Spanish engineers create a method to find the ‘black spots’ based on probabilistic analysis of the elements that generate accidents

The occurrence of accidents on a road is a problem of a random nature. This means that, in principle, accidents do not have to happen in the most dangerous areas, although they do occur more frequently. To analyze this random problem, techniques related to probabilistic safety analysis could and should be used.

The National Traffic Authority indicates in the Spanish road network what it calls “accident concentration zones”. The idea is to mark those areas or road segments where there have been a certain number of serious accidents and inform users so that they can pay attention and take the appropriate measures when they are driving by them.

This way of working, although it has a certain sense, is not the most convenient since, following this procedure, there will be zones marked as very dangerous that will not be and, what is worse, others that being very dangerous will not be marked.

To illustrate, let’s use a simile. Consider a very large group of friends who play the lottery, in principle in different numbers and numbers unknown. Let’s suppose that one of them touches the fat one year. So, if we wanted to predict who the fat man will play next year following a methodology similar to that of the accident concentration zones, we would predict that the fat man will touch the same person who touched him last year or the previous years. Here, as is evident, it would be necessary to differentiate between the amount of lottery that each player plays and how many different numbers each one plays, which, taken to terms of traffic, would refer to the severity of the accident suffered in his case and the number of times he The section of road in question is covered. If the lottery touched the one who plays the most different numbers, we will do well to predict that this person will be the lucky one for the following year, but if it was one of the many who play a single number, the prediction can be very bad. It should be noted that one thing is to count how many times you get the fat and another, the amount received. Therefore, the number of accidents that occurred is not a good parameter for the hazard analysis. Their consequences should also be included, that is, their severity.

To proceed to the identification of the danger of a section, it should not only resort to the registered accident frequencies, which is a valuable index, but also to the intrinsic danger of each section.

In recent studies of the Royal Academy of Engineering (RAI), in collaboration with the University of Cantabria (UC) has presented a new method based on the probabilistic risk analysis of roads, which allows identifying these areas without the need to wait that accidents occur.

The idea is to travel the roads recording in a video everything that is found by traversing them and identifying all the elements that can be the cause of accidents, such as curves, tunnels, viaducts, level crossings, overpasses, clearings, embankments, pavement condition, curbs, shoulders, information signs, prohibition, speed limit, etc. Once these elements are located by their kilometric point, the variables that intervene in each of them are identified, such as speed, type of conductor, states of the elements, possible decisions (correct or erroneous), weather, state of the vehicle, etc. It is also important to incorporate human error, for example, through the state of fatigue and attention of the driver,

Once these variables have been identified, it can be analyzed how they depend on each other, representing them in a graph and connecting them in the cause-effect direction. In addition to direct dependencies, there will be other indirect dependencies. This can be analyzed by what are called acyclic graphs. Finally, it is necessary to quantify the probabilities of occurrences of the events, which can be done through conditional probability tables that give the probability that a variable takes a certain value when it is known that other variables, the direct causers, have already taken some determined values. With these two components, acyclic graph and conditional probability tables, so-called Bayesian networks are constructed, which allow determining the probability of any event, no matter how complex it may be.

With all the above you can get a road scanner, that is, a graph that gives us the probability of an accident that we accumulate as we go through it. From this scanner can be identified immediately, and without the need for accidents, which are the most dangerous points and areas, which correspond to the places where the largest jumps or slopes occur.

This would allow, on the one hand, to locate dangerous points and, on the other, to order them for danger, which would facilitate the optimization of resources, using them first to eliminate the most dangerous points and not squandering the available means trying to eliminate points with smaller risks.

The Bayesian network also allows identifying the most probable causes of accidents. You just have to fix the value of the variables that were observed when the accident occurred, assigning them a probability equal to one, and then recalculate the probabilities of the rest of the variables and events. This facilitates not only the identification of the causes but also the quantification of the probabilities of the rest of the unobserved variables.

It is also very important to identify the most frequent circumstances of occurrence of each accident, since errors in the determination of these circumstances would lead us not to use corrective means in incorrect directions and not to waste the available means. For example, if a curve accident is attributed to excessive speed and not to an incorrect design of the curvature, cant or properties of the road in said area, the true causes of the same can not be corrected and a money that has not been invested will be will produce the expected results.

To proceed with this type of analysis, the RAI and UC team have developed a software that allows this detailed analysis to be carried out, having shown that, given its low cost, the expected results would not only reduce accidents and the number of killed on Spanish roads, but would produce a great benefit in savings of repairs both vehicles and road infrastructure. The methodology has also been developed for the case of railways.

These new methodologies, which have been used for many years in the analysis of the safety of nuclear power plants, could also be used immediately, when technical and political decision makers decided so.