VI. CATHODE FAILURES INTRODUCTION

Transcription

1 VI. CATHODE FAILURES INTRODUCTION Ideally, a pot should be kept operative for the time it takes, by the normal chemical and abrasive forces, to wear the bottom carbon lining evenly down to the height of the current collector bars. By normal standards this may take ten years or more (see Chapter III). This is, however, most often not the case. What is regarded as an acceptable pot life will usually vary from plant to plant. It will depend on how much effort is put into the training of the repair crew, the understanding of the chemical and physical processes that take place in the cathode, the design and condition of the steel shell, the lining construction, the choice of lining materials, and the cell preheat, startup and operational practices. These last points are probably the most important. Without sound preheat, startup and operational practices no cathode lining construction can expect to obtain a long life. Earlier, long pot life was not considered a critical parameter as long as it was above 1200 days. To-day pot life less than 2000 days is not considered acceptable, and the goal appears now to have moved to about 3000 days. Although there are smelters that have average pot life considerably shorter than 2000 days, there are also some smelters that obtain close to or even above 3000 days. All smelters make a considerable effort to increase pot life. The reasons are use of more expensive lining materials and the realization that a proper constructed and operated cell really can last up to 10 years. For some even now, and for many more in the future, the cost associated with disposal of the used materials, which for an increasing part of the industry has to be recycled or processed to inert landfill, may turn out to be the driving force, see Chapter VII. It is not always the target to obtain maximum possible pot life for all pots in a smelter. After startup of a greenfield project, a new line or a retrofit of an existing smelter it might be necessary to prematurely cut out pots in order to level the workload on the pot relining department. When a new line is cut in, a large number of pots are started within an interval of a few months. If a significant number of early failures are avoided, these cells will all fail by the usual causes some years in the future and within a limited time frame. Due to a natural failure age distribution the failure period will be broader than the

2 startup distribution but nevertheless may represent a transient workload the pot relining department is neither built for nor manned to take on. This is illustrated in the example given in Figure VI-1, which shows the original startup distribution and first failure distribution after a line was retrofit from Søderberg to prebake anode operation. The startup distribution is broader here than it normally will be during the startup of a greenfield project or the addition of a new line, since the pots were retrofitted group by group while simultaneously keeping the production up on non-retrofitted groups [1]. There was no effort to level the cutout rate. The normal answer to this is to start a planned cutout and relining of pots before they fail by themselves, preferably pick the poorest producers/worst pots in the line. Done right this will flatten and broaden the first cell reline distribution and make the relining rate manageable without the need to temporarily assign and train a large reline labour force. # of pots lined Figure VI-1. Original startup and the subsequent failure distributions (following a retrofit from Søderberg to prebaked anode cells in one potline) [2] A perfectly constructed and operated cell is not often a straight-forward task. Cryolite, aluminium, air and CO 2 at 960 C is a very corrosive system and there are several deterioration mechanisms. The whole cell may have to be shut down just because one critical link fails. The problem of failure is often complex and it is not uncommon that projected improvements has had quite the opposite effect, or that relined cells perform more poorly than earlier cells. The high amperage cells that are used in modern potlines, also tend to be much less forgiving than the older low amperage cells. The right mixture of knowledge, careful analysis, creativity and conservatism is needed to get the best results. Smelters often experience that later generations of pots obtain a less satisfactory pot life than the first generation, even with same design, same materials and same relining procedures. There may be several reasons for that, one important and often overlooked is a change in the operational framework for the smelter. Important parameters in this respect are current

3 creep (and anode size increase), pot control algorithms, changes in bath chemistry control/target and environmental regulations. The direct cause for the ultimate failure may have been triggered a long time before the cell started to show any obvious symptoms of trouble or a sudden unexpected tapout took place. A necessary and indispensable tool, both to arrive at conclusions concerning why something went wrong and to make new potlining decisions, is a full or partial autopsy of the cut-out cathode. Only in this way it is possible to account for the sequence of incidents that led to the final failure and to make the right decisions in respect to design, workmanship, materials, start-up and operational procedure. A wrong diagnosis may have serious and far-reaching consequences. Symptoms, but not the causes, of shortened life are: High initial Feconcentration in aluminium. Uneven cathodic current distribution. Uneven bottom temperature. Operational difficulties. High noise. Increased cathodic voltage drop. CATHODE AUTOPSIES As several authors have pointed out [3,4,5,6,7], a cathode autopsy can be an invaluable tool to determine reasons for pot failure, to diagnose lining problems, to improve performance, to prepare for load increase or to find out what really takes place with respect to chemistry and physics of lining materials in operating cells. Cathode autopsies mean digging out temporary or permanently shut-down cells to document their status, most often in an effort to find the reason for a pot failure. Autopsies on shut-down cells that operate satisfactorily are used sparingly, as a reline is expensive in the order of several hundred thousand USD for large cells. It is, however, sometimes necessary in order to ascertain that a new cell design or operational practice function as planned before too much money and effort are put into a new design. Since the service time for even poor cell designs can be counted in years, it can be very risky to install a large number of non-tested new cells, new or modified lining designs or new materials without proper feed-back on how it will function in an industrial environment. A point-by-point autopsy procedure is given below and described more in detail later. However, the conditions on site and the ultimate reason for the autopsy may make it necessary to skip one or several of these items.

4 Collect all relevant information o Shell history o Lining design o Materials specifications, o Preheating records o Operational data throughout pot life. Inspect pot shell and pot stall. If failed by tapout record where. Document by photos, sketches and measurements o Deformations (width, length) at several places on steel shell o Measure diagonals o Steel bottom deformation (outside) o Outward bending of sides/ends (cradles) o Bending and extension of collector bars (out, down, sideways) o Inspect outside collector bar seals o Note any cracks, hot spots, local deformation of shell o Compare with new or relined dimensions Remove bath and metal pad but take care to leave bottom ledge and side ledge at bath/metal interface intact o Remove all loose debris and vacuum surface to reveal details. Do not use compressed air for surface cleaning. o Inspect cathode surface. Photograph, sketch and measure all anomalies (cracks, holes, metal plugs, excessive wear) o Measure minimum ledge thickness around cavity o Document the extent of bottom ledge and muck Carefully remove side and bottom ledge, large muck deposits o Document state of sidewalls, endwalls o Inspect rammed bevel o Look for infiltration in peripheral joint o Measure cavity (with reference to deckplate or other fixed level) o Measure bottom block lengths where block/ring joint intersections are visible o If metal infiltration through ring joint is suspected, it may be necessary to chip away part of the rammed bevel and a centimeter or two of the block surface at the block/joint intersection in order to get a good view of metal or carbide filled gaps. Try to decide at this stage whether it is of interest to patch the pot or not o If pot is to be patched, do minimum damage to bottom blocks other than remove or patch failed block(s) If possible avoid the use of heavy machinery Check if damaged part can be removed by hand-held jackhammer or by core drilling If autopsy is to proceed start by removing a couple of bottom blocks away from suspected failure area

5 o Heavy machinery can be used o Where to start needs to be an educated guess since the first blocks that are removed are likely to be crushed. Should cause of failure be located in this area the evidence will likely be lost o Remove cathode blocks, side lining and underlying refractory and insulation to make a trench down to the steel bottom o Try to make the two cross section cuts as clean as possible (block sides, narrow joints) o Remove debris and vacuum surfaces to reveal details. o Record state of collector bars taken out Corrosion Bending Dog legs, deformation Inspect materials being removed for anything out of ordinary o Document by photos, sketch and measurements the overall view as well as details Length of bottom blocks Remaining height of bottom blocks Bending of bottom blocks Cracks in blocks Swelling of rammed parts Signs of infiltration State of refractory/insulation in lower sides Bottom heave Conversion and swelling of bottom refracttory/insulation Continue to carefully remove block by block towards suspected failure area o Continue documentation listed above o Record remaining block height above collector bar when that is possible o When failure area is reached document all relevant features What happen and why? When? Write autopsy report o Include all relevant information with respect to lining design, materials, operation and autopsy o Give a clear conclusion if possible Likely cause of failure can go beyond the apparent The problem may be very complex and one autopsy may not be enough to solve it o Recommend actions for improvement Short term Minor operational details, workmanship, materials, minor lining details Long term

6 Education and training, preheating and start, construction, pot operation strategy Future cathode linings for higher load The work is not over until appropriate actions have been taken o Discuss findings with potroom operators, relining crew o If necessary change relevant standard operational practice(s) The first stage in an autopsy is to collect all relevant historical and operational data. This information alone may give clues to the failure. A proper autopsy is performed after the cell has cooled down without watering and after the cathode pane has been cleaned as well as possible. Dimensions are then measured carefully to detect and locate heaving, cracks, displacements and expansions of blocks and potshell (Figure VI-2). Cracks that have been open during operation are often visible by the yellow aluminium carbide in them or by metal stuck in the crack. Figure VI-2. Cavity measurements in the cleaned pot. It is important to start the autopsy at a spot where one expects a minimum of information to be lost. However, if a repair is performed, one should try to start at the exact failure location and try to restrict the digging to the actual damage area. Heavy equipment that may damage neighbouring areas should not be used during a repair or a partial autopsy if the pot is supposed to be put back into operation again. In that situation only hand-held pneumatic hammers or core drilling equipment should be used. Using air cooled core drills is probably the most gentle way of removing parts of bottom blocks for either inspection or repair (Figure VI-3). During full autopsies, where eventually the entire lining will be broken up and removed, the use of heavier demolition machinery is recommended. As the blocks often can be very hard, a heavy back-hoe-mounted hydraulic hammer is advantageous (Figure VI-4a). The debris is scooped out by the back-hoe (Figure VI-4b).

7 c) Collector bar current uptake (ka) Collector bar no Collector bar current uptake (ka) Figure VI-3. Example of a) partial autopsy using core drilling equipment to find the reason why two bottom blocks did not carry load and b) during later repair (only some sidewall work and ramming remain before pot can be restarted). c) Cathodic current distribution profiles 5 days before repair (left, age 100 days) and 1233 days after repair (right, age 1338 days) [8]. Enough material is removed in the initial stage to make a ditch wide enough to inspect a cross-section of the lining down to a certain level. If a back-hoe is used to take out the broken lining material, enough carbon blocks are removed to make room for the scoop. A cleaned ditch ready for inspection during the first stage of an autopsy is shown in Figure VI-5. Colour photos and measurements (and lining samples, if necessary) are taken as the digging proceeds. It is not only looked for the final failure as careful observations may give valuable additional information. With the use of the right equipment a full autopsy may usually take three to five days, but the time may be shortened if the failure is obvious or if only limited information is sought. The report should contain the historical data, autopsy observations and a justified conclusion of the reason for the initial failure. Samples taken can be tested at realistic lining temperatures (electrical resistivity, thermal conductivity) and the data can be used in more realistic computer models for

8 Figure VI-4. The use of heavy mechanized equipment during pot autopsy. a) Hydraulic hammer mounted on back-hoe to break open the carbon pane; b) The back-hoe scoop will remove most of the debris. pot operation. In order to get a full benefit from a cathode autopsy one needs to have experience from several autopsies and a sound knowledge of material properties, thermo-mechanical effects and chemical changes. It can be beneficial to include material suppliers. Too often they seem to lack a proper understanding of the rigors their products experience in an industrial cell. It may also be important to involve the pot repair crew and pot operators during parts of the autopsy. Demonstration of weaknesses in lining construction, workmanship or operation is often the most effective way to improve the lining and operation practice. Figure VI-5. A first ditch cut through a cathode to be autopsied, cleaned and ready for inspection. Recommended future action in order to avoid a similar type of failure is the most important result from an autopsy. If the damage is very extensive or the pot is very old it may, however, be difficult to draw a conclusion with certainty. A survey of cathode failure mechanisms, with special emphasis on early failures in the carbonaceous part of the lining, is listed in a paper by Sørlie et al. [3].

9 A new lining design needs to be thoroughly tested before it is approved, since the financial implications can be huge if a large amount of cathodes are lined with a design that later proves not to satisfy the expectations. In the absence of early catastrophic incidents, failures and sign of failures may have a timeline counted in years. To speed up the approval of a new cathode construction or lining design and improve the probability of success, a good pot autopsy may be necessary. As the term implies, a test pot with no sign of abnormal behaviour is chosen for cut-out and autopsied. The age is usually 1 or 2 years. Depending on the situation and what is of primary concern, a full or partial autopsy (with subsequent repair) may be undertaken. This may reveal weaknesses in construction, lining design and/or materials that not yet have developed into full-fledged failures and which may be impossible to spot otherwise. If only specific parts of the lining are of interest, it may only be necessary to stop the pot for a few days, take some core drill samples, patch it and restart it again. Figure VI-6 show examples of such, where the conditions of the bottom refractory layer in test pots were of interest. After partial cleaning of the bottom, one core was extracted near the middle of each pot and one closer to the side. Afterwards the holes were plugged with crushed refractory bricks and ramming paste and the pots restarted. Figure VI-6a shows a core drilled through a chamotte-type refractory barrier layer [9]. A high-viscous glassy phase has been formed by reaction with percolating bath components that have protected the underlying refractory, which condition was as installed. The critical isotherm for reaction of bath components with the refractory is well below the refractory barrier seen in the photo. Figure VI-6b shows a core taken from a non-chamotte refractory barrier type that is unable to convert to glassy products by reaction with bath components [10] and where the penetration seems to have reached the liquidus isotherm. Figure VI-6. Cores of refractory barrier layer from two test pots. a) Test of new type chamotte brick [9], with glassy barrier formed in situ (pot investigated and patched at 300 days). b) Test of another type bricks without glass forming ability [10]; where bath components have penetrated entire barrier depth (pot investigated and patched at 516 days).

10 UNEVEN HEATING AND THERMAL SHOCK The cathode carbon lining used in all aluminium cells today is not impenetrable to the electrolyte. It will always have an open porosity through which bath components can percolate. Liquid metal, however, does normally not penetrate this porous structure. The various carbon qualities that are used in industrial cells, with different porosity, permeability, structure, etc., may only influence the rate of percolation, not the end result [11]. Since bath penetration in the cathode is normal, a proper lining is designed in such a way that part of the refractory materials underneath the carbon bottom blocks will have time to absorb and react with the penetrating fluorides. The reactions convert the materials into a denser, often glass-like, penetration barrier that stops or severely retards further percolation. Such reactions should run evenly and slowly after start-up of the cell. It is then of utmost importance to avoid larger cracks or open gaps in the lining, where uncontrolled amounts of bath and metal may drain in a short time. Such early crack or gap formation in the carbon lining is usually caused by thermal shock or thermal movement of lining materials during preheat and/or start-up of the cell. While some cracks may lead to almost instant failures, others may rapidly be partly or fully sealed by frozen bath, chemical reactions, thermal movements or carbon swelling. A local weakness in the carbon lining will remain, however, which can reopen at a later stage in the pot life or initiate a pothole formation (see later). The thermal gradients that are introduced in the carbon bottom pane during some preheating procedures are a major contribution to thermal shock cracking of the bottom blocks. Other important parameters are material properties associated with the structure of the carbon, such as coefficient of thermal expansion, thermal conductivity and strength parameters. The gentlest possible preheat of the cathode is probably achieved with gas burners (see Chapter II, Flame preheating). The temperature increases and the temperature distribution in the carbon bottom pane can be closely monitored and controlled and no cracking or displacement of the bottom blocks is likely if flame preheating is performed satisfactorily. The philosophy regarding the final preheat temperature setpoint may differ between smelters. The earlier disagreement between different companies regarding end temperature and time [12] seems to have been settled, in large part due to the demonstration of a low and controlled voltage during startup and early operation [2], see Chapter II. The best seems to be an end temperature near normal operating temperature (940 C±20 C) reached in about 3 days. Shorter preheat times using gas will not heat parts of the lining thoroughly enough, which may result in some increased voltage transients after startup. Resistor baking of pots, using a coke bed as resistor material between anode and cathode, is probably the most common preheat method, especially in prebaked anode cells. If poorly executed this method can, however, subject

11 the bottom blocks to severe thermal shocks, with temperatures up to 1500 C and thermal gradients of 1100 C/m at the carbon surface of the cathode [13]. Poor anodic current distribution during preheat may often be seen directly as red studs or anode cracking, but the condition of the underlying bottom block can be equally poor, although often not spotted due to the coke cover. Such preheat cracks (Figure VI-7) in one or more blocks are probably more a rule than a freak accident if cathodes with amorphous bottom blocks are resistor baked on full line load. Shunting (Figures II-10 and II-11) or replacement of amorphous block qualities with more graphitic ones will reduce or remove this problem. Although one or a few blocks may crack due to high current, extensive cracking during preheating is most often due to unsatisfactory carbon quality. Figure VI-8a shows a photograph of a cathode surface in which every single bottom block had experienced two preheating cracks, seen as two parallel lines running down the length of the cathode. High current and rapid heating in some blocks mean that other cathode blocks have much lower current densities than average. Normal quality blocks with lower-than-average current densities should, however, not crack. A cracking pattern as shown in Figure VI-8a is hence caused by inadequate thermal shock resistance of the entire lot of bottom blocks. The distance between the parallel cracks was measured to cm, which corresponds almost exactly to the inner stubstub distance of the anodes in the cell (Figure VI-8b) and is in agreement with the observation that the major part of the current is passing through the inner anode stubs during the initial preheating period. The thermal gradients in the bottom blocks makes them bend upward, thus shortening the distance and compressing the coke bed (increased the contact pressure) more toward the central channel than near the side channel. Figure VI-7. Preheat cracks (in bottom blocks no. 5 and 9) in a failed cathode. Cracks are painted in order to visualize them on photograph. Coke resistive preheating on full line load. Pot life was 446 days. The preheating cracks were not the cause of final failure.

12 If the preheat crack for some reason becomes wide enough for a sufficient amount of aluminium to remain in the metallic stage over some time, the increased current density will start to erode the bottom block in the crack area, enlarging both its width and depth. Even if the metal in the crack finally reacts to carbide, the local erosion can have created an area of permanently higher current density in which block erosion will only accelerate as the distance between the metal pad and the steel collector bar is being reduced. This was the reason behind the pot failure illustrated in Figure VI-9. Seen from the cathode surface the crack spanned a single block without extending into the narrow joints on each side. Further demolition showed a wedgeshaped eroded profile down to the collector bar. b) 80 cm Stud Inner stub Anode Coke resistor cm Cathode block Preheating cracks Figure VI-8. Photograph (a) shows cathode surface with two parallel lines of preheating cracks, spaced cm apart, running through all bottom blocks. Cracking pattern corresponds exactly to inner stub-stub distance (b) and observation that the major part of current initially passes through the inner stubs. Pot life was 1233 days. The preheat cracks were not the cause of final failure. As can be seen from Figure II-9b, resistive preheat on metal may also result in a severe cathodic current imbalance during the early stage of the procedure. For the cathode shown in Figure VI-10, this imbalance was severe enough to crack one bottom block from thermal shock, later resulting in a failure similar to the one previously described. Non-linear thermal gradients may also result in lateral cracks in the cathode blocks that are not readily observed after a preheat, but may be detected indirectly by permanently lower current uptake in one or more blocks. Uneven current distribution can, however, also be the result of other types of cracks and failures which will be discussed later. A stationary linear vertical thermal

13 gradient in a block will be the "equilibrium" state from which rapid heating and cooling will generate strains and stresses, illustrated schematically in Figure IV-59. Rapid heating from the surface will cause the largest stress concentrations in the central part of the block while rapid cooling from the surface will generate maximum stresses near the edges. Figure VI-9. Pot failure caused by thermal shock cracking of one bottom block. Increased current density over time in the crack area resulted in severe local erosion. Pot life was 1119 days. The procedure that most effectively may reduce or avoid thermal shock cracking during resistive preheating is by shunting the pot (Figures II-10 and II-11) or use graphite instead of coke as resistor material. Other remedies can be reduction of the differences in resistivity of the coke beds underneath each anode by use of flexible contacts and allow the full weight of the anodes to rest on the resistor coke bed (Figure II-8). The evenness of the coke layer is important, both with respect to its height, contact area and granulometry. Graphitic and graphitized bottom blocks are normally not prone to thermal shock cracking during resistive preheat. The main reason for this is the increased thermal conductivity of such blocks, which effectively reduces the thermal gradients. Although thermal shock cracking can be common during full line load coke preheating of amorphous or standard 30% graphite blocks, it need not be fatal. During the continued heating of the pot the crack will come under compressive load and minor amounts of metal that may penetrate the crack after startup will soon react to aluminium carbide and seal the crack, as happened in cathodes illustrated in Figures VI-7 and VI-8.

14 Figure VI-10. An eroded preheat crack spanning the width of one bottom block. The pot tapped out from this point. Pot life was 856 days. Coke resistive preheating may also be used for Søderberg pots, but a requisite to avoid thermal shock cracking is a flat milled anode surface. Since the working surface of the anode adjusts to the shape of the metal pad surface, it is virtually impossible without milling to obtain even contact with the resistor coke/cathode over a large enough cathode surface (Figure VI-11). The result of such a preheat may be catastrophic. In extreme cases the unevenness of the cathodic current distribution has been found to melt the steel collector bars in their slots (Figure VI-12) and almost instantly graphitize parts of the amorphous cathode blocks, a thermal process that needs temperatures well above 2000 C. Figure VI-11. Cleaned cathode surface shows the uneven heating (almost no heating in middle of pot) after resistive preheating with a nonmilled Søderberg anode. Pot life was 2.5 hours.

15 Thermal shock crack Collector bar remains Bath in collector bar slot Steel plate (molten collector bar) Bath Refractory layer Cathode surface Broken carbon Debris Figure VI-12. Extreme uneven current distribution resulted in melting of collector bar in slot and instant graphitization and cracking of bottom block in current path. Pot life was 2.5 hours (bath tap-out through collector bar). INSTALLATION DAMAGE OF RAMMED PARTS Virtually all modern cathode designs use a combination of prebaked blocks and a carbonaceous ramming paste in the construction of the carbon lining. Such a paste is almost always used in the peripheral seam between bottom blocks and sidewall carbon and in the narrow seam between bottom blocks, although in some designs these blocks can be glued together using a carbonaceous glue or cement [14,15]. The use of ramming paste in the pot serves several purposes. Its most important purpose is as a sealant to fill the voids between the prebaked bottom blocks and between blocks and sidewall materials, thus to prevent liquid metal and bath to penetrate rapidly into the interior of the lining. Another important purpose is to act as a cushion to absorb some of the thermal expansion of the bottom blocks during cell preheat. The rammed parts undergo physical, chemical and thermomechanical changes from a plastic stage during installation to a baked and solid form during preheat or early operation. They therefore tend to be a weak link in a lining that otherwise consists of mostly preformed and thermally inert parts. The quality aspects of the ramming paste itself combined with the way it is installed may therefore have an impact on the quality and service life of the cathode. A ramming paste will sometimes bond quite well to the adjacent prebaked block thus making it difficult for liquids to infiltrate the sub-cathodic materials through a direct interfacial route. However, during demolition of failed cathodes it can often be seen that no bonding at all has taken place and that the ramming/cathode block interface has been readily accessible to liquid

16 infiltration sometimes during the operation of the pot (Figure VI-13). In other instances it may be experienced that pressure alone between lining parts has been enough to avoid infiltration, even when no bonding between the ramming paste and carbon blocks formed during bake-out of the cathode. Most lining designs today uses cold or tepid ramming paste, but hot ramming paste is still used at some locations. Apart from environmental and hygienic objections, the main problem with hot ramming paste is that its properties change fast during cooling and that it is practically impossible to inspect the quality of the installation once it is in place. Proper and even densification can only be achieved by good standard operating procedures and a numerous, well trained and fast-working ramming crew. Stratification of the rammed parts due to cooling can be a problem when hot ramming paste is used. The compaction must be finished within a limited time interval in order to avoid cooling and subsequent change in paste rheological properties. Overworking a room temperature type paste, especially when the paste is of the "wet" type (Figure V-21), may lead to an enrichment of binder and fines in the surface layer (Figure VI-14a). This surface layer will have shrinkage characteristics during baking that may differ from the rest of the compacted paste. When new layers of paste are applied on top of this, cracking and stratification between paste layers may occur during baking. In severe cases large segments of the peripheral bevel and border may become dislodged during operation. a) Figure VI-13. Aluminium metal and aluminium carbide covering the block side from top to bottom illustrates the lack of bonding between ramming paste and bottom block. This interface had been accessible to metal infiltration at least once during operation. a) Downstream side of block. b) Upstream side of block. Crushing of aggregate particles is another consequence of overworked paste, especially for "dry" pastes (Figure V-21). The grain damage usually takes place close to the tamped surface (Figure VI-14b) and is most

17 predominant when electro-calcined anthracite (ECA) is used as the aggregate. Some types of ECA are more susceptible to crushing than others [16]. The result is the formation of new aggregate surface area that lacks the binder coating. When more paste is rammed on top of this, this new layer will not be bonded to the underlying paste but represents a non-bonded and weak zone that may fail during baking or operation. It is not uncommon to have to fish considerable amounts of baked ramming paste lumps out of the bath of newly started cells. In some cases, and then in connection with other destructive mechanisms in the cathode bottom, a slow shedding of rammed parts may continue for years. Figure VI-14. Incorrect compaction practice may lead to stratification of rammed layers. a) Surface enrichment of binder-fines; b) Crushing of aggregate particles in surface layer. a) b) Low strength of coarse aggregate particles or poorly designed paste granulometry may further enhance stratification of rammed parts. A strong tendency for particle crushing in one particular ramming paste is seen in Figure VI-15 where electrical resistivity in baked specimens is plotted versus compaction (green density). Figure VI-15. Electrical resistivity of baked specimens as a function of compaction density of some commercial cold ramming pastes. Shaded area shows scatter caused by extensive particle crushing in one particular brand. Filled symbols/fully drawn curves: Anthracitic pastes; Open symbols/broken curves: Graphitic pastes (from Faaness et al. [17]). Electr. resistivity ( µω m) Green density (kg/m 3)

18 Electrical resistivity will generally decrease as the number and area of particle-particle contacts increase. However, in one paste a large scatter of values found above a certain green density (shaded area), indicates cracking and generation of new aggregate surface within the paste body. Similar phenomena can be seen from mechanical strength measurements of compacted ramming paste bodies. The strength increases with increasing green density up to a certain point. At higher green densities particle crushing starts and the baked carbon body loses its strength, illustrated through a sudden scatter in measured values (Figure VI-16). Tensile strength (MPa) Figure VI-16. Tensile strength of baked ramming paste specimens (anthracite) plotted against green density. Shaded area illustrates scatter due to particle crushing (from Faaness et al. [17]) Green density (kg/m 3) Figures VI-17 and VI-18 show examples of stratification failures in the peripheral seam/bevel area. Figure VI-17 shows a typical failure from overwork of a room temperature ramming paste. The detail shows the stratification found in the rammed bevel slope during autopsy of a failed cathode. This particular paste had been compacted with mechanized compaction equipment, readily seen from the serrated pattern made by the compaction wheel (roller), but similar stratification by overwork of the paste with handramming tools is equally possible. Figure VI-18 illustrates that the upper part of the peripheral seam and parts of the entire bevel lost the anchoring. There was poor paste bonding to the carbon filler block and the sidewall carbon. Vertical bath/metal-covered shrinkage crack surfaces can be seen in the remaining bevel (background) as well as on the vertical face of the peripheral seam (foreground). Not enough densification is mainly a problem associated with overfilling with paste or with pneumatic handrammers equipped with improper size tools. A too deep layer of loose paste will often result in poor compaction of the bottom part. As the top part densifies, the increased friction against the seam walls reduces the possibility to extend the ramming force to the lowest level of the paste seam. With respect to the surface area of the ramming head,

19 this should not be too large, maximum 20 cm x 20 cm but preferably smaller, in order to be able to exert a large enough compaction pressure upon the paste. Figure VI-17. Photograph of peripheral paste bevel shows stratification and lack of binding between paste layers caused by extensive crushing of paste surface particles during compaction (grooves are from serrated wheel of compaction machine). Pot life was 314 days. Figure VI-18. Loss of parts of the peripheral seam incline caused by stratification in upper part of seam and poor bonding between prebaked carbon (filler block and sidewall block) and rammed parts. a) Photograph of the damage; b) Schematic representation. Pot life was 9 days b) Poor bonding Stratification

20 a) b) Figure VI-19. In most modern designs the a) large volume of ramming paste in the sides have been replaced by b) preformed carbon (filler) blocks, c) profiled sidewall blocks and/or d) refractory materials. c) d) The routines for ramming the narrow joints between the bottom blocks and the peripheral seam may by themselves result in weakness that in turn may lead to infiltration and shortened pot life. The narrow joints are normally rammed first with ends closed by retainer strips to avoid paste from spilling into the peripheral joint. When the narrow joints are finished, the strips are removed and the ring joint can be rammed. The problem arises if the ramming of the narrow joints and the side ramming is not performed on the same or following shift. The surface of the exposed ramming dries rapidly by loss of softener and loses its property to bond to newly applied paste during baking. If there has to be an extended delay between the ramming of the narrow joints and the peripheral joint, the exposed surface layer of the rammed part should be roughened up to expose soft binder before new ramming is applied. The T-joint formed between the narrow paste and the paste in the peripheral seam (Figure VI-20) is hence a place where bonding may become problematic. Figures VI-21 and VI-22 show examples of infiltration through the T-joint that resulted in failure. Autopsies of failed cathodes have shown that a disproportionate number of early failures, i.e., pot lives less than 2-3 years, are caused by weakness in the rammed parts of the carbon lining [17]. Such failures are either caused by improper paste installation, bake-out or cell operation, in addition to poor mechanical and thermomechanical characteristics of the paste itself. Possibilities for later pot failures introduced during installation may be caused by improper ramming procedures as well as poor paste properties. It is important to know the compaction characteristics of a given paste and its temperature window (Equation V-33), since a good standard installation practice for one may be unsatisfactory for another brand (Figure I-72).

21 Figure VI-20. Top half of image shows peripheral ramming (front part of bevel removed), bottom half is bottom blocks and narrow seams. A few millimeter thick aluminium carbide deposit at the block/ring joint interface may indicate infiltration through gaps. T-joints between narrow seams and peripheral seam (arrows) are cracked and carbide-infiltrated. Cathode was taken out for repair at 105 days. Figure VI-21. Image a) shows a vertical core drilled through one T-joint (cathode in Figure VI-20) with crack and carbide deposits from metal infiltration. The sketch b) shows the location of the core. Core Gap Block Ring joint Narrow joint Crack Block Core view b) The damage inflicted on the cathode represented by the core in Figure VI-21 was two collector bars cut by infiltrated aluminium. The vertical infiltration path was through the crack in the T-joint, but the severe damage to the collector bars was caused by poor compaction in the lower part of the peripheral joint. This can be seen by the large amount of aluminium carbide now filling what was open porosity prior to metal penetration. If the ramming had been done according to standard operating procedure, most of this lateral infiltration would have been stopped, possibly only resulting in minor corrosion on the

22 underside of the collector bars. The poor densification was local and restricted to the vicinity of the two damaged bars. In neighbouring parts of the pot, proper densification of the lower part of the ring joint had resisted any lateral flow of metal and the collector bars showed no damage, even if similar infiltration paths existed through T-joints. Figure VI-22. Bath and metal filled gaps seen at a) intersection between bottom blocks and side ramming and through T- joint indicate infiltration. Top part is peripheral ramming, bottom part blocks and narrow joint. b) Metal plugs shown stuck in T-joints after upper part of blocks and ramming has been removed. Pot life was 468 days. In smelters where pots are not rebuilt in the potroom stalls, leveling of the cathode relining crew workload may result in that lined or partially lined cathodes will have to be stores for periods. Such stored cathodes should either have no ramming paste installed or all ramming done. A cathode should never be stored for any period of time when it is only partially rammed. On the other hand, once all ramming paste has been installed, a cathode can be stored for extended periods of time at least for one year provided that it is stored in a dry environment. Critical ramming paste properties do not change significantly during long storage once it has been rammed to its desired density. This has been verified by baking and testing rammed carbon bodies stored in different environments for long periods of time. Extended storage of lined cathodes has little relevance in the day-to-day operation of a smelter but may have to be evaluated during construction of a greenfield facility, smelter expansion or for maintenance of the cathode inventory during temporary shut-down of a line.

23 SHRINKAGE OF RAMMING PASTE While baked carbon materials display a positive coefficient of thermal expansion at temperatures lower than the original baking temperature (Figure IV-1), carbonaceous ramming pastes will normally shrink after solidification (Figure IV-2). This is caused by the carbonization reactions in the binder. Since the binder coke is denser than the liquid binder, the binder coke bridges holding the aggregate particles together will contract and result in a volumetric shrinkage of the paste. The magnitude of this shrinkage will mainly depend on the granulometric composition of the paste (Figure IV-3) and only to a lesser extent on the binder content. The compaction density of the paste will normally not have any large influence on its shrinkage behaviour (Figure IV-4). The calcined paste aggregate will expand over the entire temperature range, and as the binder carbonization reactions go towards completion, the bulk paste starts to expand again, usually in the temperature range C. The net linear shrinkage of the paste will be the net thermal movement from the temperature it solidifies to 950 C. Since the carbonization reactions in the binder are not only a function of temperature but are also time dependent, the isothermal contraction during heat soaking at a high temperature, i.e., operating temperature, has to be added (Figure IV-2). The prebaked bottom block has a positive CTE at all actual temperatures (Figure IV-1) and will expand until operational temperature is reached. The block will expand further during early operation due to sodium adsorption. The thermal movement of the rammed peripheral border must be considered in two parts, i.e., swelling in the temperature range where the paste is plastic, and usually some shrinkage between the hardening temperature and the final temperature of operation. The combination of paste swelling and block expansion in the lower temperature range (normally up to about 500 C) may slightly extrude the paste from the seam and will not be detrimental to the baked block-paste seal. Cracks or gag openings in the lining are likely to happen in the temperature range where the paste shrinks. As an illustration the following simplified model is assumed. The heat is applied evenly over the crosssection (carbon pane) of the cell. A 3 m long bottom block with average CTE 5.4 x 10-6 K -1 in the temperature range C, will expand 8.1 mm. Any sodium induced swelling will be in addition to this. A peripheral seam with width 40 cm (80 cm total) and total shrinkage 0.4 % will shrink only 3.2 mm, i.e. considerably less than the thermal expansion of the prebaked block under ideal circumstances. The cross section of the cathode (carbon) is, however, not evenly heated. The peripheral seam is usually protected during the preheat, by banking it with cryolite or some form of insulation. Baking will thus proceed indirectly by heat transfer from the bottom block. Measurements have shown that the

24 main body of peripheral paste may still be plastic while the bottom block has completed most of its thermal expansion [17]. Even if moderate to small amounts of paste is used in the peripheral seam, parts of the seam may not have exceeded 500 C when a 48 hours preheating is finished and most of the block thermal expansion is done. Figure VI-23 shows the carbon pane bottom and side surface temperature distribution during a gas preheat. Although the surrounding ramming, refractory and insulation are not included in the images, they are included in the model calculations. The model calculations indicate no baking of peripheral rammed seam after 24 hours (Figure VI-23a), only partial baking after 48 hours (Figure VI-23b) but fully baked after a 72 gas preheat (Figure VI-23c). If there is a significant time delay between the heat-up of the bottom block and the baking of the peripheral seam an excessive paste shrinkage in a wide ring joint may lead to opening of gaps at the interface between paste and the prebaked bottom blocks (Figure VI-24). Sodium expansion of the bottom block may tend to counteract this type of interfacial gap formation, but will partly be delayed in time and may not prevent damage caused by early leakage of aluminium and bath into the interior of the lining. Ramming paste shrinkage perpendicular to this direction, i.e., vertical shrinkage cracks across the peripheral seam, is a more important consideration. Again assuming a paste with 0.4 % net shrinkage, which is quite normal but far in excess of what should be tolerated, the total crack width for each running meter could amount to as much as 4 mm. By assuming a circumference of the peripheral rammed seam in a small-tomedium sized cathode to be 30 m, the worst possible situation will correspond to a total crack width of 12 cm, i.e., with a seam width of 40 cm this corresponds to a total gap with a cross section of 480 cm 2. This will, of course, not be the situation in an industrial cathode. The paste shrinkage will instead give rise to a great number of transverse cracks in the peripheral seam (Figure VI-24), which will give several openings in the carbon lining through which bath and metal can penetrate. For pastes with very poor shrinkage characteristics such shrinkage cracks have been found with less than 1 meter spacing almost along the entire periphery of the cell. Cathode autopsies have verified this shrinkage mechanism, followed by the bath and metal penetration described above. It is therefore strongly recommended that ramming pastes used for lining cathodes show no or very low linear shrinkage (< 0.2 %) in the temperature range C. This is particularly critical if the cell is preheated on molten aluminium (not a recommended method, see Chapter II). Figure VI-25 shows photographs of such shrinkage cracks in the peripheral seam. These were only two out of several. Figure VI-25a shows one still filled with metal at the shutdown of the pot while Figure VI-25b shows one where part of the crack eroded to a hole through which the pot finally tapped out.

25 Figure VI-23. Modelled isotherms in a ¼ section of a cathode carbon panel seen from side and below. a) After 24 hours preheat on gas. b) After 48 hours. c) After 72 hours [18] b) t ( oc) c) t ( oc) Figure VI-24. Early metal and bath penetration can take place through vertical shrinkage cracks in the peripheral seam or through interfacial gaps that open between seam and bottom blocks. Deckplate Shrinkage crack Ramming paste Interfacial gap Bottom block

26 The rammed peripheral seam has often been found to be the weakest part of the lining where massive amounts of bath and metal can drain into the cathode interior in early life. Figure VI-26 shows an autopsied cathode with a vertical shrinkage crack through the rammed seam. The crack surface is covered with bath and metal. A solid cone of aluminium is present in the sidelining below the crack and large amounts of metal have flowed into the bottom lining. A collector bar in the vicinity of the failure was exposed to molten aluminium and the cell was cut out because of iron contamination in the metal. The rammed joints between the bottom blocks do usually not develop these shrinkage cracks. The joints are only a few centimeters wide and the friction forces between the paste and the prebaked blocks are high enough to distribute the shrinkage stresses in such a way that a large number of microcracks will develop in the ramming paste rather than a few major cracks. This micro-cracking will only result in added porosity and will not have any detrimental effect upon an already porous baked carbon material. In addition to selecting ramming pastes with more acceptable shrinkage characteristics, narrowing of the ring joint width can be used to minimize the possible development of fatal shrinkage cracks in the peripheral seam. (See Chapter I, Bottom block installation.) Figure VI-25. Vertical cracks in peripheral seam. a) Filled with bath and metal. b) The crack eroded to 3 hole causing a tap-out. Pot life was 350 days. COLLECTOR BAR INDUCED FAILURES The steel collector bar in aluminium cell cathodes is normally sealed to the bottom block with cast iron, carbon ramming paste or carbonaceous glues/cements. Since the thermal expansion coefficient of the current collector bar in parts of the temperature range can be four times that of carbon, severe thermomechanical stresses can develop when the assembly

27 is heated. Such stresses can lead to crack formation during preheat or early operation that may destroy the cathode bottom or strongly increase the cathodic voltage drop. Figure VI-26. Example of cathode failure caused by massive bath and metal penetration through shrinkage crack (surface covered with frozen bath and metal) in the peripheral seam. Pot life was 5 days. It is not the task here to go into the details of strain and stresses generated in the cathode lining, but just point out a few basic concepts and show how they are likely to affect the carbon blocks. A negative thermal gradient downwards into the bottom block will tend to flex the central part upwards relative to the ends. Swelling induced by sodium absorption may give additional heaving that is larger than the thermal effect (Eqn. IV-12), i.e., the longitudinal profile of the bottom block will tend to produce a convex surface curvature (Figure VI-27). Although an outer pressure will counteract part of the expansion due to sodium absorption, an absolutely non-yielding support will cause crack formation and/or displacement of bottom blocks. The steel collector bar must be able to slide in the slot. The surface of the collector bar (with rammed or glued connection) or the slot (with cast iron connection) should be smooth and without faults or excess roughness that can lock the bar in the slot and restrict the relative movement. In case the collector bar-carbon connection is such that movement becomes restricted during heating, the higher CTE of the steel relative to the carbon will set up tensile stresses in the lower parts of the block. These stresses can initiate or extend existing corner cracks (see Figure I-58) and, in the most extreme cases, lead to complete crack failure of the block. It is the authors view that cathodes should be constructed with full length bottom blocks and split collector bars. This concept allows the bar to slide both inwards and outwards in the block. If, however, a very strong refractory (LCC) castable dam is used as a sealant for the collector bar between the bottom block and the lower side insulation, this can effectively fix the collector bar at this position. The only alternative for the bar is then to slide inwards. It is then important that the two collector bar ends do not meet inside the slot or

28 that a hard, rigid spacer block (carbon or hard refractory spacer) is used to separate the two. If the inward movements of the blocks are hindered, stresses may develop that gives vertical transverse cracks in the block, starting from the bottom face and sometimes propagate all the way to the cathode surface. A crushable insulation material as spacer will remove this problem. Over time the spacer will be replaced by bath components that penetrate the bottom block porosity. By then, however, it will have served its purpose. b) High temperature, high Na concentration Lower temperature, lower Na concentration Max deflection a) Figure VI-27. Flexing of carbon blocks (all movement indications are grossly exaggerated). a) Vertical sodium and temperature gradients in the cathode will tend to bend the middle of the block upwards. b) Similar bending of the carbon bottom (treated as a solid plate). If most of the lengthwise bar movement can be inside the block, as with the split bar design, the movement of the bar relative to the collector bar seals in the shell will be minimized, thus maintaining a better seal against air ingress. High temperature creep in the bar is another mechanism that is operational at elevated temperatures as the yield strength decreases. At about 900 C the collector bar is malleable and will slowly deform under pressure. If, on the other hand, the steel-carbon interference occurs fast or at a temperature where enough steel strength is preserved, the horizontal cross-sectional pressure from the steel may exceed the strength of the carbon. This will lead to initiation or propagation of wing cracks, which will reduce the friction between steel and carbon and allow more unrestrained movement of the current collector bar. This will, of course, also reduce the contact pressure, deteriorate the electrical contact and increase the cathodic voltage drop. Thermomechanical interference of collector bar and block was discussed in Chapter IV. In practical collector bar rodding this becomes most important with use of cast iron. There are still smelters that rod their bottom blocks with cast iron and no preheating of the bar, which are inviting bottom block cracking and reduced pot life. The way it is done is to dam sections of the gap between the slot and the bar and pour the iron compartment-wise (Figure

29 VI-28). Section 1 is poured with both block and bar being cold. Upon cooling, part of the heat from the cast iron will dissipate to the neighbouring Sections 2&3 (Figure VI-29), which is subsequently poured. This continues until the entire block is rodded. Figure VI-28. Examples of section-wise cast iron rodding of collector bars without external preheating of the bar. a) Example of pouring sequence. b) Section-wise rodding in shop. c) View after 2 nd pour (from Caruso et al. [19]) a) When Section 1 is poured, the cast iron freezes almost immediately in the space between the cold bar and block. It may be argued that the rest of the compartments will have received some preheating by heat dissipation from this first pour, but as Figure K shows, it may not be enough. Anyway, when the assembly is heated prior to cut-in, the lateral expansion of bar and cast iron will soon overtake the expansion of the slot. Block cracking around the Section 1 pour will hence be almost inevitable, as verified both experimentally (Figure VI-30) and model-wise [19]. Whether this cracking will be contained locally or initiate catastrophic crack propagation throughout the block, depends on the resulting stress development in the assembly during preheating, startup and operation of the pot.