1The close links between the two authors has enabled the resources of the historical data and artefacts in Sheffield to be brought together with the engineering and computing resources in Nottingham. Following discussions about the internal structure of cutlery and tools which might indicate manufacturing processes, the non-destructive techniques of ultrasound were attempted on modern and historical artefacts. The initial results of this collaboration are presented in this paper, which is divided into two parts. The first part describes the background to the development of the cutlery manufacturing industry in Sheffield and the second explains the potential application of advanced non-destructive evaluation techniques to study the internal structures, and hence the manufacture, of post-medieval ferrous metal products.

Cutlery manufacturing industry in Sheffield

2The Sheffield metalworking trades, notably cutlery, have had centralised control since at least the mid 16th century and following the establishment of the Company of Cutlers in Hallamshire in 1624, there is an impressive collection of documents showing trade and manufacturing organisation. Hallamshire was an administrative unit centred on the parish of Sheffield, together with the adjacent parishes of Handsworth and Ecclesfield and in the 16th century, it was administered by the Lord of the Manor of Hallamshire, the powerful Earls of Shrewsbury.1 The Lords had taken an active and financial interest in the local trades, which employed a large proportion of the male population. By the early 17th century, the Manorial Court had become the focus for controlling the numbers of apprentices, enforcing a number of holidays throughout the year and registering the personal marks of the trained craftsmen. Following the death of the last resident Lord in 1616, the Hallamshire metalworkers approached Parliament for an Act of Incorporation to maintain this order in the industry. The Act established the Company consisting of a Master Cutler, two Wardens, six Searchers and twenty-four Assistants, elected annually from among trained craftsmen, who were known as the Freemen.2

3The Act of Incorporation in 1624 set out the overall organisation of the metal manufacturing industry in Hallamshire, and extended the Company’s control to the craftsmen in an area six miles beyond Hallamshire’s boundaries. The main aims of the Act were similar to those of the old Manorial Court–to restrict the number of apprentices entering the trades, to maintain a strict demarcation between the different craft groups and to insist that all Freemen register their identifying marks which they were to stamp on all their manufactured goods. The Act also included a remarkable clause, insisting that all manufactured knives, etc. had to have an edge of steel (the older Worshipful Company of Cutlers in London having no such rule). The Searchers of the Company were entitled to enter premises to search out “deceitful wares” such as those not having steel in the edge, and the craftsmen could be fined– hence, the importance of carefully recording of Freemen’s marks. The cutlers undoubtedly understood the qualities of steel for cutting edges even though the chemistry of steel making had not yet been developed. They knew that steel, when hardened and tempered, could retain a sharp cutting edge far longer than iron, but because of the relative manufacturing costs of iron and steel, steel was confined to the cutting edges of their wares. This resulted in a variety of manufacturing methods for attaching the steel.

4The craftsmen who were to be governed by the Company in 1624 were the cutlers, who made knives, the scissorsmiths, shearsmiths and sicklesmiths and in the 1670s and 1680s, further groups submitted to the Company’s organisation–the scythesmiths, the awlbladesmiths (making leatherworking tools) and the filesmiths. The principal processes in cutlery and edge-tool manufacturer were forging, grinding and assembling, which might include other processes such as file cutting and putting together scissor blades. These processes could be done by one trained man and in very simple premises. The primary production unit was a master craftsman who might train an apprentice and who may or may not also have had journeymen working with or for him. The workplaces and tools were relatively simple, but some of the processes (heavier forging and grinding) required power sources, such as water-powered wheels.

5Although Sheffield was to become a world supplier of iron and steel in the 19th century, there was only a small amount being produced locally in the 16th and 17th centuries. With a tradition of local iron smelting dating from the Middle Ages, imported iron ores were being used to produce iron at a small number of furnaces around Sheffield and locally produced steel only became available following the construction of cementation furnaces in Sheffield at the beginning of the 18th century.3 Prior to this, Sheffield cutlers were dependent on merchants for their supplies of imported iron and steel. Evidence for the types and quantities of metals used by Sheffield craftsmen comes from the archives of the Cutlers’ Company. These archives are excellent in quality and continuity, having for example, a set of mark registers from the Manorial Court of 1614 to the end of the Company’s role as a trademark sub-registry in 1998. The archives also contain account books from 1625 and minute books from the 1720s, as well as records of its commercial activities in sourcing raw materials, quarry ownership and steel manufacture.

The Storehouse of the Cutlers’ Company, 1680s

6In the early 1680s, in a brave attempt to prevent the local craftsmen, and scissorsmiths in particular, being exploited by merchants, both as suppliers and buyers, the Cutlers’ Company established a Storehouse for the sale of raw materials and the purchase of finished goods, which were then sold on to merchants. The records of this short-lived enterprise detail scores of men buying their iron, steel and handle material, such as ivory, from the Company store and selling back their finished goods. From these records of a few years’ commercial activity, it is possible to calculate the relative costs of the raw materials, to estimate the running costs of the craftsmen and estimate the amount of production in Hallamshire.4

7The accounts show that a London merchant, Mr Guillims, seems to have been the main supplier of imported material, supplying in one instance, 13 hundredweight of German steel for £24.14s.0d. By the 1680s, wrought iron appears to have been supplied solely by the local forges of Wadsley and Attercliffe, with the Storehouse recording a cost of £14 per ton. This gives a cost of almost two pounds per hundredweight for steel and almost 15 shillings per hundredweight for iron. From these records, it is clear that the relative costs of iron and steel dictated that the use of steel was restricted to the edge only and that the bulk of the knife, scissors or scythe blade was made of wrought iron, a fact clearly appreciated by the authors of the 1624 Act. The way in which the craftsmen purchased their iron and steel confirms this.

8The Storehouse records have dated entries for the purchases of iron and steel–the weight of metal bought and the cost to the craftsmen. For example, in November of 1681, three craftsmen, chosen randomly, were two scissorsmith John Hobson and Matthew Stevin and Joseph Dungworth, a cutler. These three men were in their twenties, having finished their apprenticeships and become Freemen in the mid 1670s. On one day, they purchased the following amounts of iron and steel:

9These sales of raw materials give a price of 2½d per pound of steel, while iron per pound was approximately half that price. Interestingly, on the same day as the above sales of metal, a Thomas Hobson purchased just over fifteen pounds of German steel, which was almost double the price of “ordinary” steel, but it is not clear from the Storehouse records what the origin of this “German” steel was. These calculations show that the Act of Incorporation, which insisted on cutlery having an “edge of steel,” was in line with the market in metals at the time. The cost of cutlery would have been much higher had the article been made completely from steel. Therefore, the cutting edge alone was of the more costly metal, but this involved the craftsmen in a more complex production process, necessitating the attachment in some way of the steel to the iron making up the bulk of the articles.

10In order to appreciate how the cost of these raw materials was converted into the selling price of the manufactured goods, it is necessary to look at the sale of finished items back to the Storehouse. With some commercial naivety, the Cutlers’ Company appears not to have added any overheads to the costs of the raw materials or to the price of finished goods sold on to the merchants. The Storehouse was established to help the craftsmen, especially the scissorsmiths, who agreed to sell their goods only through the Storehouse. With a guaranteed sale, the craftsmen could hardly have believed their luck and the records suggest that craftsmen each had a cupboard in which their finished goods were stored. The Storehouse records show scissorsmiths were selling their wares to the Storehouse at prices ranging from 9d to 20d per dozen and generally the Storehouse paid out a total of about £35 a week for scissors. The merchants visiting the Storehouse could then purchase goods from one or more craftsmen. The records show that about 120 craftsmen (about fifty per week) availed themselves of the Storehouse facilities to buy and sell goods. One craftsman was followed for a few months in 1680/1681, chosen simply because his unusual surname was easily identified in the list of names.

11Edward Brittlebank was the son of a husbandman from a village called Thorpe, to the north of Sheffield, and was apprenticed to John Radcliffe in 1646, gaining his Freedom in 1656. He lived in Sheffield, being taxed in 1672 for two domestic hearths and one smithy hearth. He trained his son, Richard, and five other apprentices during his working life, which continued into the late 1690s. He would have been in his fifties when he was dealing with the Storehouse.5

12In three months from November 1681, Edward went to buy iron about once or twice a week, buying between 28-49 pounds and spending on average, 72d. He also bought smaller quantities of steel–between three and seven pounds, which cost him between 7d and 1s.5d Two or three times a week, he sold the scissors he had made, in varying amounts from six dozen (72 pairs) to one gross six dozen (216 pairs). The numbers of scissors he sold was reflected in the metal he bought, which suggests that he was not carrying any reserve stock, only buying what he would use in a week. Table 2 is a summary of a short period of Brittlebank’s dealings with the Storehouse.

13Edward Brittlebank sold his scissors for 18d per dozen (½d per pair) and from the amounts of iron and steel he purchased from the Storehouse, assuming he bought all his metal from the Company, it is clear that steel was a very small part of each pair of scissors–only along the cutting edge. These records confirm, in detail, the manufacturing practice of using as little steel as possible. These records are also an indication of the scale of cutlery manufacture in Sheffield at the end of the 17th century. One man was able to sell almost 2,500 pairs of scissors in a three month period. From the number of scissorsmiths using the Storehouse, it is estimated that about 725,000 pairs of scissors were sold each year.6 This is an astonishing amount, which would have increased over the decades as the number of scissorsmiths also increased. It is regrettable that the cutlers did not sell their knives through the Storehouse, in order to obtain an estimate of their output.

14Blades were forged from bars of metal, but the craftsmen had to weld the steel to the iron and this was achieved in a variety of ways, as surviving examples show. One method, especially for the larger agricultural implements such as scythe blades, was to make an iron-steel-iron sandwich of the metals, a technique which continued into the 20th century in the manufactured of “crown” scythes. Another way was to weld a thin section of steel onto the blade, as can be seen in the 17th century pair of scissors (fig. 1). Here, the scissors have been made from wrought iron, with a very thin layer of steel along the inside face of each blade. A third method of attaching the steel can be seen in examples of table cutlery made from iron and shear steel. The cutler manufactured the iron bolster and tang of the knife and then attached the steel for the blade. The steel was laid on top of the iron of the bolster, with a small section of iron being on the back face of the blade. As the blades became less polished in use and were attacked by a variety of chemicals and food acids, the outline of the iron on the back of the blade became more pronounced and looking like a “thumb print.”

15Turning from the documentary evidence of the activities of Sheffield craftsmen which has shown the use and quantities of iron and steel in cutlery blades, the second part of this paper will describe methods for studying the construction of the blades, from which one can extrapolate the probable methods of manufacture. Assessment of historical artefacts often results in their destruction and while non-destructive methods, such as X-rays are of value, there is another method of non-destructive examination which can add more information about the knife blade, such as the thickness of the metal layers and the effectiveness of the welding of metals.

Advanced non-destructive evaluation techniques

16Ultrasonic techniques for the Non Destructive Evaluation (NDE) of engineering structures are widely used in many manufacturing industries. With increasing use of adhesive bonding and the introduction of complex, multilayered, composite materials in the aerospace and automotive industries, for example, it is imperative that testing procedures are developed to ensure quality assurance in safety critical applications. A range of information can be determined such as the location and size of cracks, delaminations and other defects, the overall dimensions, estimation of particle size and porosity content.

17An increasingly wide range of ultrasonic testing equipment is becoming commercially available and academic research, in partnership with industry, is competitive and dynamic in this area. This has been facilitated by increases in available computer power, the application of sophisticated signal processing techniques to measured data and greater understanding of the propagation of ultrasound through complex materials.

18Although some work has been done using simple ultrasonic techniques to study archaeological finds, there is little published in the literature about the application of ultrasound to the study of layered metal structures. There are a number of practical considerations that make this area of research challenging. The surfaces of found metal objects are likely to have a certain surface roughness making the insertion of sufficient ultrasound energy into the structure difficult. The other main issue is the scattering nature of such metals due to their highly crystalline structure and the presence of a variety of inclusions. However, there are now a number of signal processing techniques that can be applied in such cases to minimize these effects such as spatial averaging and the use of higher energy non-pulse signals such as chirps and pseudo-random binary sequences.7

19Despite these complications, there are a number of advantages to the development of sensitive ultrasonic NDE techniques for archaeological applications. Currently available techniques are often destructive in nature where a section of a find must be permanently cut out. Any information obtained from this cut-out section may well be useful in determining chemical composition and structure but only pertains to that point and variations in structure across the entire object are not usually feasible. Although it is not possible to use ultrasonic techniques to determine chemical composition, it may not be possible in some circumstances to use X-ray techniques to determine structure either. Clearly ultrasonic techniques could not replace current diagnostic tools used in archaeology but it is hoped that ultrasonic techniques will be a useful addition.

20In this paper we consider the application of ultrasonic NDE techniques developed for the assessment of engineering structures to two metal artefacts.

Background theory

21Most ultrasonic NDE techniques involve sending a short pulse signal (equivalent to an audible click) into the object under test and using knowledge of the propagation characteristics to determine information about the structure. Ultrasound is simply high frequency sound usually in the frequency range 20kHz upwards. Sound, in turn, is simply a mechanical vibration. As such, sound waves or signals will not travel through a vacuum and instead must have a medium through which to propagate. The nature of the medium will determine the velocity at which the sound travels, with a sound signal travelling faster through a solid than a liquid, and faster through a liquid than a gas. For solid media, velocities are generally higher in metals such as steels and aluminium than in most polymers and plastics. Having said that, the measurement of velocity cannot be used as an accurate diagnostic tool for precise material identification. However, relative variations of velocity within a structure are useful indications of variations in the manufacturing process parameters and chemical composition (for example, in the processing of structural epoxy-based adhesives8 or heat treatment of metal surfaces).9

22The other parameter describing the propagation characteristics of a sound signal is attenuation which is a measure of the loss of energy in a signal as it travels through a medium. The exact nature of attenuation is complex although, in polymer materials, it is primarily due to absorption and in crystalline metals it is principally due to scattering from grain boundaries and defects.10 Exact measurement of attenuation is difficult since a variety of experimental issues will also have an affect the signal amplitude.11 Surface roughness will reduce the incident signal level leading to possible signal-to-noise issues. The liquid couplant used between the transducer and the object can also affect signal level to an unknown extent and also affect measurement reproducibility. Finally, attenuation (whether due principally to scattering or absorption) is highly frequency dependent and increases (following some function) with increasing measurement frequency. In other words, some sort of frequency domain analysis of the measured data must be performed carefully to ensure an accurate assessment of the attenuation of a particular material. Time domain analysis is not sufficient and will only result in a crude estimation which is unlikely to be particularly useful for NDE purposes.

23When a sound wave propagates through a medium consisting of different materials, for example, in a layered structure, the energy in the signal splits at the interface between the two materials. Part of the energy is transmitted across the interface and part is reflected back. The ratio between the transmitted and reflected energy is determined by the relative values of the acoustic impedance of each material, where the acoustic impedance (Z) is given by:12

where c is the velocity in the material and ρ is the density. For a sound signal travelling across a boundary from material 1 into material 2 then the reflection coefficient (when considering the signal amplitude) is given by:

where Z₁ and Z₂ are the acoustic impedance values of materials 1 and 2 respectively. The reflection coefficient gives a value between -1 and +1 representing the proportion of incident energy reflected back into material 1 and any signal inversion. For most situations it is acceptable to assume that there is no energy loss within the interface itself.

24There are two extreme interface reflections. The first occurs when there is a large acoustic mismatch between the two layers, for example, between a metal and air. In this case a signal travelling through the metal into air will be almost entirely reflected back from the interface. The other scenario occurs when the acoustic impedance values of the two materials are very similar (for example, between iron and steel). In this case, very little energy would be reflected back at such a boundary and almost all the sound signal will be transmitted across it. When testing found metal objects consisting of layers of similar metals both these situations will occur. The ultrasonic pulse will enter the object and propagate to the metal-metal interface when most of the energy will be transmitted across it. The pulse will continue through the object to the other side when all of the energy will be reflected back due to the acoustic mismatch between the metal and air.

25The thickness of a single-layer object or a layer in a multi-layer object can be determined from the ultrasound data by measuring the time between the relevant reflected echoes. For example, the time (t) between an echo from the transducer-top surface interface and the back-surface-air interface will be the transit time for the pulse to travel twice through the thickness (d) of the object. If the velocity (c) is known a priori then the thickness can be calculated using the simple formula:

Similarly if the time and thickness are known then clearly the simple time of flight velocity can be determined for a particular layer (neglecting of course, any frequency dependence in the propagation velocity).

Experimental techniques and samples

26For this project, a pulse-echo technique has been used whereby an ultrasonic pulse signal is transmitted into the test object and the same transducer receives the reflected signal pulses or echoes that are reflected from the back of the test object and all interfaces inside it. The transducers used were two TMP3 contact transducers one with a centre frequency of 10MHz and the other of 20MHz together with a Desktop Ultrasonic Instrument (DUI) (NDT Solutions Ltd.) which acts as the pulser, receiver and digitizer. All the experimental data has been processed in Matlab using a generic pc to give A-scan and C-scan representations. Standard 3-in-1 oil has been used as the couplant between the transducer and the test object throughout. In order to increase the signal-to-noise ratio averaging has been used at the point of data capture, 1000 times in all cases. For the measurements on the scissors the experimental data was processed further in the frequency domain. The Fourier analysis was performed in Matlab and the ultrasonic propagation modelling was carried out using in-house software PROPMAT.13 In order to compare the ultrasonic results X-ray measurements were also performed by Vanessa Fell at the Centre for Archaeology, English Heritage Fort Cumberland, Portsmouth, UK.

27To test the projected use of non-destructive evaluation using ultrasound, a relatively simple structure was used. This was an early 20th century plane iron from the Hawley Collection held at University of Sheffield, the aim being to determine the internal structure and to calculate the thickness of the layers across the plane iron. The effective cutting edge of the plane iron was of steel which had been brazed on to the underside of the iron body of the tool, using a layer of brass. The lower, front part of the body had been milled to remove a section to take the piece of steel. The advantage of using this plane iron was that the layers were clearly visible along the edges, allowing for actual thickness measurements and hence technique validation (fig. 2).

28The second sample considered was a pair of 17th century scissors. These were evaluated in order to determine the thickness of the steel layer on the inside faces of the blades. The scissor handles, shanks and blade backs are made of iron and it was possible to sample part of the scissors which was iron only (fig. 1).

Results

The three-layer plane iron

29An initial X-ray study was carried out on the 3-layer plane iron (as described above), together with a photograph of the blade section of the plane iron. The three layers can clearly be seen in the photograph but the X-ray results gave no indication of the number or even presence of the individual layers. For this particular sample, no useful structural information could be obtained from the X-ray data.

30Pulse-echo ultrasound measurements were then performed on the sample. Figure 3 shows typical A-scan data for a measurement through the iron only section (top) and a measurement through the three-layer section (bottom). The y-axes in both cases have been cropped due to the large contrast in amplitude between the first pulse (a reflection between the transducer and top surface of the object) and the first echo from the back of the object at the metal-air interface. In the bottom graph showing the A-scan through the three-layer section, two additional peaks are observed. These are due to reflections from (first) the iron-brass interface and then (second) the brass-steel interface. Theoretically other reflections will occur, for example, multiple reverberations in the brass middle layer, however, the amplitude of these reflected signals will be so small as to be unmeasurable. In both A-scans a certain amount of noise can be observed, this is due to both system noise and some scattering from the metals. The noise has been significantly reduced by averaging, without which, it would not be possible to distinguish the reflections from the internal interfaces in the three-layer section. However, although small in amplitude the internal structure has been detected using the ultrasound technique and is clearly illustrated in the experimental data.

31From thickness measurements and the A-scan data for the iron section the average time of flight velocity for the iron has been calculated.

32Using the iron velocity and a standard value of velocity for brass (taken from the literature) and the overall thickness of the three-layer section, a value of time of flight velocity for the steel section was also determined. Further A-scan measurements were then taken in a grid formation across the entire three-layer section of the plane iron and the thickness of each individual layer calculated at each point.

33The layer thickness information was then transferred to C-scan format to give colour maps of the variation in the structure of the plane iron (tab. 4). The three C-scan images are shown in table 4. It can be seen that the layer thicknesses vary across the plane iron. The top iron layer varies in thickness from 1.82mm to 2.42mm, the thickest areas tending to be in the centre. The bottom steel layer varied only between 1.78mm to 2.15mm. Here, however, the thickest areas are at the right and left hand side close to the solid iron back. The “step” in the iron body of the plane iron was cut with a milling cutter resulting in an uneven internal surface, which is seen in the thickness variations. The brass flux forming the bond between the two ferrous metals, also shows the variations in thicknesses, from 0.46mm to 1.1mm.

16th-century scissors

34The second metal artefact considered in this project was a pair of 16th century scissors. Measurements were made along one blade using the ultrasonic pulse-echo technique described above, this time using a 20MHz centre frequency transducer. Unlike the three-layer plane iron the scissors are slightly more complicated to test because the blades are comparatively thin. The pulse signals will reflect at each interface within the structure and from the back face in the same way as for the plane iron but since there will be a lower level of attenuation overall (the distance travelled being so much less) so multiple reverberations will be observed. In many situations the presence of multiple reverberations is not necessarily a problem provided that they can be separated in time, in other words, the time for the pulse to travel there and back through the layer is less than the pulse width. For the scissors, however, this is not the case and the multiple reverberations overlap each other in time causing constructive and destructive interference and hence the A-scans cannot be interpreted in the time domain easily, if at all. In this case any reflection from the interface between the iron and steel within the blade is completely masked by the multiple reverberations of pulses travelling between the transducer and the back of the blade-air interface. There are a number of possible solutions to this problem. The first is to analyse the data in the frequency domain using Fourier Analysis techniques. A digital filter can be set up to remove the masking reverberations from the A-scan data leaving only the iron-steel interface reflections. This has been carried out with some success. A second method is to use an ultrasound wave propagation model to simulate the propagation through the blade and by iteratively changing the layer thickness parameters and comparing with the experimental data until good agreement between the two is found.14 This has also been carried out on the scissor blade A-scan data to find the thickness of the steel layer in the blade. Both the filter and Propmat model techniques gave similar results the average of which are given in table 5. It is clear that the steel thickness varies gradually over the length of the blade as would be expected, narrowing towards the tip. The agreement in the results of the two techniques gives confidence in these thickness measurements but further work needs to be carried out on samples where the thicknesses are known a priori to complete the validation.

Summary

35In the ultrasonic part of this paper, we have shown the potential usefulness of applying established ultrasonic nondestructive evaluation techniques that have been developed in the engineering industries to the analysis and characterization of found metal artefacts. Ultrasonic NDE is a fast moving field being driven by the need for quality assurance and process monitoring in the manufacture of safety critical components and structures. Although a very small amount of work has been done previously using ultrasound in archaeology it seems sensible to revisit this field in order to exploit the advances in technology that have been made in recent years. Many of the issues such as highly scattering materials, thin layers and complex structures which have discouraged archaeologists from using ultrasound properly are increasingly being solved for industrial situations.

36The data presented here shows positive first steps in the application of ultrasound techniques to find structural information in found metal objects which might well be difficult to determine using conventional techniques and certainly not without destroying the artefact itself.

Conclusions

37This paper aims to show how the diverse fields of history and electronic engineering can be combined to interpret archival records and ferrous artefacts. Documentary sources and surviving historical artefacts provide evidence that Sheffield craftsmen welded steel to the cutting edges of knives, scissors and tools, resulting from the relative costs of iron and steel. The important resource of the Company of Cutlers archive, and especially in this case, the Storehouse records, demonstrates the complexity of cutlery manufacture and marketing.

38Interpreting the structure of post-mediaeval artefacts relies on a number of methods, each of which can provide different information about their characteristics. Ultrasonic non-destructive evaluation and simulations can provide additional information about the number and thickness of the layers of metal. The non-destructive method of X-rays has been shown here to give some indication of the structure of the three-layer plane iron, in that it identified differences at the cutting edge, but could not identify the number of layers, their thicknesses and the quality of the interface.

39The method of non-destructive evaluation of antique metal artefacts presented here has been shown to add information to the results of other investigative methods. In order to demonstrate the value of exploring this method of testing and of the simulation of probable results, the samples used here were of known composition, having a reasonably smooth surface, and it is acknowledged that many archaeological metal artefacts will not. It is also acknowledged that the use of 3-in-1 machine oil as a couplant may be banned for use with some artefacts. However, this method has been shown to be of value in determining the manufacturing methods, confirming the documentary sources and allowing for a qualitative analysis of manufacturers’ practices.