
Download PDF
Download ePub

Nondestructive characterization and artificial intelligence recognition of acoustic identifiers of ancient ceramics

Heritage Science volume 11, Article number: 144 (2023) Cite this article

Abstract

Cultural heritage identity management is the most basic and important work in the process of cultural heritage protection. It is of great significance to provide a unique and identifiable digital identity for ancient ceramics. At present, the identification information of ancient ceramics is mainly composed of external visual characteristics, and there is no report on feature identification method that can reflect the properties of ancient ceramics. Audible sound signals not only have advantages in non-destructive testing, but also can be used as voiceprint information to identify, monitor and analyze ancient ceramics. In this paper, seven ancient ceramics and 12 similar modern ceramic cups are taken as research objects, and an acoustic identifier (AID) is constructed. We put forward a reliable acoustic identification method for ancient ceramics, and established a digital code of acoustic characteristics of ancient ceramics. The results show that audible sound waves can reflect the attribute information of ancient ceramics. Sufficient acoustic data combined with deep learning can not only accurately match the identity of ancient ceramics, but also detect the real-time identity information of ancient ceramics, and make a comparative analysis of its cracks and whether it has caused damage. This method can provide a variety of practical applications for audible signal feature recognition technology in the exhibition, protection, trading, recognition and safety management of ancient ceramics and other cultural relics.

Introduction

The identification and management of cultural relics can be more efficient and accurate through the unique identification code of cultural relics [1, 2]. At present, the identification of cultural relics is mainly in the form of their name, category, number and descriptive information. The identification information of cultural relics mainly consists of external visual features such as 2D images and 3D scanning data [3, 4]. For imitations of the same size and extremely similar appearance, the characteristic information of the above-mentioned data is limited and cannot be the only indicator to identify the identity characteristics of cultural relics. Cultural relics should be endowed with unique digital identifiers and identification information. Therefore, it is necessary to conduct more comprehensive and in-depth research on the calibration, recording and nondestructive identification of cultural relics [5].

Sound signal is rich in information, which not only has unique advantages in nondestructive testing, but also can be used as voiceprint information for identification, monitoring and analysis [6, 7]. The audible sound has been widely used in nondestructive testing, such as wood, mechanical parts, steel bars, food quality testing and so on [8,9,10]. The speed of sound propagation will also change with the structure and density of the object. The properties or defects of objects can be analyzed according to the sound characteristics of solid sound transmission [11]. In addition, the acoustic behavior of an object can be predicted preliminary according to its natural frequency, which is related to its composition, structure, mass, and size [12,13,14].

The raw materials of ancient ceramics are basically from the place of origin, and the firing methods of different kilns have their own characteristics [15, 16]. In the case of ancient ceramics, it is necessary to identify the information reflecting its own attributes. In this study, the acoustic characteristics of ancient ceramics were analyzed to establish stable and reliable acoustic identifiers (AID), which is the vibration spectrum information of a solid after sound transmission through ancient ceramics. The AID of ancient ceramics reflects the inherent characteristics of their material and structural state.

Deep learning is a core technology in the development of artificial intelligence, which has achieved optimal and even surpassed human results in research fields of voice classification, voice recognition and voice processing [17]. Deep learning is a branch of machine learning. Deep learning can automatically learn features, while traditional machine learning requires manual feature calibration to extract features [18]. Deep learning can automatically learn complex data features through multi-layer mapping of neural network, which reduces the workload of feature engineering and relies less on original data [19]. In this paper, a nondestructive detection and identification system for ancient ceramics is designed. Researchers can conduct nondestructive testing on ancient ceramics and record their AID at that time for further study. At the same time, the measured acoustic waves of ancient ceramics are recorded into the database as their AID, and the artificial intelligence platform is used to establish sound classification models for the deep learning of these acoustic data. Through these recorded acoustic data, the integrity of the cultural relics in the future can be compared, analyzed, and identified as to whether they are newly damaged, repaired or replaced. Therefore, the identity and status of cultural relics can be accurately identified and further traced. It is of great significance to the collection, management, deterioration analysis, authenticity identification and analytic cognition of cultural relics.

Methods

The purposes of this study are as follows: (1) design acoustic measurement methods and devices suitable for ancient ceramics; (2) construct AID that can represent the identity and state of ancient ceramics; (3) customize the classification model and result characterization method of ancient ceramic AID recognized by artificial intelligence.

Samples

As shown in Fig. 1a, seven ancient ceramics of different types and sizes are taken as research objects, including two blue and white porcelain plates, two blue and white porcelain bowls, two celadon jars and one celadon bowl. The samples were provided by the key laboratory of materials and technology for underground cultural relics protection, Ministry of education, Shaanxi University of Science and Technology, and their basic information was shown in Table 1.

Fig. 1
figure 1

Ceramics samples; a seven ancient ceramics samples; b 12 Chinese zodiac modern ceramic cup samples

Full size image
Table 1 Basic information of ancient ceramic samples
Full size table

Figure 1b shows 12 ceramic cups produced in the same batch, which use the same raw materials, forming process and firing system. The only difference between them is that their cups are printed with different designs of the Chinese zodiac, including rat, ox, tiger, rabbit, dragon, snake, horse, goat, monkey, rooster, dog, and pig.

Experimental method

The acoustic analysis system of ancient ceramics based on artificial intelligence technology is composed of two parts: acoustic parameter acquisition device and artificial intelligence analysis platform (Fig. 2). The first part is the acoustic parameter acquisition device, including audio power amplifier (AV-699BT, China), vibration pickup device (Korg, CM-300, Japan), dual microphone preamplifier (M-Audio, Audio buddy, USA) and computer. The audio vibrator is a kind of piezoelectric microphones, which can play audible sounds ranging from 20 Hz to 20 kHz. According to the shape and size of the ancient ceramics, choose appropriate pickups, such as patch pickups, pickups clip, piezoelectric sensing film, etc. The pickup device selected in this study is a pickup clip. The pickups can only receive the vibration transmitted by ancient ceramic structures, not air-borne vibration. When the ancient ceramic vibrates, the magnetic induction lines will be cut in the magnetic field inside the pickup, and the coil around the magnetic core will generate induced current, so the pickup can convert the vibration signals collected from the ancient ceramics into electrical signals without interference from other vibration waves from the air medium.

Fig. 2
figure 2

Acoustic parameter measuring device and the process diagram of the vibration classification model for ancient ceramics

Full size image

After measuring the acoustic signals of the ancient ceramics, the acquired acoustic signals are converted into digital signals using a dual microphone preamplifier, and the data is recorded and analyzed using computer software. The sound is then deeply learned and recognized using Easy DL, an open-source platform for artificial intelligence. Easy DL supports customized model training and supports a full range of functions in model development [20]. Easy DL platform is built on the PaddlePaddle framework, which consists of multi-machine parallel architecture, multi-GPU parallel architecture, sequence model and large-scale sparse training, which is suitable for this study. PaddleBook and the available online validation model set PaddleModels also add some knowledge of deep learning, which makes PaddlePaddle the reason for choosing the deep learning framework [21,22,23]. The basic flow of the acoustic classification model is shown in Fig. 2. Once a predetermined amount of data has been measured, a customized acoustic classification model for ancient ceramics can be obtained through the functions of data set management, model training, model evaluation, model verification and model publishing.

Results and discussion

Acoustic signals of ceramic samples

The acoustic parameters of seven ancient ceramic samples were measured with a sinusoidal variable frequency generator of 20 Hz to 20 kHz. As shown in Fig. 3, it is the audio signal result of generator, and results of the signal passing through seven samples respectively. The duration of this sine wave sweep signal is 10 s. The sampling frequency is 44,100 Hz, and the frequency resolution is 0.1 Hz. Within the range of 2 s to 6 s, the pickup receives strong vibration transmitted by ancient ceramic structures. When the amplitude of the measured vibration transmitted by ancient ceramic structures was greater than 0.5 dB, there will be a slight displacement in the experiment of the ancient ceramic sample, which is the resonance phenomenon. In the resonance process, the sample will produce violent vibration and knock the generator. At this time, the level signal obtained by the pickup clip exceeds 100%, and the measured time-domain signal will be saturated. The saturation of the signal in Fig. 3 means these ancient ceramic samples were resonated. As shown in Fig. 3, it has been found that ancient ceramic samples may have resonance between about 200 Hz and 900 Hz. Resonance is the phenomenon of a surge of energy in the whole system when the external excitation frequency of the system is equal to some specific value, and these specific external excitation frequencies are the natural frequencies of the system [24, 25]. As shown in Fig. 3, it was found that the resonance did not occur at a single frequency, but in the resonance band from about 200 Hz to 900 Hz.

Fig. 3
figure 3

The time domain curves of the generator and the vibrations transmitted by ancient ceramic structures at 20 Hz to 20 kHz

Full size image

According to our previous research, the mass and elastic modulus of ancient ceramics are the two main factors that affect the resonant frequency of ancient ceramics. When the material properties (density, Poisson’s ratio, etc.) of ancient ceramics are different or the shape is different, the affected parameters are still mass and elastic modulus [26, 27]. When the boundary conditions of a structure are different, the natural frequencies must inevitably be different, because the boundary conditions will affect the distribution of elastic modulus of the structure. Meanwhile, the elastic modulus of ceramics is related to the phase type, particle size, distribution, proportion, and porosity of ceramics [28, 29].

Deep learning and recognition of ancient ceramic AID

For the detection and identification information collection of precious cultural relics such as ancient ceramics, the most basic premise is to ensure the safety of cultural relics. Ancient ceramics are brittle and may be damaged or potentially damaged if the test signal causes them to resonate. Therefore, in the subsequent detection, we chose to avoid the resonance frequency band of the samples, and chose the non-resonance frequency 1000 Hz, which is relatively close to the resonance frequency band, as the generator frequency. This sound wave can reflect the state parameters of the sample, such as small differences due to cracks, pores and attachments in the sample [30]. This is to ensure the stability of the test data, and it is more important to avoid any possible damage to ancient ceramics caused by resonance. The sampling frequency is 44,100 Hz. Each sample was tested 50 times, each test lasted for 4 s, and a single test collected about 176,400 data sampling points. These acoustic data points make up the AID of ancient ceramics.

As shown in Fig. 4a, it is the 1 kHz audio generator and the time-domain signals of seven samples. The intensity of the acoustic amplitude of each sample is different. Each sample has different sound loss. The mass of the sample and the propagation path of the signal are the main factors affecting the sound loss. Mass is the product of density and volume, the greater the mass of the sample, the more the diffusion attenuation of sound waves. The longer and more complex the propagation path of sound wave in ceramic, the greater the heat loss of sound wave in ceramic. In addition, cracks and materials in ancient ceramics will also affect the measured path of sound wave propagation. It is these factors that constitute the unique acoustic identity of each ancient ceramic.

Fig. 4
figure 4

Acoustic parameters characterization of ancient ceramic samples at 1 kHz; a time domain curves of generator and ancient ceramic samples; b frequency domain curves of generator and ancient ceramic samples

Full size image

In addition, the frequency domain characteristics of samples can be seen after Fourier transform of the time domain signal, as shown in Fig. 4b. The sampling frequency is 44,100 Hz, and the frequency resolution is 0.1 Hz. In order to make the characteristics of the curve in the frequency domain more obvious, we smoothed the curve with a smoothing index of 1/96. It can be found that the main frequency 1 kHz amplitude of each sample is different. In addition, there are a lot of harmonics in the frequency domain of each sample, and the distribution and amplitude of these harmonics are also different. These characteristics constitute the characteristic information of each ancient ceramic AID.

We uploaded 350 audio data into the ancient ceramic sound classification model created on Easy DL, label the data, and then train the model. In the model training, the system carries out deep learning on the 350 audio data. After deep learning, 105 audio data was randomly predicted by the random test set of the model. The number of correctly predicted performance was 103, the number of incorrectly predicted performance was two, and the accuracy rate of the model was 98.1%. Therefore, we increased the number of test data for a single sample from 50 to 100 times, and the total number of data reached 700. The model randomly predicted 210 audio data again, and the accuracy of the model increased to 100%. When the amount of data in the model is larger and the data features are more stable, the recognition accuracy of the model will be higher.

Then, we measure these seven samples twice again, and put the measured data into the trained model for verification. The verification results were shown in Table 2. The recognition accuracy rate of all samples is 100%, and the recognition matching degree is 96.24% to 100%. In order to explain the reasons for the fluctuation of recognition results, we take the test results of sample BW-bowl 2 as an example for analysis. As shown in Fig. 5a, the BW-bowl 2 sample has some cracks near the measurement point, and the position of the pickup clip is held along the direction of the cracks. In the data acquisition process of Result 1, the position of the pickup clip is the same position and direction. However, when data was collected for Result 2, the position and direction of the pickup clip was changed, and it crossed the crack.

Table 2 Audio recognition results for seven ancient ceramic samples
Full size table
Fig. 5
figure 5

Analysis of influencing factors of acoustic parameters of sample BW-bowl 1; a photos of measuring position of BW-bowl 1; b frequency domain curves of BW-bowl 1 twice test data; c time-domain curves of BW-bowl 1 twice test data

Full size image

As shown in Fig. 5b and c, when the measurement direction changes, the contact position between them will deviate, and the signals picked up in the time domain and frequency domain will be significantly different. It can be clearly seen from the frequency domain waveform that the amplitudes of some characteristic peaks have changed, but the position of the characteristic peaks has not changed obviously. This leads to the change in the matching rate of the recognition, but it does not lead to recognition error. This can be improved by keeping the consistency of measurement points during the test. At the same time, when the amount of deep learning data increases, the recognition accuracy will also be improved in this case.

Deep learning and recognition of modern ceramic samples

We use the laser positioning sensor (ZW-LV100R-NP, China) to make bidirectional positioning measurement on the sample, to avoid the inaccurate identification accuracy due to the inaccurate location of the collection point, as shown in Fig. 6a. It can make sure the collection points are in the same position at each measurement. The reflective film at the laser positioning point was shown in Fig. 6b. The reflective film can reflect the laser light to the positioning sensor, so that the sensor can determine the position of the sample, as shown in Fig. 6c and d. We collected the information of the same test point on the sample for three times by using 1 kHz sine wave audio with a duration of 4 s, and it was necessary to reposition the sample and reposition the pickup clip each time. It was found that the acoustic data measured by this method had the same characteristics, as shown in Fig. 6e. This shows that the laser positioning method may reduce the measurement error and improve the accuracy of recognition. To further confirm the results, vibration signals were collected from 12 samples.

Fig. 6
figure 6

Ceramics measuring point positioning measuring device and schematic diagram; a schematic diagram of positioning laser and directional laser; b reflective film at the location of the directional laser point; c directional laser and reflective film at positioning point; d positioning laser and reflective film at measuring point; e frequency domain curves of the ceramic cup sample were measured three times

Full size image

Consistent with the measurement of ancient ceramic samples, 1 kHz was chosen as the generator frequency. The sampling frequency is 44,100 Hz. Each sample was only tested 20 times, each test lasted for 4 s, and a single test collected about 176,400 data sampling points. Acoustic parameters of 12 Chinese Zodiac modern ceramic cup samples at 1 kHz as shown in Fig. 7. As shown in Fig. 7a and b, even 12 ceramic cups produced from the same batch, with the same composition, manufacturing method and firing method, have different vibrations transmitted through the samples. Acoustic parameters can show the structural differences between them, showing their unique “voiceprint information.” Among them, the time domain signal of the Dog-cup is significantly higher than that of other cups. It is found that there is a crack at the bottom of the Dog-cup, which will cause the sound wave cross-section to reflect, resulting in an increase in the amplitude of the time domain frequency.

Fig. 7
figure 7

Acoustic parameters characterization of 12 Chinese zodiac modern ceramic cup samples at 1 kHz; a time domain curves; b frequency domain curves

Full size image

At the same time, we found that after laser localization of the samples, even though the modern ceramic samples are very similar, and the number of measurements is reduced to 20 times compared with the ancient ceramic samples of 50 times, the success rate of recognition and similarity rate are 100%. This shows that if the location of the measuring point is consistent, the identification information of ancient ceramics at this measuring point can be repeatedly obtained, so as to realize the identification of ceramics with high precision.

Conclusions

This study proposed an AID for ancient ceramics based on acoustic vibration spectra, and designed an acoustic analysis system for ancient ceramics, which is composed of acoustic parameter acquisition device and artificial intelligence analysis platform. The acoustic parameter acquisition device enables non-destructive measurement and stable recording of acoustic parameters of ancient ceramics of various sizes and vessel types. The AID of ancient ceramics reflects the vibration spectrum characteristics of ceramics, which is closely related to the material and structural state of ancient ceramics and records the comprehensive performance characteristics of ancient ceramics.

Through the acoustic classification model of ancient ceramics, the deep learning of AID is carried out to realize the non-destructive, accurate and fast recognition of ancient ceramic AID by artificial intelligence. When the number of data tests of a sample reaches 100, the recognition accuracy of all samples is 100% and the recognition matching degree is 96.24% to 100%. Through the laser positioning of 12 Chinese zodiac modern ceramic cups, it is found that the recognition accuracy can reach 100%. The recognition and matching degree can be further improved by increasing the amount of measurement data and the clamping accuracy of device. It is possible to establish and publish API at a later stage to complete the application of the AID database and intelligent identification program.

By deeply learning the acoustic transmission parameters of ancient ceramics and comparing the sound information of solid acoustic transmission of ceramics with PaddlePaddle platform, it can accurately identify whether ceramic samples are replaced, repaired or damaged. This is of innovative significance for the establishment of digital coding of cultural relics, and has important applications for the collection, management, deterioration analysis, authenticity identification, analysis, and cognition of cultural relics.

Availability of data and materials

The data are not publicly available due to their containing information that could compromise the privacy of the National Key Research and Development Program of China (Grant No. 2019YFC1520100 and 2019YFC1520200).

Abbreviations

2D:

Two-dimensional

3D:

Three-dimensional

AID:

Acoustic identifiers

API:

Application programming interface

References

  1. Belussi A, Migliorini SA. Spatio-temporal framework for managing archeological data. Ann Math Artif Intell. 2017;80:175–218. https://doi.org/10.1007/s10472-017-9535-0.

    Article  Google Scholar 

  2. Ognjanović Z, Marinković B, Šegan-Radonjić M, Masliković D. Cultural heritage digitization in Serbia: standards, policies, and case studies. Sustainability. 2019;11(14):3788. https://doi.org/10.3390/su11143788.

    Article  Google Scholar 

  3. Wulong X, Xijie S, Shihui P. Visual dissemination of intangible cultural heritage information based on 3D scanning and virtual reality technology. Scanning. 2022;2022:8762504. https://doi.org/10.1155/2022/8762504.

    Article  Google Scholar 

  4. Jacek K, Marek M, Jerzy M. Problems of acquisition and postprocessing of 3D scans of large architectural objects. MATEC Web Conf. 2019. https://doi.org/10.1051/matecconf/201925203001.

    Article  Google Scholar 

  5. Paskin N. Toward unique identifiers. Proc IEEE. 1999;87(7):1208–27. https://doi.org/10.1109/5.771073.

    Article  CAS  Google Scholar 

  6. Sun C, Yang Y, Wen C, Xie K, Wen F. Voiceprint identification for limited dataset using the deep migration hybrid model based on transfer learning. Sensors. 2018;18(7):2399. https://doi.org/10.3390/s18072399.

    Article  Google Scholar 

  7. Yang W, Wang X, Zhou S, Zhao H, Huang J. An improved method for voiceprint recognition. Complex, Intell, Softw Intensive Syst CISIS. 2018;772:735–46. https://doi.org/10.1007/978-3-319-93659-8_67.

    Article  Google Scholar 

  8. Lbatawi IE. An acoustic impact method to detect hollow heart of potato tubers. Biosyst Eng. 2008;100(2):206–13. https://doi.org/10.1016/j.biosystemseng.2008.02.009.

    Article  Google Scholar 

  9. García DP, Utrilla MM, Alpuente HJ, Martínez RJ. A study of the optimal waveforms for non-destructive spectral analysis of aqueous solutions by means of audible sound and optimization algorithms. Appl Sci. 2021;11(16):7301. https://doi.org/10.3390/app11167301.

    Article  CAS  Google Scholar 

  10. Rojas JAM, Alpuente J, Postigo D, Rojas IM, Vignote S. Wood species identification using stress-wave analysis in the audible range. Appl Acoust. 2011;72:934–42. https://doi.org/10.1016/j.apacoust.2011.05.016.

    Article  Google Scholar 

  11. Naaman M, Pearson M, Pullin R, Almudaihesh F, Grigg S. Evaluating the usefulness of audible acoustics as a damage detection method in large composite structures. EWSHM. 2023;270:849–61. https://doi.org/10.1007/978-3-031-07322-9_86.

    Article  Google Scholar 

  12. Stefan S, Konstantinos K, Berthold H. An auditory feature detection circuit for sound pattern recognition. Sci Adv. 2015;1(8):e1500325. https://doi.org/10.1126/sciadv.1500325.

    Article  Google Scholar 

  13. Schuyer J. Molar sound velocity of solids. Nature. 1958;181:1394–5. https://doi.org/10.1038/1811394b0.

    Article  CAS  Google Scholar 

  14. Li T, Zhang X, Wang H, et al. Sound absorption and compressive property of PU foam-filled composite sandwiches: effects of needle-punched fabric structure, porous structure, and fabric-foam interface. Polym Adv Technol. 2020;31(3):451–60. https://doi.org/10.1002/pat.4781.

    Article  CAS  Google Scholar 

  15. Jin X, Wang X, Liang Y, Wang F, Luo H. Celadon colour data association classification and its dynasty-kiln site characteristics. Ceram Int. 2021;47(21):29567–75. https://doi.org/10.1016/j.ceramint.2021.07.124.

    Article  CAS  Google Scholar 

  16. Jin X, Wang X, Liang Y. The mechanism of “flint red” and its relationship with celadon glaze color. J Eur Ceram Soc. 2022;42(7):3332–8. https://doi.org/10.1016/j.jeurceramsoc.2022.02.031.

    Article  CAS  Google Scholar 

  17. Ruihui M, Xiaoqin Z. A review of deep learning research. KSII Transa Int Informa Syst. 2019;109(5):820–38. https://doi.org/10.3837/tiis.2019.04.001.

    Article  Google Scholar 

  18. Du X, Cai Y, Wang S, Zhang L. Overview of deep learning, 2016 31st Youth Academic Annual Conference of Chinese Association of Automation (YAC), Wuhan, China, 2016. 159–164. https://doi.org/10.1109/YAC.2016.7804882.

  19. Shinde PP, Shah S. A review of machine learning and deep learning applications. In: 2018 fourth international conference on computing communication control and automation (ICCUBEA), Pune, India, 2018; 1–6. https://doi.org/10.1109/ICCUBEA.2018.8697857.

  20. Easy DL AI development platform. 2023. https://ai.baidu.com/easydl/. Accessed 14 Jan 2023.

  21. Du W, Zheng J, Li W, Liu Z, Wang H, Han X. Efficient recognition and automatic sorting technology of waste textiles based on online near infrared spectroscopy and convolutional neural network. Resour Conserv Recycl. 2022;180:106157. https://doi.org/10.1016/j.resconrec.2022.106157.

    Article  CAS  Google Scholar 

  22. Li X, Xiong H, Li X, Wu X, Chen Z, Dou D. InterpretDL: explaining deep models in PaddlePaddle. J Mach Learn Res. 2022;23(197):1–6.

    Google Scholar 

  23. Ma Y, Yu D, Wu T, Wang H. PaddlePaddle: an open-source deep learning platform from industrial practice. Front Comput Sci. 2019;1(1):105–15. https://doi.org/10.11871/jfdc.issn.2096.742X.2019.01.011.

    Article  Google Scholar 

  24. Ashcroft NW, Mermin ND. Solid state physics. Philadelphia: Saunders College Press; 1976. p. 116–37.

    Google Scholar 

  25. Matsushima K, Noguchi Y, Yamada T. Omnidirectional acoustic cloaking against airborne sound realized by a locally resonant sonic material. Sci Rep. 2022;12:16383. https://doi.org/10.1038/s41598-022-20591-z.

    Article  CAS  Google Scholar 

  26. Jin X, Wang X, Cao X, et al. Construction and recognition of acoustic ID of ancient coins based on deep learning of artificial intelligence for audio signals. Herit Sci. 2023;11(46):1–7. https://doi.org/10.1186/s40494-023-00891-x.

    Article  CAS  Google Scholar 

  27. Wareing RR, Davy JL, Pearse JR. Variations in measured sound transmission loss due to sample size and construction parameters. Appl Acoust. 2015;89:166–77. https://doi.org/10.1016/j.apacoust.2014.10.001.

    Article  Google Scholar 

  28. Flores M, Ouamara N, Remondiere L, Jouin J, Fiore G, Oriol S, Rossignol S. Synthesis and robocasting of YAG xerogel: one-step conversion of ceramics. Sci Rep. 2022;12:8454. https://doi.org/10.1038/s41598-022-12204-6.

    Article  CAS  Google Scholar 

  29. Naaman M, Pearson M, Pullin R, Almudaihesh F, Grigg S. Evaluating the usefulness of audible acoustics as a damage detection method in large composite structures. In: European Workshop on Structural Health Monitoring. EWSHM 2022. Lecture notes in civil engineering. 2023;270:849–61. https://doi.org/10.1007/978-3-031-07322-9_86.

  30. Abeele E, Carmeliet J, Cate J, Johnson P. Nonlinear Elastic Wave Spectroscopy (NEWS) techniques to discern material damage, Part II: single-mode nonlinear resonance acoustic spectroscopy. Res Nondestruct Eval. 2000;12(1):31–42. https://doi.org/10.1007/s001640000003.

    Article  Google Scholar 

Download references

Acknowledgements

Thanks to Baidu Easy DL for providing this research with an open-source platform for deep learning of artificial intelligence.

Funding

This work was supported by the National Key Research and Development Program of China (Grant No. 2019YFC1520100 and 2019YFC1520200).

Author information

Author notes

  1. Xiufeng Wang

    Present address: Key Laboratory of Materials and Technology for Underground Cultural Relics Protection, Ministry of Education, Shaanxi University of Science and Technology, Xi’an, China

  2. Xiaoxue Jin

    Present address: School of Bioresources Chemical and Materials Engineering, Shaanxi University of Science and Technology, Xi’an, China

Authors and Affiliations

  1. Key Laboratory of Materials and Technology for Underground Cultural Relics Protection, Ministry of Education, Shaanxi University of Science and Technology, Xi’an, China

    Xiaoxue Jin

  2. School of Materials Science and Engineering, Shaanxi University of Science and Technology, Xi’an, China

    Xiufeng Wang

  3. School of Bioresources Chemical and Materials Engineering, Shaanxi University of Science and Technology, Xi’an, China

    Chaohua Xue

Authors
  1. Xiaoxue Jin
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Xiufeng Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Chaohua Xue
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

XJ: Investigation; software; data curation; visualization and writing-original draft. XW: Conceptualization; methodology; supervision and writing—review. CX: Supervision and formal analysis.

Corresponding authors

Correspondence to Xiaoxue Jin or Xiufeng Wang.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Jin, X., Wang, X. & Xue, C. Nondestructive characterization and artificial intelligence recognition of acoustic identifiers of ancient ceramics. Herit Sci 11, 144 (2023). https://doi.org/10.1186/s40494-023-00990-9

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s40494-023-00990-9

Keywords