A Methodology to Manage and Correlate Results of Non-Destructive and Destructive Tests on Ancient Timber Beams: The Case of Montorio Tower

Abstract

Intending to safeguard architectural heritage, the assessment of the health of timber structures is crucial, though challenging, due to the organic nature of wood and to the uncertainties of its preservation state. To this end, useful support is provided by the ICOMOS guidelines, which provide conservation strategies based on thorough diagnosis and safety evaluations. In this context, the study summarized in this paper focuses on the medieval Tower of Montorio, which suffered considerable damage due to the strong earthquake that occurred in those area in September 2003. Its subsequent process of rehabilitation and restoration involved a widespread experimental campaign and the substitution of some timber beams. This circumstance has offered a rare opportunity to study these ancient elements in detail, although they are limited in number. Six beams made of oak and removed from an intermediate floor of the tower were evaluated through a comprehensive approach that included both non-destructive tests (NDT) and destructive tests (DT). Particularly, they were subjected to visual inspections, ultrasonic, sclerometric, and resistographic methods, and destructive four-point bending tests. Overall, the study presented here provides a useful qualitative comparison between them. Results highlighted that relying only on NDT might lead to an overestimation of mechanical properties and that combining NDT with DT is crucial for a more accurate assessment. Therefore, the need to deepen the research on correlations between NDT and DT to obtain reliable values of mechanical properties while respecting the conservation aim was confirmed.

1. Introduction

When dealing with architectural heritage, the health assessment of timber structures is a topic of crucial importance for planning proper conservation strategies, and it is a difficult task as well. Most of the problems are mainly due to the organic nature of the wood, the resultant uncertainties concerning the actual state of preservation, and the presence or absence of ongoing deterioration processes, which can affect the structural capacity towards static and dynamic loads. To overcome these difficulties, it is not superfluous to remark that a necessary step is to adopt an approach that meets the basic principles of conservation and structural restoration of architectural heritage. For this purpose, the ICOMOS charters and documents [1,2] still provide a valuable methodological synthesis and a sound guide to perform the diagnostic process. According to a principle stated in [2], “Diagnosis and safety evaluation of the structure are two consecutive and related stages on the basis of which the effective need for and extent of treatment measures are determined”, where diagnosis is based on different approaches—historical, qualitative, and quantitative—and safety evaluation is the final phase, which should reconcile qualitative with quantitative analysis (i.e., visual examination, historical and archaeological research, structural analysis, and testing). For the practical application of these basic concepts, there is a wide range of methods and techniques to rely on and a wealth of literature on the subject. For the development of guidelines for the assessment of timber structures, COST Actions reports and publications are particularly relevant [3,4]. Regarding heritage structures, for a comprehensive overview of existing procedures, guidelines, and standards with a rich bibliography, Riggio et al. [5] is an essential read. Also, Perria [6] provides an interesting methodological proposal. Nevertheless, although methodologies and tools are well known, the evaluation of the load-bearing capacity of timber structures remains a difficult problem to solve, especially when the timber structure should fulfill strict regulatory serviceability requirements. In situ load tests should provide conclusive proof, but high costs and practical difficulties make this choice, if not impossible, quite problematic in most cases. Furthermore, whatever the methodology used, a problem—recurring but not frequently addressed—concerns costs, which almost always means a compromise between the achievement aims and available resources. For all these considerations, it is evident that the development of reliable comparison and calibration processes to find good correlations between NDT and DT is a primary goal to meet [7,8].

The case study presented here concerns one of the timber floors of the Tower of Montorio, a medieval architectural complex in north Italy. (For a description of the architecture of the monument, see Carpani 2013 [9].) The load-bearing walls date back to the early 13th century, and the timber structure is relatively recent, having been built at the end of the 19th century. During renovation works in the second half of the last century, an additional thick, unreinforced concrete slab and new partition walls significantly increased the superimposed loads, causing a marked inflection of most of the beams. The situation worsened following the earthquake that damaged the tower in September 2003 [10]. In the framework of post-quake rehabilitation and restoration works, an experimental campaign was carried out on the second-floor timber beams [11]. The in situ evaluations revealed that they were made of local oak (Quercus robur L.) and showed that their state of conservation was extensively compromised. Subsequently, it was decided to replace six of them, offering a rare opportunity to carry out a comprehensive experimental campaign in the laboratory, including both non-destructive and destructive tests (four-point bending tests) [12,13].

This paper will focus on the laboratory experimental campaign, describing and studying it in detail, to suggest a methodological approach and to establish some useful correlations between the results of NDT and mechanical parameters obtained by DT. The paper also presents an innovative way of data visualization based on a geometric representation that allows for a quick interpretation of the experimental results.

2. Methodological Approach

Six timber beams removed from an intermediate floor of the Montorio Tower after the earthquake of September 2003 were analyzed at the Laboratory of Mechanical Analysis of Brasimone ENEA Research Centre.

For each specimen of the sample, the evaluation was carried out through three phases: (i) visual grading; (ii) non-destructive tests (NDT), which included the thermo-hygrometric, the sclerometric, and the resistographic tests; and (iii) destructive 4-point bending tests (DT) until the specimen broke. In the first phase, a brief evaluation of the state of preservation, leading to the strength class, was performed; subsequently, NDT and DT allowed us to obtain a measure of the mechanical characteristics. Finally, some correlations between the results of NDT and DT were analyzed and discussed.

In the following paragraphs, each phase is described in detail in terms of the most significant results.

It is worth noting that the limited number of beams did not allow us to obtain statistically significant results. However, when dealing with architectural heritage, ancient elements are rarely available to remove and test through destructive methods. For this reason, the paper aims to provide a methodological approach and to share, within the scientific community, data that are especially valuable because they are rare.

2.1. Visual Grading

The visual grading (VG) was performed according to the UNI 11119:2004 [14]. Such an inspection aimed to evaluate the conservation status of timber elements to assign to each of them a class of resistance from I to III. (Class I corresponds to the best achievable outcome.) Firstly, a photographic survey was carried out; then, the pictures were arranged and classified, as shown in

Table 1. Visual inspections of the defects due to the events/needs for use and to the tree’s growth conditions.

	Smoothed Edges	Knots	Grain Slope
Beam 1
Beam 2
Beam 3
Beam 4
Beam 5
Beam 6

and

Table 2. Visual grading of the defects due to the serviceability conditions of the beams.

	Cracks	Deformations	Biological Attack
Beam 1
Beam 2
Beam 3
Beam 4
Beam 5
Beam 6

.

Pictures should be as such to document the presence (or absence) of defects and degradations, and they should be grouped according to the following causes: environmental conditions where the trees have grown (which may result in knots and grain slop); events/needs for the use of the beams (which may result in smoothed edges); and serviceability conditions (which may result in cracks, deformations, and biological attacks). Such an organization is useful, and recommended, for proper data archiving and consulting.

Subsequently, defects and degradations should be quantified and recorded on appropriate forms (Figure 1). A conceptual representation of the beam should also be included to indicate the two ends (left and right, respectively); the four sides (A, B, C, and D); and the three main cross-sections (left, middle, and right).

Forms in Figure 1 were designed to comply with the classification criteria provided by UNI 11119:2004 [14] and filled in with data referred to in the case study. For each element, forms should contain the parameters listed below.

• Size of the beam (longitudinal length (L), base (b), and height (h) at the left, the middle, and the right sections, respectively).
• Wood species.
• Size of the smoothed edges.
• Minimum diameter (d_min) of the knots (S indicates single knot, and G indicates group of knots) and the corresponding dimension of the effective section (Hf).
• Grain slope refers to the longitudinal axis of the element.
• Cracks due to shrinkage in terms of the sizes of the openings (O) and depth (D). N.D. stands for not detected, which means that crack thickness was not significant.
• Ring shake (R) and cracks due to lightning/freezing/damage (L/F/D).
• Deformations due to arching (A), sickle (S), twist (T), and warping (W).
• Degradation due to fungi (blue stain (BS), white caries (WC), and brown caries (BC)).
• Insect attacks.

The main information obtained from data collection and recorded in Figure 1 can be qualitatively summarized as follows: all the beams were made of oak (Quercus robur L.); shrinkage cracks with a considerable number of openings were evident; knots determined a local grain slope, which appeared to be very steep in beams 1, 3, 5, and 6; beams 1 and 3 had smoothed edges (sides C–D and A–B, respectively); and beams 2, 3, 4, and 5 had previous bending deformations. Moreover, previous biological attacks, as well as galleries of insects and cubical brown rot, were noted.

As a result of the VG, a class (from I to III) was assigned first to each of the surveyed parameters and then to each of the specimens of the sample (

Table 3. Classes of resistance based on visual grading.

	Smoothed Edges	Cracks Ring Shake	Single Knots	Group Knots	Grain Slope (Gradient %)	Shrinkage Cracks	Element Class
Beam 1	III	III	I	I	II	II	III
Beam 2	I	III	I	I	II	III	III
Beam 3	III	III	I	I	I	III	III
Beam 4	I	III	I	I	I	III	III
Beam 5	I	III	I	I	I	III	III
Beam 6	I	III	I	I	I	III	III

).

2.2. Non-Destructive Tests

After the visual assessment described above, an extended NDT campaign was carried out. It included (i) ultrasonic tests to evaluate the degrees of homogeneity of the material within the single specimen and within the whole sample; (ii) sclerometric tests, which examined the surface hardness of the material; and (iii) resistographic tests to evaluate the homogeneity of the material and the extent of internal defects or degradations. Details of the campaign can be found in [11], which also reported the moisture content of the beams, the results of which were quite homogeneous, with an average value of 12%.

Tests were carried out on both transverse (S_T) and longitudinal (S_L) sections, except for the ultrasonic analyses, which were carried out only on transverse sections.

In Figure 2a, a schematic representation of the surveyed sections (S_L,L, S_L,R, S_T,L, S_T,M, and S_T,R) is shown, in addition to some pictures of instruments taken during the test phases.

For ultrasonic tests, the Novatest^® I-SONIC device was used. It was composed of a data acquisition control unit and a pair of contact probes (both working at a frequency of 53 kHz) for the transmission and receipt of compression waves through the material. Tests were performed using the direct method, which consisted of placing the transmitter and the receptacle on two opposite sides of the beam and computing the wave propagation velocity (V) as the ratio between the path length and the flight time.

Sclerometric tests were carried out by employing a Woodpecker^® system. It consisted of a steel needle, characterized by a 60 Rockwell hardness, which was fixed into the wooden tissue by five subsequent blows. In the meantime, an external comparator measured the portion of the needle that did not penetrate the wooden tissue.

Each test area had a 50 × 50 mm² surface that was broken up into a grid with nine measuring points (Figure 2). Therefore, adding to the three transversal sections (S_T,L, S_T,M, and S_T,R) already subjected to the ultrasonic evaluations the two longitudinal sections (S_L,L and S_L,R), in total, 126 points were analyzed.

Resistographic tests were conducted through the IML Resi F400^® device. It measured the timber penetration strength of a drilling needle propelled by a sophisticated power tool. The instrument output consisted of a chart where the Resistograph^® amplitude (Am) was plotted as a function of the penetration depth (see the resistographic profile in Figure 3). Since Am is a measure of the energy expended in penetrating the material, it may be related to the internal quality of the beam, and it may evidence density variations.

The results of the tests are summarized in

Table 4. Ultrasonic test results.

	V [m/s]
	ST,L	ST,M	ST,R
Beam 1	1612.7	947.8	951.8
Beam 2	1115.4	684.5	missing
Beam 3	1043.9	1330.5	1118.8
Beam 4	773.9	758.6	785.3
Beam 5	734.3	943.5	754.0
Beam 6	897.8	1106.5	906.1

,

Table 5. Sclerometric test results.

	P [mm]
	ST,L	ST,M	ST,R	SL,L	SL,R
Beam 1	12.29	8.89	9.94	10.78	10.50
Beam 2	10.74	10.65	12.11	10.34	10.06
Beam 3	8.55	7.93	7.98	10.16	11.60
Beam 4	12.02	11.04	12.48	17.65	12.93
Beam 5	8.88	10.46	9.22	10.88	10.07
Beam 6	10.43	11.70	12.32	10.62	*

* Full penetration of needles.

and

Table 6. Resistographic test results.

			RM [%]
	ST,L	ST,M	ST,R	SL,L	SL,R
Beam 1	-	12	22	28	27
Beam 2	15	31	23	48	52
Beam 3	26	45	26	86	10
Beam 4	17	11	17	10	29
Beam 5	45	19	44	50	3
Beam 6	17	35	12	62	3

. For each beam and each section, the tables show the average ultrasonic velocity V (Table 4), the average sclerometric penetration length (P) (Table 5), and the resistographic measure (RM) (Table 6). The RM is the ratio between the area under the resistographic profile (not considering the damaged zones) and the penetration depth.

The comparison among results highlighted a consistency among Resistograph^®, ultrasonic, and sclerometric tests: all of them confirmed the poor preservation state of the beams. In the transversal direction, they revealed the significant degradation state of beam 1 and beam 4 and the partial degradation state of beams 2, 3, 5, and 6. Moreover, in the longitudinal direction, beams 5 and 6 appeared to be significantly degraded at the right end, whereas beams 3 and 4 were partially degraded at the right end and at the left end, respectively. However, owing to the very low values, the Resistograph^® amplitudes obtained for the transverse sections of beam 1 were not considered in deriving correlation curves.

2.3. Destructive Tests

The experimental destructive test setup was carried out following UNI EN 408 [15]. As shown in Figure 4, each specimen was simply supported at two points close to the ends, and it was loaded with vertical forces, F/2, applied at two points near the middle of the element. The response of the beam was detected through five linear variable differential transformers (LVDT), which measured vertical displacements (w): D3 was positioned at the lower edge of the middle section of the beam, whereas D1, D2, D4, and D5 were positioned along the neutral axis. The load was increased until the beam broke. This was the four-point bending test.

In

Table 7. Beam configurations at the failure.

	Failure Configuration	Failure Detail
Beam 1
Beam 2
Beam 3
Beam 4
Beam 5
Beam 6

, some pictures showing the final stages of the test are reported. As evident, none of the specimens (except for beam 3) highlighted a typical bending behavior, which would have been characterized by an almost vertical breaking.

Data taken from destructive tests allowed us to compute the bending modulus and bending strength through Equations (1) and (2), respectively.

$$E_{m,g}={\displaystyle \frac{L^3\left(F_2-F_1\right)}{b{h}^3\left(w_2-w_1\right)}}\left[\left({\displaystyle \frac{3a}{4L}}\right)-{\left({\displaystyle \frac{a}{L}}\right)}^3\right]$$

(1)

$$f_m=a{\displaystyle \frac{F_{max}}{2W}}$$

(2)

where W is the section modulus, (F₂ − F₁) is the load increment, (w₂ − w₁) is the corresponding displacement increment within the linear portion of the force–displacement curve, and F_max is the breaking load. The other parameters (L, a, b, and h) had geometrical meaning and may be inferred from Figure 4.

Due to the specimen inhomogeneity, in terms of geometry defects and degradations, values of E_m,g and f_m were scattered, as shown in Figure 5, where measured values (MV) are indicated by blue diamonds. Moreover, they had values lower than the ones obtained by extending the manufacturer correlation curve (MCC), superimposed by a dotted red line in the same figure. This correlation curve was almost linear and was derived by combining the two curves, which related penetration lengths with the Young’s modulus and with bending strength, respectively, provided by the Woodpecker^®. In the same figure, regression equations and r-squared coefficients (R²) are superimposed. As evident, a linear equation did a good job of fitting the manufacturer data.

Results of Figure 5 can be interpreted with the help of the pictures in Table 7, which show the inner part of the specimens. Although the behavior of the timber elements was difficult to code, and general rules cannot be established, through this approach, useful qualitative evaluations were carried out.

• Beam 1 was the most rigid element of the sample, even if its bending strength was lower than expected. This may be due to the poor adherence among longitudinal wood fibers, which resulted in a diagonal breaking that started from pre-existing damage.
• Beam 2 exhibited a low bending modulus, likely due to significant pre-existing damage, which caused a relative motion among fibers. The bending strength also seemed to be compromised by damage, and indeed, the breaking started from it.
• The bending modulus and the bending strength of beams 3, 4, and 6 were below the expected values, even though they showed a linear trend close to the theoretical behavior. Beam 3 was the only one that exhibited an almost vertical breaking, even if the resistant cross-section was not fully involved.
• Beam 5 had a bending modulus like beam 6, whereas its bending strength was only 69% of beam 6. In this regard, it is worth noting that beam 5 had a section reduction near the breaking point.

3. Comparison and Correlation between Destructive and Non-Destructive Tests

3.1. Comparison between Bending and Sclerometric Tests

The average values of the penetration length (P_T) at the sections of the beams of the sample are summarized in Figure 6a. This graphical representation delivers a comprehensive overview of results, especially effective in the case of larger samples. The following rules apply.

• The more regular each pentagon is, the more homogeneous the corresponding beam is expected to be. (In case the pentagon is equilateral, the element would be isotropic).
• The more similar the pentagons are to each other, the more homogeneous the mechanical behavior within the sample is expected to be.
• The smaller each pentagon is, the more rigid the corresponding beam is expected to be.

In the case study, pentagons were irregular and allowed us to catch the main differences among the specimens of the sample: beam n.4 was the most irregular, and the behavior of the sample was more homogeneous in the longitudinal direction than in the transversal direction.

If we focused on the three transversal sections (S_T,L, S_T,M, and S_T,R), excluding the two longitudinal ones (S_L,L, S_L,R), the pentagon became a triangle (Figure 6b), which is more adequate to describe the behavior of an orthotropic material like wood.

The triangle of beam n.3 was the smallest, and it appeared to be almost equilateral. This result is interesting because this beam was the only one that had an almost vertical breaking section when subjected to bending tests (see Table 7). The triangle of beam n.4 was almost regular was well. However, this was one of the bigger ones, and consistent with this result, values of fm and Em were among the smallest we obtained using bending tests (see Figure 5). Triangles of beams n.1 and n.5 were small but irregular. This result is interesting because these two beams had the highest bending modulus values, but the bending strengths were lower than expected.

In Figure 7, Figure 8, Figure 9, Figure 10 and Figure 11, the correlation between the penetration lengths and the bending test results (in terms of bending modulus and bending strength) is shown in detail. In the same plots, manufacturer correlation curves (MCC) are superimposed with solid red lines. The dashed line is the MCC in terms of the Young’s modulus reduced by 3 GPa (MCC—3 GPa), whereas dotted lines represent MCCs in terms of the Young’s modulus and bending strength reduced by 6 GPa (MCC—6 GPa) and 55 MPa (MCC—55 MPa), respectively. These curves did a good job of fitting measured data at the middle transversal sections of the beams, as detailed in

Table 8. Regression equations and r-squared coefficients of correlation curves at the middle transversal sections of the beams.

Penetration Length x [mm] vs. Young’s Modulus y [GPa]	Correlation Curve	Regression Equation	r-squared coefficient
	MCC—3 GPa	$y=-0.5345x+14.448$	R2 = 0.80 referred to Beams 1, 5 and 6
	MCC—6 GPa	$y=-0.5345x+11.448$	R2 = 0.78 referred to Beams 2, 3 and 4
Penetration Length x [mm] vs. Bending Strength y [MPa]	MCC—55 MPa	$y=-2.9735x+52.283$	R2 = 0.51 referred to Beams 1, 3, 4 and 5

, where regression equations and r-squared coefficients are summarized.

The main aspects that stand out from the figures before and from Table 8 are summarized below.

• Values of the bending modulus and the bending strength experimentally computed were significantly lower than the MCC values. This circumstance pointed out the fact that the results of sclerometric tests alone would lead to an overestimation of the mechanical parameters of the elements. On the other hand, as a result of visual grading, all the beams were classified as elements of class III (which is the worst class established by UNI 11119:2004 [14]).
• A certain correlation within the set of values of the bending modulus and the bending strength experimentally computed was evident, with particular reference to the middle transversal sections. Bending modulus values of beams 1, 5, and 6 appeared to be distributed close to a correlation curve, which was the Young’s modulus MCC reduced by 3 GPa. Bending modulus values of beams 2, 3, and 4 appeared to be distributed close to a correlation curve, which was the Young’s modulus MCC reduced by 6 GPa. Bending strength values (except for beams n.2 and n.6, which had the lowest and the highest values, respectively) appeared to be distributed close to a correlation curve, which was the bending strength MCC reduced by 55 MPa. As further confirmation of this point, measured values of the Young’s modulus and bending strength at the middle transversal sections were compared with values computed through the regression equations given in Table 8. Results are summarized in Table 9.

Table 9. Comparisons among measured values (MV) and values computed through regression equations for the Young’s modulus and bending strength at the middle transversal sections of each beam.

Beam	Young’s Modulus [GPa]			Bending Strength [MPa]
	MV	MCC—3 GPa	MCC—6 GPa	MV	MCC—55 MPa
1	9.81	9.70	6.70	22	26
2	5.74	8.76	5.76	12	21
3	7.37	10.21	7.21	28	29
4	4.65	8.55	5.55	15	19
5	8.50	8.86	5.86	24	21
6	8.50	8.19	5.19	35	17

Table 9. Comparisons among measured values (MV) and values computed through regression equations for the Young’s modulus and bending strength at the middle transversal sections of each beam.

Beam	Young’s Modulus [GPa]			Bending Strength [MPa]
	MV	MCC—3 GPa	MCC—6 GPa	MV	MCC—55 MPa
1	9.81	9.70	6.70	22	26
2	5.74	8.76	5.76	12	21
3	7.37	10.21	7.21	28	29
4	4.65	8.55	5.55	15	19
5	8.50	8.86	5.86	24	21
6	8.50	8.19	5.19	35	17

• Results of sclerometric tests in the longitudinal direction at the two ends of the beams (S_L,L and S_L,R) showed the worst correlation in terms of bending parameters. In particular, within the sample, the penetration lengths of the specimens were very close to each other, whereas both the bending modulus and the bending strength were very different between one beam and the next. On the other hand, the two ends of the beams were far from the section where the bending modulus was computed (at the middle of the beam). Moreover, beam n.4 recorded a penetration length that was significantly different from the MCC.

3.2. Comparison between Bending and Ultrasonic Tests

In each section (S_T,L, S_T,M, and S_T,R) and for each of the two paths (AC and BD, respectively), four measures were repeated, and the two average values of the velocities (VAC and VBD, respectively) were obtained. The section’s velocity V was calculated as the average of VAC and VBD, and the results are summarized in Figure 12.

As evident in Figure 12, triangles were very different from each other (both in shape and in size), suggesting that the elements of the sample had different mechanical behaviors. In general, all triangles were not equilateral, and mechanical properties were not homogeneous along each beam, except for beam n.4, whose triangle was also the smallest. However, in this case, a small triangle was suggestive of poor mechanical properties. What is very interesting is that the triangle of beam n.3 was one of the biggest, consistent with the fact that in Figure 6b, it was the smallest. The red triangle was not closed. It referred to the beam n.2, for which the value of the velocity at the S_T,R section was missing.

In Figure 13, Figure 14 and Figure 15, the propagation velocities were placed in relation with the results of the bending tests. In general, as expected, mechanical properties (bending modulus and bending strength) increased with propagation velocities. However, a certain correlation was evident only in terms of fm at the S_T,M section.

3.3. Comparison between Bending and Resistographic Tests

In Figure 16a,b, the values of RM, corresponding to transverse and longitudinal sections of each beam, are summarized. In this case, figures were irregular and differed significantly from one another. Then, based on resistographic tests, the behavior of the whole sample, as well as the behavior within each beam, was not homogeneous. This is also confirmed by Figure 17, Figure 18, Figure 19, Figure 20 and Figure 21, where no correlation can be established between the RM and the bending parameters.

4. Conclusions

A comprehensive approach combining visual grading and destructive and non-destructive tests was conducted on some timber beams taken from an intermediate floor of the Montorio Tower damaged by the earthquake that occurred in those areas in September 2003. After this tragic event, some timber beams were replaced, offering the rare opportunity to study these ancient elements in detail, although limited in number. The limited number of beams (a total of 6) did not allow us to obtain statistically significant results. However, when dealing with architectural heritage, ancient elements are rarely available to remove and test through destructive methods. For the same reason, these data are especially valuable. The aim was to assess their preservation state, as well as to deepen possible correlations between the results of NDT and the mechanical properties of the beams. Visual grading revealed significant defects and degradations, leading to the classification of all beams into class III. The NDT (including ultrasonic, sclerometric, and resistographic tests) provided further insights into internal degradations and highlighted the variability in terms of the hardness and density of the material. Moreover, the actual mechanical properties computed through DT were generally lower than expected, considering the resistant sections. This is consistent with the results of NDT, which estimated a poor preservation state of the structural elements. Despite being local tests, the NDT provided significant information about the element’s integrity. A strict numerical correlation between mechanical properties obtained through NDT and DT has not yet been obtained, but it is well known that this is an ambitious goal. However, this study achieved a twofold objective. On the one hand, it allowed us to obtain and share data deriving from destructive tests, which are especially valuable because they are rare. On the other hand, it demonstrated that non-destructive tests are useful and reliable for in situ evaluations when the conservation of the ancient elements is a priority and reducing destructive actions is needed.