HONORS 392 21 November 2013 Deflection: An Illustration of the Duhem-Quine Thesis We both looked at each other with looks of skepticism and concern. It was more than just a serviceability issue, or so it seemed. Bridges just arent designed to deflect that much. As we stood on the sidewalk, we could feel the bridge bounce beneath our feet as trucks drove by, such that I had to grab the handrail to keep my balance. Bouncing bridges were something we were used to, wed inspected hundreds of bridges before. Bridge spans are designed to vertically deflect up to a ratio of 1:360 with the span length (that is, L/360) and we had grown used to feeling the rocking and swaying beneath our feet. But not to this degree. Something was wrong. Bet the girders have thermally expanded due to the heat and have produced excess camber in the main span causing that deflection. We better check the spacing in the joint headers to see how much more give our bridge can take said Greg confidently. Greg had been the lead bridge inspector for WSDOTs Bridge Preservation Office for fifteen years and had designed bridges for even longer. His vast engineering knowledge and past experiences gave him confidence that he knew the source of the problem. This rocking wasnt a shock to him; hed seen cases like this one many times before. I on the other hand, the intern and co-inspector of two months experience, nodded my head and got out my tape-measure to record the joint spacing. Looks like about 1- 3 / 8 to me I said, writing down the number in my report. No, no, you measure between the compression seals, not the whole joint width. Its going to be less than that Greg told me, inferring my mistake without even seeing me perform my measurement. Yeah, thats what I did. Greg looked up, puzzled. The report showed that the joint spacing was measured at 1- 3 / 8 two years ago as well, and assuming that both measurements were taken correctly, it seemed to indicate that no change in spacing had occurred. Guess there wasnt any thermal expansion after all. In that case it must be Well wait, there could still have been thermal expansion, right? Weve just checked this one joint. And it is summer time, Im sure they expanded a little bit, I mean it is hotter out than before. No, the girders dont have to expand just because its summertime. Its an elastic system, and so the internal forces in the frame must have sufficiently braced the connections preventing a thermal expansion. Again, Gregs past experiences were of great benefit to him. Relying on past experience had always treated him well. It had given him success, allowed him to predict outcomes, produce convincing explanations, and in turn receive a hefty paycheck. This time was sure to be no different. Well shouldnt we at least check the joint on the other end? Maybe it only expanded in one direction and only at the other joint. Greg agreed to check the other joint, partly to prove his point, partly to train Junior (as he semi-affectionately called me) in the ways of Bridge Inspection, and partly to get me to be quiet and to just follow his lead. Upon measuring the other joint header, Gregs contented face shined as there was indeed no difference in measurement from previous years data as he expected. There. Are you convinced now, Junior? The joints are unchanged, so there must not have been any thermal expansion. I was unconvinced. One minute Greg was sure there was thermal expansion, and the next minute he wasnt. Was there some crucial test he had performed which showed that indeed no thermal expansion had performed? I knew that men who have an excessive faith in their theories or in their ideas are not only poorly disposed to make discoveries but they also make very poor observations. They necessarily observe with a preconceived idea and, when they have begun an experiment, they want to see in its results only a confirmation of their theory. Thus they distort observation and often neglect very important facts because they go counter to their goal (228). In this case it seemed to me that Greg had based his entire argument on one data point, namely the joint expansions. He was willing to change his theory in no time at all because he wanted the joint measurements to be a guaranteed indicator of whether the span had expanded. What if I proposed an alternative theory to why the joints measurements were unchanged? Just because the joints dont appear to have contracted doesnt mean the span hasnt expanded, right? I mean, the piers beneath the joints could have shifted under the force of thermal expansion, pushing the joints further away from the slab, giving the illusion that the slabs remained the same but really everything shifted proportionally. You yourself said the frame was elastic! And besides, how do we know the measurements were taken correctly two years ago? Maybe some other intern like me made a goofy mistake. Surely I had trapped him. I had been taught that evidence can never condemn an isolated hypothesis but only a whole theoretical group (230). Furthermore, when the experiment is in disagreement with predictions, what is learned is that at least one of the hypotheses constituting this group is unacceptable and ought to be modified; but the experiment does not designate which one should be changed (233). Since the joints were not measured to be tighter than in previous years as expected, one of the hypotheses we raised must have been wrong. Either the bridge span truly did not expand, or the frame did not remain fixed and thereby the joints themselves translated, or the measurement was not taken properly in the past, or perhaps something else entirely. Where was the fault, which theory was to blame for the unexpected lack of movement? Surely this was non-deductible even for Greg! I awaited his response. Theyre right. I took the measurements two years ago. Gregs cold glare was sharp. There was no changing his opinion. It was settled then. We moved on in our inspection. We entered the UBIT (Under Bridge Inspection Truck) and began to inspect the substructure of the bridge. It was a fairly routine inspection. Nothing seemed out of the ordinary until we inspected the bridge piers beneath the central spans. Pack rust had built up between the gusset plate members, up to thick. The pier caps beneath the joint headers were pushed outward by this amount. This pack rust is new. See how the epoxy is still coating the plate around the connection? It hasnt had time to leech along the crack and erode it, so this must have happened within the last two years. Looks like you were right about the frame translating after all, and the center span has thermally expanded in turn. Nice work, Junior. Well thanks, Greg, but I dont think we can be so sure of this either. I just complemented you. Thats high praise. Now you want me to take it back? Well Im just saying that just because theres pack rust here doesnt mean there was thermal expansion in the center span, any more than when we thought that since the joint headers hadnt shifted there wasnt thermal expansion in the span. What makes this any more of a crucial experiment than measuring the joint headers? I figured my point was justifiable. Experimental contradiction does not have the power to transform a physical hypothesis into an indisputable truth; in order to confer this power on it, it would be necessary to enumerate completely the various hypotheses which may cover a determinate group of phenomena; but one is never sure he has exhausted all the imaginable assumptions (Duhem 235). The result is merely an ambiguous falsification. There is certainty in the existence of a fault, but no certainty in the source of the fault. I still believed there was no way to know with certainty which of our assumptions was incorrect, and that this new test did not have ultimate authority in determining whether the bridge had indeed expanded. Thats how inspections are done, Junior. Its called abductive reasoning. Its called inference to the best explanation. Right now thats the best explanation logically, so its the one Im going to go with. Im going to use my good sense and conclude based on evidence that this is what happened, and then Im going to write my report. And then Im going to get paid. While I wanted to point out that his good sense was itself very ambiguous and no more absolutely certain as our previous interpretations of data, I decided not to test my luck further. After all, I too hoped to be paid. We had almost finished our inspection of the bridge, though we had yet to inspect the connections between the girders supporting the main span and the piers. These connections in particular were designed such that 5 bolts 7 / 8 in diameter fastened the members together. Connection failures are sudden and inelastic. When a girder begins to buckle, it slowly bends and deflects, and even after yielding completely it still bears its shape until finally it fails. So when you notice the girders beginning to deflect, you know something is starting to go wrong. There is no such warning when a bolt shears. It suddenly and violently fails, splitting in two. As such, connections are designed with a much higher factor of safety. If a bridge is going to fail, you want the failure to be a ductile failure in a girder, not a brittle failure due to a sheared bolt. It is thereby not altogether surprising that I had never seen any problems with connections over the course of two months of inspections, and yet we always inspect them anyway. However, when I pulled out my hammer and tapped on a bolt, half of it slid out of its socket, and plummeted to the ground far below us. Greg was stunned. We had just found a bolt as part of a fracture critical member connection that was sheared in two. Though we felt safe in that there about were about two hundred bolts across the bridge, finding even one sheared bolt was alarming. I looked to Greg, expecting his usual charisma and his confident assurance in a new explanation. Give me the hammer was all that he said. Taking the hammer, he struck another bolt, and it too slid out of its connection, and fell down into the river below. Two sheared bolts? This was unheard of. Over the course of the next few hours as we inspected the rest of the connections, we found 76 bolts at girder-pier interactions that were sheared in half. At one connection there remained only a single intact bolt holding up an entire bridge span. It was now completely and absolutely clear in Gregs mind the reason for the excessive deflection in the bridge. It was not the thermal expansion of the girders. It was not the pack rust in the piers. The deflection was beyond any doubt now due to the sheared bolts in the connections. Greg was back to his old self. He once again was completely certain of his analysis, and he had his own separate set of reasons for why the bolts were sheared. While I contemplated my safety working as a bridge inspector, I also considered the sequence of events leading to Gregs final conclusion. If I had initially suggested that our assumption that there were no sheared bolts was incorrect as a possible hypothesis for the deflection, Greg would have laughed at me. This is indeed a drastic suggestion. The probability of a single sheared bolt is remarkably low, and the chances of multiple sheared bolts is increasingly unlikely. But was there any more certainty now in this new explanation for the joint headers reading the same spacing than the previous hypotheses? All of the hypotheses had evidences for them, and they all lead to the same conclusion. Simply finding unlikely evidence does not necessitate a conclusion to be based primarily on this special evidence more than other evidence. Perhaps it is even less warranted. Greg had fewer experiences involving both deflections and sheared bolts than deflections and pack rust or deflections and thermal expansion. Evidence of sheared bolts is not a crucial experiment any more than evidence of pack rust was. Again more assumptions could be questioned which would allow for the same interpretation of data. Perhaps thermal expansion created undue forces on the connections which acted to shear the bolts. Or perhaps the bridge deflected first, which then placed a torque on the bolts which caused them to shear. We are left with no more certainty than before. There remains [no] irrefutable procedure for transforming one of the two hypotheses into a demonstrated truth (235). The bridge system was reliant upon so many assumptions about how each element worked together, that to call into question how one element worked called the whole system into question. Gregs analysis of the bridge was also based on his understanding of forces and structural kinematics. His hypotheses were reliant on background assumptions of modern physics. His notions of thermal expansion assumed that a standard thermal coefficient existed for the center span, and that the steel had certain predictable behaviors to temperature, and on other material properties of the components involved. These assumptions are also included in the theoretical group of hypotheses under question when Greg incorrectly predicted the expansion of the joint headers. Suppose our knowledge of steels susceptibility to expansion is flawed. If by changing our understanding of thermal expansion to allow for girders in this particular situation to behave exceptionally leads to the same experiential conclusion that the joint headers would be unmoved, is this any less logical? Any statement can be held true come what may, if we make drastic enough adjustments elsewhere in the system (266). Because any contrary piece of evidence calls into question not a single hypothesis but an exceedingly large group of assumptions, it is possible that any number of modifications to the existing assumptions can result in an equally logical means of accepting the contrary piece of evidence. In our case, the joint expansion could have been explained initially by any number of possible corrections. Thus theories cannot be accepted nor rejected based on observation and evidence alone, as there is no necessary connection linking data back to a theory. What are we then to do? Any statement, even one from logic or mathematics, could be revised or abandoned in light of experience (377). As a guide, each man is given a scientific heritage plus a continuing barrage of sensory stimulation; and the considerations which guide him in warping his scientific heritage to fit his continuing sensory promptings are, where rational, pragmatic (268). Trying to explain this to Greg would not be easy. Greg youve changed your mind three times today. We still dont know why we didnt measure any changes in joint spacing. You cant be so sure that this time youre right. Yes I can. I changed my mind because I found new evidence each time, and changed my assumptions accordingly. But have we really considered everything? Can we say for sure we wont find something else that will have us change our minds again? How do we know this is where the fault lies? Couldnt we allow for any number of possibilities which can change our system and lead to the exact same result? Were inspectors, Junior. We go out, look at bridges, and figure out whats wrong. At the end of the day, Ill tell you whats right and whats wrong. Im not paid to come out here and just guess. But Greg, thats the only thing weve done. He was right though. Hes paid not to determine with certainty what assumptions about the bridge (no pack rust, unsheared bolts, typical steel, etc.) were in error. Hes paid to figure out what is to be done. The day arrives when good sense comes out so clearly in favor of one of the two sides that the other side gives up the struggle even though pure logic would not forbid its continuation (248). We shut down the bridge and replaced all 76 sheared bolts and moved on to the next bridge. We will never know why the joint headers did not show any movement when we measured them. In fact we will never know with absolute certainty that any of our assumptions in regards to bridge behavior are true. There may come a day when we are led to change our understanding of basic bridge mechanics, in light of new experience or data. However I continue to think of the conceptual scheme of science as a tool, ultimately, for predicting future experience in the light of past experience (267). We should be careful not to claim more than we do. Science is not certain, not insusceptible to change. Pragmatically, it is a practice of good- guessing.
Work Cited
Curd, Martin, J. A. Cover, and Chris Pincock. Philosophy of Science: The Central Issues. New York: W.W. Norton, 2013. Print.