The CMS Central Calorimeter, consisting of the barrel (HB) and endcap (HE) calorimeters, sits inside the CMS solenoid vacuum tank. HB covers the h range 0 < |h |< 1.3 and extends radially between r = 1.806 and r = 2.95 meters, while the HCAL endcaps (HE) cover the pseudorapidity range 1.3 < < 3.0 . Both components share many features. They use scintillator/fiber as the active detectors are made of the same absorber material, and share many similarities of construction.
The presence of the 4 tesla magnetic field places unusual requirements on the calorimeter absorber. First, it should be non-magnetic to eliminate magnetic forces on it, and to avoid distortion of the uniform central field in CMS. Second, it should have a short interaction length to best take advantage of the limited space. It should have a relatively low z to limit multiple scattering of muons. It should be a structural material, because it must support itself as well as all other interior components in the CMS central detector. Finally, it must be affordable. These requirements led us to choose a copper alloy as the absorber material. Copper has an interaction length of about 15 centimeters, compared to 17 cm for iron. Thus we can pack substantially more interaction lengths into the same region by using copper. We chose a copper alloy, 90-10 brass (90% Cu, 10% Zn) due to its better machinability compared to pure copper.
The HB is a cylindrical structure created by connecting 18 calorimeter "wedges" into half-barrels. Each wedge subtends 20 degrees of j, and extends from the CMS detector mid-plane about 4.3 meters in z and weighs 25.7 metric tonnes (tonne). Fig. 2. 1 shows the overall view of the HCAL half barrel. Fig. 2. 2 shows the cross section of a single 20 degree wedge. The side view of the wedge is shown in Fig. 2. 3. The wedge is composed of copper alloy absorber plates that are bolted together into the wedge structure. Machined slots in the plates create slots for the scintillator trays. These 9mm thick slots are regions of constant radius that extend the full z length of the wedges. The wedge is composed of 31 copper alloy plates that are bolted together to make the body of the wedge. In addition, for strength, the inner and outer plates are made from stainless steel. The wedges are bolted to each other with bolts that are located in the front and back stainless steel plates.
The HCAL will sit on rails attached to the inner shell of the vacuum tank of the superconducting coil. The 2 horizontal wedges of each half barrel will have a mating rail structure that will rest against the rails of the vacuum tank.
The HCAL design is unusual in that brass is used as a structural element. Because there is not much experience using this material structurally, we have chosen a cautious design path. We have made a variety of finite element analyses of the calorimeter structure, using different assumptions on boundary conditions and other details. We have chosen the models that predict the largest forces in our structure (worst case models). In all cases, we design using the yield strength of soft copper, rather than cold worked material. Using that yield strength, we have designed the calorimeter to have a factor of 2 safety margin.
Each half-barrel is supported on two rails attached to the vacuum tank in the median plane (3 o'clock and 9 o'clock). The weight of each half-barrel is distributed uniformly along the length of the rails using a plunger/disk spring system. In turn, each supermodule of the crystal electromagnetic calorimeter (ECAL) is supported from the innermost stainless steel plate of an HB wedge via a 4 point suspension system attached to guide rails machined into the stainless steel plate of the corresponding wedge. The CMS tracker is supported on the innermost stainless steel plate by a four-point mounting frame, also attached to the plate in the median plane.
There is a unified CMS numbering system. The CMS global coordinate system is as follows (See Fig. 1.1):
- y is nearly vertical, x is horizontal and points inward to the center of the LHC rings, and Z is aligned with the beam. This results in y being inclined at 1.23% from the true vertical plane as the plane of the LHC is not horizontal.
- The azimuthal angle j is increasing from the x axis toward the y axis, while the pseudorapidity increases (decreases) with increasing (decreasing)z.
The HCAL half-barrel at positive z is labeled +1, while the negative Z barrel is labeled -1. The wedges are numbered 1 through 18 for each half-barrel, starting with 1 at the X axis and proceeding counterclockwise towards positive y.
Each HB wedge contains 17 active layers of scintillator trays, starting with the first one between the front (or inner) stainless steel plate and the first brass absorber plate. This one is numbered 1, and the last layer, outside the outer stainless steel plate, is numbered 17. There is an additional layer in front of the inner stainless steel plate, labeled 0. It is possible that there be even an additional scintillator layer inside the ECAL, immediately in front of the ECAL strong-back plate. If so, this layer will be labeled -1.
Each HB wedge weighs about 25.7 tonnes (metric). The ECAL supermodule weighs 4 tonnes (metric). The total weight of the tracker is approximately 4 tonnes (metric). The combined weight of the HCAL barrel plus components supported by the HCAL barrel is then about 1130 tonnes (metric).
We are designing the HCAL structure to have a factor of 2 safety margin above the static load. Therefore all dynamic loads must be less than this absolute allowed loading. This requirement places constraints on the crane in the assembly hall, the gantry that will lower the half-barrels into the collision hall, and the insertion tooling. In addition, this also places constraints on the shipment of the wedges from the absorber factory to CERN.
The entire weight of the HCAL rests on the rails on the inner shell of the vacuum tank. Therefore it is essential to ensure the integrity of the welds that connect the rails to the vacuum tank. Each weld will be x-rayed to check for defects. The welds are designed to have a safety factor of 2 to the yield strength of stainless steel.
Specialized tooling is required for the construction and installation of the HCAL barrel. A plate lifting fixture is needed for manipulating the plates. An assembly table is needed during the process of stacking the plates into wedges. A wedge lifting fixture is needed to allow the lifting and rotation of wedges for construction of the half-barrels. A cradle and spider support mechanism is needed for assembly of each half-barrel. An insertion/extraction tool is needed to insert and remove the half-barrels from the vacuum tank.
In addition, tooling is required for quality control of the absorber structure. A load test fixture is needed to test the mechanical integrity of each wedge, starting with the first prototype wedge.
The crystal electromagnetic calorimeter barrel (ECAL) is constructed out of supermodules, each ECAL supermodule matching an HB wedge. As indicated earlier, the ECAL barrel is supported from the HCAL barrel supermodule by supermodule. The ECAL barrel supermodule is supported by a rail system integrated into the design of the HCAL front stainless steel plate. The requirement that ECAL be supported by the HB (inner) front plate places requirements on the HB front plate flatness and placement in the CMS global coordinate system. The ECAL needs to be in the form of a cylinder centered within 2mm of the r=0 of CMS and coaxial to the beam line. The ECAL supermodule-to-supermodule placement needs to be controlled to less than 1mm. These requirements can be translated into requirements on the HCAL global placement, the HCAL front plate flatness, and the HCAL rail system.
The HCAL half-barrels are much more rigid than the vacuum tank rails upon which they rest. A sophisticated weight distribution system has been designed which is described later in this chapter. In addition, a clearance of 5 cm is required between HB maximum radial extent and the vacuum tank minimum inner radius.
The Tracker weight of 4 tonnes (metric) will be carried on the HCAL structure. Two supports on each end will carry the load from the tracker to plates that are attached to the front face of the HCAL at z beyond the extent of the ECAL. The supports will be at +- 30 degrees in from the -y direction. These supports will trap the ECAL supermodules interior to them in z. Therefore the design of the tracker support will include the addition of temporary supports so that one of the permanent supports can be removed. This will allow maintenance of the freed ECAL supermodule if necessary.
The HCAL is composed of 36 wedges. There are 4 special wedges of 2 types, those at 3 and 9 o'clock. These wedges have the rail structure mounted in the back plate. The remaining 32 wedges are identical. Figure 2.2 shows the cross section of a "typical" or "regular" wedge, while Fig. 2. 4 shows the cross section of one of the 4 special wedges that provide the support structure for rail mounting.
Each wedge has an inner stainless steel plate, followed by 31 brass plates, ending with an outer stainless steel plate. The wedges are bolted together across the inner plates and the outer plates. The phi edges of the wedge will have thin "skins" installed, to retain the scintillator packages. The wedge structure will have a "cutout" at the large z large r end to contain the photodetector/electronics box. To avoid eddy currents in the HCAL in case of a fast discharge of the solenoid, the stainless steel inner and outer plates will have an electrical isolation coating applied to them. (Note that the brass plates of adjacent wedges do not touch.) Forces transverse to the wedge (and bolts) are distributed along key bars (shear keys) inserted snugly into each plate. 18 wedges are bolted together to form a half-barrel. Stainless steel ties, inserted and bolted into machined slots in the outer plates take up any shear forces present. The inner plate has aluminum ties bolted on to take up shear forces. The 2 half-barrels rest on rails inside the vacuum tank. The half-barrels are not connected together.
We have chosen 90% Cu/10% Zn brass for our absorber material.
It has roughly the same density for the same price as 99% Cu.
We performed a market survey to see what
materials were commercially available in the form we required.
We solicited proposals regarding material choices from copper
foundries in the United States, Europe, Russia, and China. Our
first choice was Phosphor-Bronze, a very strong material. Unfortunately,
this material was available from only one vendor, and was three
times as expensive. Our second choice, 90-10 Brass, C22000, was
available from a number of vendors.
Issues regarding the choice of brass are its availability in proper form, strength, density, machinability, and potential for creep.
C22000 has a yield strength of 70 MPa (10 ksi), (1 ksi = 1000 lb/in2 = 1000 psi) 1000 psi) and a ultimate strength of 260 MPa (37 ksi). It has a machinability rating of 20 where C36000 free-cutting brass is 100 and pure copper is 20. It has a fair machinability. C22000 has a density 3% less than pure copper. Its nuclear interaction length is 15 cm.
We are concerned with the possibility of long term creep of the brass material. We have searched the literature regarding creep in brass or copper. At the loads we anticipate, creep is not an issue at room temperature. However it is prudent to experimentally verify this result. We are preparing a creep experiment to test this prediction.
The inner and outer plates are 7cm thick stainless steel. The interior plates are 2.9cm thick and are composed of C22000 brass. Brass ingots are cast, then hot rolled to the approximate dimensions of the plate. Finally, the faces of the plate are machined, to form the 9mm slots for scintillator trays, and to add the bolt holes, threading, and slots for keying. Fig. 2. 5 and Fig. 2. 6 Fig. 2. 6show the layout of the stainless steel front and back plates. Fig. 2. 7 and Fig. 2. 8 show the layout of the 2 types of brass plates which allow for inner and outer scintillator trays respectively.
The absorber plates are bolted together with M16x2 low head stainless steel bolts. Fig. 2. 9 shows the bolt geometry.
As seen in Figs. 2.7 and 2.8, the plates are bolted together along 4 rows. Each row has bolts at 20cm centers. There are 2000 bolts per wedge, and about 72,000 bolts in the total HB structure. To supply a preload to the plates (to prevent them from separating due to the forces on the wedges in the barrel structure), the bolts are torqued to 170 N m (125 foot-pounds) of force, supplying 51.7 kN (11600 pounds) of preload per bolt. (The threads of the bolt will be lubricated with silicone.) There is a maximum separating force of 25.8 kN (5790 pounds) per bolt in the barrel structure. Thus the 51.7kN (11600 pounds) preload gives a factor of 2 safety margin.
The strength of the bolted connection depends on the bolt strength, the strength of the brass plate under the bolt head, and the strength of the threaded plate the bolt connects into. The FEA calculation of the maximum force on any bolt is 25.8 kN (5790 lbs.). We require a safety factor of 2, so the maximum force = 51.7 kN (11600 lbs.).
The force on the material under the bolt head is
with D the diameter of the bolt, h the thickness of the plate under the bolt, and t the allowable shear stress of the material, 5 ksi for soft copper.
- For our plate and bolt design, D = 23.8mm (15/16"), h = 20.3mm (0.80") and thus P = 52.6 kN (11,800 lbs.) for a safety factor of more than 2 . 2.8.3
Bolt tests on the efficacy of the head socket (torque test) and the strength of the bolt (tensile test) are being carried out at the University of Mississippi and at an ASTM certified testing lab - MMA Laboratories, Inc., Newtown, PA. For our research and development work we have selected American standard size 5/8"-11 UNC bolts because they are readily available. The tensile stress area for the 5/8"-11 bolts is 146 mm2 (0.226 in2) whereas for M16x2 bolts it is 157 mm2 (0.243 in2). This gives us an additional margin of safety on the bolt. All tests so far have been carried out on thin head 5/8 inch diameter, #11 rolled threads , 2 inch long bolts fabricated from 304 stainless steel and from Nitronic 50. Two bolts of stainless steel and three of nitronic were tested and the results are presented in Table 1.1.
Tensile Strength
Testing was performed in accordance with ASTM-A370-95. The yield point is not the point where the bolt fails, but corresponds to the point where the bolt becomes permanently deformed and will not recover totally elastically. These tests are being duplicated at the University of Mississippi. It should be noted that the bolts are stronger than the bolt base material, which was also tested by the lab. We ascribe this additional strength to work-hardening that occurs when the thread is rolled. The bolts are able to withstand the required 51.7 kN (11600 lbs.) of force.
Torque test
To achieve the 51.7 kN (11600 lbs.) preload, the bolt must be torqued to 170 N m (125 foot-pounds) of torque. We tested the bolt design to verify that this torque is achievable. The torque tests are carried out using a modified ASTM F880 procedure. The bolt is threaded into a hardened steel fixture until the bottom of the head is nearly flush with the fixture. Then a hardened steel bolt is threaded from the bottom of the fixture until it bears against the bolt under test. A hex key bit is inserted to the full depth of the bolt socket and the torque applied with a torque wrench. The Nitronic 50 bolts have a socket strength of greater than 197 N m (145 ft-lbs.). The 304 SS bolt sockets fail at about 190 N m (140ft-lbs.). Although this is greater than the needed 170 N m (125 ft-lbs.), we are modifying the bolt design. We will change the Allen socket for a Torx socket, which can withstand more torque.
Thread tests
We plan on bolting the bolts into threads cut into the absorber plate. Since this is a soft material, we need to know at what point the bolt or the threading will fail. We obtained samples of hot-rolled copper from Non-Ferrous Metals Company, Sofia, Bulgaria (a potential vendor for the HCAL absorber). Hardened steel bolts are fitted through a 19 mm (3/4 inch) thick piece of the copper alloy and then threaded into a hole tapped into a second piece of copper alloy thereby clamping the two pieces together. Standard feed holes and threads are used (either UNC ½-13 or UNC 5/8-11). The bolts are then torqued until thread failure. It was noted that about 13.6 N m (10 ft-lbs) before thread failure a noticeable change in the way the bolt was reacting to the torque occurred. We attribute this to the copper flowing around the bolt into the non-threaded section.
The results of the test are given in Table 1.2.
| |
The maximum torque's for the 5/8" threading are considerably larger than the required 170 N m (125 ft-lbs.) needed to supply the 51.7 kN (11600 lbs.) preload. In addition, the brass should have better mechanical properties than the copper. Tests on brass are under way.
The shear forces between the wedge absorber plates cannot be held by the friction induced by the bolt preload, as it is too small. Shear forces will be taken by keys, as shown in Fig. 2. 4.
The keys are 12.7 mm (0.5 inches) square in cross section. The keyway extends the whole length of the wedge. The keys are press fit into the lower plate keyway. Each key is 20cm long, and the keys are placed continuously for the full length of the wedge.
In calculating the forces on the keys, we will conservatively assume that the friction force = 0 and that the keys will have to take up all of the shear force. The maximum shear force F is applied to the joint between the outer SS plate and the first inner absorber plate. The maximum shear force is exerted on the horizontal wedges, at 3 and 9 o'clock. From the FEA analysis, the maximum force on a bolt is 25.8 kN (5790 lbs.), with the distance between two bolts of 200mm (7.87"). The allowable shear stress for soft copper, tmax , is 35 MPa (5 ksi).
tmax =Fmax/A = 5970/(0.57.87) = 1.52 ksi (10.6 Mpa), (2)
This well below the allowed limit. In reality, the key will be made of cold drawn material which will have an even larger tmax. The stress in the plate is small. The normal stress is s.
s = Fmax/A = 5970/(0.257.87) = 3.0 ksi (21 Mpa) (3)
Again this is well below the maximum allowed stress in copper of 70 Mpa (10 ksi). The technique for installing and loading the keys during the wedge assembly process is described in the section on wedge assembly.
The keys and keyways structure take the shear forces in the phi direction. We also need a pin to take shear forces in the z direction, along the long length of the wedge. The wedges are built and always remain in the orientation with the long axis of the wedge parallel to the floor. Then there are nominally no shear forces in the z direction. However, for safety, and to serve for plate alignment during assembly, we will place one shear pin between each pair of plates.
The FEA model used for the wedge is 2-dimensional. We justify this approximation because the wedges are continuously attached together, and continuously supported. The bolts were modeled as point hinges between plates (i.e. like spot welds). The plates were free to rotate about the point attachments. In addition, the absorber plates were allowed to penetrate each other. These simplifying assumptions were used because it is a "worse case". In actuality, the bolt head and nut will supply some torque to prevent plate rotation. Thus this model will have more deformations than the final wedge.
The two simplifying assumptions above were justified by studying their effects via finite element models. To test the effect of the "spot weld" treatment of the bolts between absorber plates, we made a model where the absorber plates were "welded" together along their length. This model was more rigid than the "spot weld" model. Therefore we chose the spot weld model because of its more pessimistic predictions.
To test the effect of absorber plate penetrations, we made a FEA model of the barrel structure where the absorber plates could not penetrate each other. In this model we found that the forces on the plates were small. Therefore we elected to use the model where the plates can penetrate.
The wedges make contact with each other only at the inner and outer stainless steel plates. We studied two models of the inter-wedge attachment. The first model had the inner and outer SS plates welded to the neighboring wedges along the full length of the wedge. The other model of wedge attachment had SS plates are attached by hinges. The plates could pivot about those points. We found that the maximum tensile force on the bolts connecting wedge plates was 1.6 times larger for the hinged model than for the welded model. The maximum tensile forces on bolts connecting wedges was very close in the two models 78.9 kN (17688 lbs.) for hinged, 79.7 kN (17863 lbs.) for the welded.
Therefore we made our default model one where the absorber plates are connected by spot welds, and can penetrate each other. The adjacent wedges are connected by hinges. This model will give us pessimistic results. The model is shown in Fig. 2. 10. The brass absorber plates do not touch. The force vectors on the front of the calorimeter represent the 4 tonnes (metric) ECAL modules.
The deformations on the calorimeter are shown in Fig. 2.11. The values are in polar coordinates, centered in the nominal center of the barrel. The maximum deflection is in the vertical Y-Z plane, and is about 1.3mm (0.05"). The scintillation slots also suffer distortion. The maximum closing down of a slot is 0.4mm, Fig. 2.12. This is large enough that we will need to take it into account in the preparation of scintillators for those slots.
The stresses on the plates are low, 42 MPa (6000 psi), with sy = 70 MPa (10000 psi), as shown in Fig. 2.13. and Fig. 2.14 . The maximum stresses are found in the horizontal rail wedges at the connection between the stainless steel outer plate and the first copper plate. A maximum bolt load of 25.8 kN (5790 lbs.) occurs in this region.
We have also compared this FEA to other studies. One study treated the bolts similarly, but was 3-dimensional. We found consistent results between these 2 models.
Table of Contents
2.1 MECHANICAL OVERVIEW 972.2 STATIC LOADS 1012.3 DYNAMIC LOADS (ASSEMBLY/INSTALLATION) 1012.4 TOOLING 1012.5 ECAL REQUIREMENTS 1022.6 VACUUM TANK REQUIREMENTS 1022.7 TRACKER MOUNTS AND REQUIREMENTS. 1022.8 WEDGE DESIGN 1022.8.1 Overview 1022.8.2 Plate layout 1042.8.3 Bolts and bolt patterns 1092.8.4 Shear keys ,keyways, and shear pins 1112.8.5 Calorimeter FEA summary 112