Dependable Red Hot Action

We present a brief observational, 'ethnographic', study of the Roughing Mill in a steel plant and use material from recorded activities to provide ‘illustrative vignettes’ of some aspects of the accomplishment and problems of everyday work. The account provides a 'bottom up' method for developing a more sophisticated and situated view of the problems of dependability. The paper documents the social organisation of work in the Roughing Mill, the interaction between the computer scheduler and the skill of the mill operator in accomplishing 'dependable' production of steel plates from slabs. Introduction: dependability and socio-technical systems. "Dependability is defined as that property of a computer system such that reliance can justifiably be placed on the service it delivers." (Randell 2000) As computer-based systems embracing humans, computers and engineered systemsbecome more complex and organisationally embedded, so the challenges of dependability of building systems involving complex interactions amongst computers and humans -increase. In these systems failure, lack of dependability, can result in financial or human loss and, consequently, improved means of specifying, designing, assessing, deploying and maintaining complex computer-based systems would seem of crucial importance. Much of the work on dependability has necessarily, and naturally, focused on massive, extraordinary, public failures such as the London Ambulance Service failure of 1992, the space shuttle catastrophe of 1986, or the Ladbroke Gove train disaster of 1999. This paper is, however, concerned with rather more ordinary, everyday instances of dependability and failure. Instances of undependability in many settings are not normally catastrophic but are rather mundane events that occasion situated practical (as opposed to legal) inquiry and repair. Dependability can then be seen as being the outcome of peoples’ everyday, coordinated, practical actions. Workers draw on more or less dependable artefacts and structures as a resources for their work of achieving overall dependable results in the work they are doing (Voß et al. 2002, Clarke et al. 2002). In this paper we wish to explicate how overall dependability is practically achieved in the operations of a steel rolling mill, a rather different setting than most studies of dependability in IT systems have looked at. Here the research is not situated in bright, clean offices of the services industries or the technologically advanced and safety critical sectors of the nuclear industry or aircraft safety, but in the noisy, dirty and dangerous world of steel manufacture. Although the focus of activity is transforming a slab of red hot steel into steel plate rather than, for example, the provision and control of information, similar dependability issues of timeliness, responsibility, security etc can arise, and need to be resolved, in the interaction between computer systems and human skills. Our research consists of a brief observational, 'ethnographic', study (Hughes et al 1992; 1994) of the Roughing Mill in a steel plant. Although ‘quick and dirty’ the fieldwork covered all 3 daily working shifts and a number of roughing mill operators of varying levels of skill and experience. In the paper we offer ‘illustrative vignettes’ of aspects of this particular work in the Roughing Mill as an example of a more ‘bottom up’ method for developing a richer situated view of the practical problems of dependability (Suchman, 1995). The paper provides us with an opportunity to respecify the problem of dependability, and hence the lessons for IT systems design, by documenting ‘real world, real time’ practices whereby dependability is rooted within the practical ongoing social organisation of work. Our argument is hardly radical in emphasising the point that any abstract ‘rules for dependability’ – such as procedures, models, proscriptions, prescriptions, etc. – have to be applied within the context of some socially organised work setting in which those who have to apply such rules have to deal with all the contingencies and other demands on their attention and effort. What is perhaps more radical is the intention to treat this observation seriously as both a research endeavour and as an input to system design. Our interest in the social organisation of work is in how the work activities (which are often glossed and idealised) are actually carried out and accomplished as day-to-day activities with whatever resources, including technological, are to hand and facing up to whatever contingencies arise. As far as system design is concerned – and as we have said elsewhere – such an interest seeks to understand the work as a first priority so that any innovation in system design better resonates with the work as actually done. This point of view is based on two suppositions: first, that most systems fail because they do not resonate with the work as it is actually done as a ‘real world, real time’ phenomenon; second, that even when the intention is to change the work (to make it more efficient, reliable, etc.) it is always best to have a good idea of what may be lost in doing so to put against any putative gains. The research reported upon in this paper was motivated by several observed ‘problems’ in the Roughing Mill, some of which may be viewed as relevant to issues of dependability. The dependability issues were manifested in the complex interrelationship of skill, teamworking, and awareness that could result, for example, in a range of ‘troubles’. These included: • ‘Cobbles’ or 'turn-up' of the part rolled slab that makes it difficult, and sometimes impossible, to manipulate the slab through the Mill • Badly shaped slabs coming into the Mill that produce, for example, ‘fishtails’ or other defects in the finished slab. • Slab defects produced by the furnace, for example, 'thermic shock' requiring the Mill operator to make adjustments in how the slab is rolled and that may mean the final rolled plate will not yield all the ordered plates. • Various kinds of marking etc. on the slab produced by difficulties in rolling that may influence the quality of the final plate • A variety of computer problems related to the identification, measurement and sequencing of the slabs. Picture 1: Problems getting 'turn-up'. Picture 2: A 'cobble' being lifted from the line. As in any tightly structured sequence of interdependent activities such ‘troubles’, even though they are often regular and routine, are ‘troubles’ which detract from the dependability of the system by producing waste, slowing production, creating frustration and increasing overall costs. However, and again as with most systems of high interdependency, achieving ‘smooth’ operation day in and day out is extremely difficult and requires a great deal of experienced skill on the part of the operators of the technology. The analysis that follows uses a framework of 'sensitising' concepts (Blumer, 1954; Blythin et al 1997) that have been developed over the years in doing ethnography as a contribution to system design. It provides a means of bringing out the grounding of dependability on the social organised skill and competences of those involved in the work setting. These concepts are distributed coordination, the situated orientation to plans and procedures, and the achievement of various forms of awareness of work. To begin, we describe the work of the Rolling Mill and the rolling process in a rather idealised fashion. In this way we can begin to bring out the situated actual activities done by the operators as routines of their daily work. The roughing process Like many unfamiliar work activities, the process of rolling a slab of steel appears complicated beyond belief. Ideally, however, it is simple enough. Slabs, or blocks of steel, are rolled into steel plates of varying thickness. The process begins with the available steel slabs being assembled in the slabyard and moved to furnaces. Usually more than one steel plate will be made from each rolled slab so ‘as rolled’ plated will be cut on one of the shear lines. ‘Build rules’ are used to determine how many plates of required sizes can be made from standard slab sizes with minimum wastage. There are two furnaces each with two ‘strands’ of slabs passing through them. The temperature for each slab is calculated and passed to the furnace controller. The slabs are heated to around 1250°C. Each slabs temperature is repeatedly calculated as it is heated and when it has reached the required value is flagged as ready to roll. A another slab is then pushed into the furnace so moving the strand one along with the slab ready to be rolled dropping out of the discharge end. The mill – really the Roughing and the Finishing Mill – use reversing mills or rollers. The incoming slab will have already been specified, and displayed to the operator, as to be rolled in one of two orientations: length to width or length to length. However, this requirement is not imperative and is sometimes overridden. If, for example, there are flaws in the flaws in the slab which make it difficult to follow the requirement. The slab is reduced in thickness by a series of ‘passes’ back and forth through the mill until the desired thickness is reached. The Roughing Mill ‘stand’ (where our observations were concentrated) is a large structure which supports two steel rollers turned by two large electric motors. The distance between the rolls – the ‘roll gap’ – is adjusted by the ‘screws’. Slabs are transported on roller tables that are controlled in sections to give more delicate control over the movement of the slab. The slab can be turned on the ‘turning table’ which consists of alternate rolles, thinned on alternate sides, and which can rotate in opposite directions. Moving side guides are used to square up and centralise the slab for passage through the rollers. The thickness of the slab can at any stage be inferred from the screw position the last time the slab passed through the roll gap. This process is typically fraught with problems since the whole mill is significantly elastic under the forces generated by rolling along with the fact that the rolls expand as they heat and wear as more steel is rolled. The process of ‘zeroing’ the mill – adjusting it so that the unloaded roll gap is actually zero when the indicator says it is – is difficult to carry out and often poorly understood The length and width of the slan can be measured by an optical gauge known as the ‘Kelk’ or ‘Accuplan’ but only when it is held still on the turning table. At this stage it is important to achieve the final plate width and the keep the slab – by now almost a plate – as close to the ideal rectangle as possible. The computer calculates the sequence of screw settings and turns. The screws are reset automatically after each pass. The computer requires a width reading when necessary and corrects to achieve an acceptable width and keeps track of what has been rolled. The operator is responsible for manipulating the slab to turn it and enter it through the rollers centrally and squarely after the screws have been automatically set. This involves hand and foot controls. If necessary the operator can also take control of the screws. At each ‘pass’ through the mill the steel is reduced in thickness by the ‘draft’. As the volume remains the same, the other dimensions must increase. Most of this increase appears as extra length in the rolling direction and is quite easy to predict. (There is also some ‘spread’ outwards as it passes through the rollers. This may be large in terms of product tolerances but is always small in relation to the elongation of the slab. However, it is difficult to predict.) The first target is to elongate one of the slab’s dimensions until it reaches the width of the final product. It may be rolled in both orientations until this is achieved. It is then turned through 900 and rolled in the same orientation from then on. The drafts from then on will, ideally, be the maximum possible in order to reduce rolling time and minimise heat loss. Different limits apply at different parts of the process. Although there are variations according to the composition and quality of the slab, the general procedure in the Roughing Mill is as follows. The slab is pushed from the furnace, through the wash boxes to remove scale and then aligned and centred on the rollers. Information on the monitor in the ‘pulpit’ – the control room where the operator works – tells the operator the slab quality, its present width and length, the width and length required, the orientation, the ‘turning point’ (the measured point at which the operator should turn the slab to roll for final length), and the ‘finish point’ (the point at which the operator should send the slab through to the Finishing Mill). The rolling operation itself begins with pre-broadside passes through the Mill. The slab is also sprayed to remove scale. The operator then ‘goes for width’ by rolling the slab to produce the desired width up to the ‘turning point’. Measurement of the slab through the Accuplan is displayed in the ‘pulpit’. One red light indicates that measuring is taking place, two that the slab has achieved width. Green lights are displayed for the operator to turn the slab to roll for length. As the operator puts the slab through the mill he turns and aligns it (Picture 3). The scheduler reduces the gauge at each pass – displayed on the overhead monitor and the ‘clock’) until finish point is achieved. The final pass is a reverse pass. The rollers are then lifted and the plate sprayed on its way to the Finishing Mill. Picture 3: Aligning the Slab the mill lights are green Of course, in actuality the process rarely goes as smoothly as this. ‘Troubles’ of various kinds are a regular feature. One prominent trouble is when the part rolled slab ‘turns up’ to form a U or W shape that makes it impossible to manipulate. There are a number of techniques, all involving heavy manual labour, to recover from such events, but they cause delays and do not always succeed. Although the process is not fully understood, the cure is straightforward. The screw settings should ensure heavy drafting at critical points in the process but this requires considerable experience on the part of the controller. Indeed, many of the more experienced operators will go into manual mode for the last few passes. A related problem is when the plan view of the plate is not the ideal rectangle. If the problem is severe the final rolled plate will not yield all the plates required. The operator does have a degree of control over this but the automatic controller gives no help. Sometimes the length or the width do not turn out as planned and further action is necessary. In the Pulpit the operator has various monitors and controls. On his left is the furnace monitor (in this case misaligned after a mishap with a crane), load measures, mill light, screw inject (this can also be done through the central control pad) and levers for screwing the rollers up and down (Picture 5). To the right of the monitor is the main control pad, the main monitor, rack lever, amp meter, the monitor for sprays, and temperature gauges (Picture 4). To the front of the operator are a number of foot pedals for turning the lab on the rollers and sending it through to the Finishing Mill. There is also a head level display that provides the reference points for the slab currently being rolled. Outside, on the Mill itself, the mill ‘clock’ and measuring lights provide further information. Indeed, the mill ‘clock’ – the way in which the ‘hands’ alter to reflect the changing of the screws – was a clear and persistent focus as the operator worked. Picture 2: Pulpit Controls: Right Side Picture 3: Pulpit Controls" Left Side Some concepts for analysis We now want to move onto some analysis of the fieldwork materials using the presentation framework developed out of previous ethnographic studies in a variety of domains. It is important to stress that the framework of concepts is in no way a theory of work, organisation or whatever. The rationale for this insistence on the framework not being a theory would take us too far afield, but, briefly, is connected to Garfinkel’s critique of constructivist sociology. It is in an important part a heuristic and practical device for bringing out the generic everyday features of socially organised work settings and, at the same time, presenting these in a form useful for designers. In this particular instance we are interested in whether the heuristic of the framework facilitates the identification of dependability issues. The features we want to illustrate in this paper are as follows: ‘distributed coordination’, ‘plans and procedures’ and ‘awareness of work’. Distributed coordination This points to how work tasks are performed as coordinated activities, that is, as activities which have interdependence with activities done by others who may not be co-located. It is clearly a notion closely tied to the idea of a division of labour but goes further in emphasising the ubiquity of the day-to-day need to achieve coordination within a division of labour. Depending on the work setting and the activities concerned, distributed coordination can take many forms, involve varied technologies and operate at different periodicities. As with the other concepts that of ‘distributed coordination’ is a methodological injunction to treat work settings, persons and activities as embedded in an organised ensemble. The activities and the persons who perform them are interconnected as part of some organisation of activities and persons which has to be coordinated in order to ‘get the work done’. Plans and procedures ‘Plans and procedures’ refers to one of the more obvious means by which distributed coordination is achieved and supported. Project plans and schedules, manuals of instructions, procedures, workflow diagrams are all ways of enabling persons to use as resources for coordinating work activities. There is no implication here that any particular set of plans, etc., is successful at coordination, or conforms to some ideal standard. The explicit point of plans is to coordinate the work of different persons so that separate work activities, either in parallel or serially, have a coherence and, typically, through this meet other goals such as efficiency, meeting time constraints, beating the enemy, growing the company, and so on (Schmidt, 1997). Although ‘plans and procedures’ are, of course, about coordination – and often an important resource in its achievement – ‘plans’ are abstract construction that require implementation within the specifics of the circumstances in which it is to be followed (Suchman, 1987; Dant and Francis, 1998). The accomplishment of a ‘plan’ is dependent upon the practical understanding of what the plan specifies in these circumstances, using these resources, and facing up to these contingencies. In many cases of ‘real time, real world work’ accomplishing the plan often involves using local knowledge, ‘cutting corners’, ‘bending the rules’, even revising the plan in order to meet its overall objective.