Lift Report

The speed profile curve is used to extract the value of maximum acceleration and jerk during the stop. These two variables are then correlated with the quality of the stop as rated by the subjects. A linear multiple regression is carried out to fi nd a relationship between the quality of the stop and the values of maximum jerk and acceleration for this escalator.

One of the major risks on an escalator is the risk of passenger falls. Passenger falls could range from a single fall in the simplest case, to multiple avalanche falls in the extreme cases. An avalanche fall is the case where one passenger fall causes a second passenger fall and so on. Thus one of the more important aims of escalatordesign and operation is the minimization (and eventual elimination of) of the risk of passenger falls.

Passenger falls can be triggered by design related aspects, passenger behaviour or a combination of both. Design related aspects include safety device initiated automatic stops. Passenger behavior could include inattentive passengers, not holding onto the handrail or facing the wrong direction of travel.

The designer has little control over the passenger behaviour, but can control the design related aspects. If the designer can control the way in which the escalator stops such that the risk of passenger falls is minimized, then half of the problem is addressed. The other half relating to passenger behaviour can be then addressed by educating passengers and making them aware of their escalator environment.

This paper will only concentrate on the design related aspects of the risk of passenger falls. Passenger education and awareness in order to reduce the risk of passenger falls is outside the scope of this paper and will not be discussed any further.

The risk of passenger fall is at its highest during an escalator stop. This is explained as follows:

1. Normally, an escalator is not started with passengers on board, so there is no risk of passenger falls due to escalator starting.

2. Once an escalator attains its top speed is usually runs at constant speed.

3. On escalators where speed changes take place regularly (either for traffic handling or for energy saving), this is always done in a smooth manner that eliminates the risk of passenger falls.

So the main area that the designer has to concentrate on is the stopping of an escalator (either in response to an operational stop request, or due to an emergency condition such as a emergency stop switch being pressed or a safety device tripping). This is corroborated by data that suggests that 2.5 % of all escalator stops lead to a passenger fall (Metro Consulting Engineers ).

It is the operation and characteristics of the escalator braking system that influence the risk of passenger falls2). Thus, for the designer to influence the risk of passenger falls, he/she has to concentrate on the method of operation of the braking system. The problem currently for the designer is the lack of definition for the relationship between the braking characteristics (namely the kinematics) of the stoppage and the risk of a passenger fall.

The aim of this paper is to try to empirically find the relationship between the kinematics of the escalator stopping speed-time profile and the risk of passenger falls. These parameters will be called performance criteria.

Passengers are more likely to lose their balance and fall if an escalator stops violently. However, no quantitative objective measure exists to relate the risk/probability of a passenger falls to the quality of the stop. This section discusses existing criterion for braking performance assessment, and recent work carried out to attempt to find a relationship between the braking performance and the probability of passenger falls.

The emergency stopping criteria for the escalators are twofold and conflicting: To stop the escalator within a specified distance, but to achieve that stop in a way which does not cause passenger falls. These two requirements necessitate a compromise setting, and it is important to specify a criterion that satisfies both.

The European Standard (CEN, 1983: EN 115) achieves these two requirements by specifying minimum and maximum stopping distances. It requires that the stopping distance of the escalator be within the following values, at the two extremes of loading. The minimum values are for unloaded escalators and the maximum values are for downward moving loaded escalators, and are shown in Table 1 for various values of speed. However, it is not clear how these stopping distance requirements translate into deceleration or jerk values (as we cannot assume uniform acceleration during the stop). Moreover, there is no evidence that compliance with those minimum values of stopping distance will ensure safe stopping for passengers in an emergency.

The draft for public comment of EN 115 (prEN 115:2005) suggests the following changes:

So if the second of these proposed changes is accepted, the criterion for limiting the value of deceleration becomes explicitly set.

The American code (ASME, 1993), on the other hand, specifies the performance requirements in terms of maximum values of acceleration (or deceleration in this case). It stipulates a maximum deceleration of 3 ft/s2 (0.9144 m/s² )(A17.1: Clause 804.3b, ASME, 1993). The standard does not specify any maximum stopping distance.

Previous work carried out by the Health and Safety Laboratory sponsored by the HMRI (Her Majesty’s Railway Inspectorate) in the United Kingdom, (Pittard, 1995; Swift & Brennan, 1995; Pittard & Swift, 1995) conducted measurements of acceleration on three London Underground3) escalators at various loading conditions and various combinations of brakes applying. It concluded that the worst case for acceleration values is when the escalator is unloaded and travelling in the down direction [“For descending stops, higher accelerations can be expected with unloaded escalators” ..and that “most severe accelerations were produced by an unloaded escalator stopping in the downward direction.” (Swift & Brennan, 1995, p8)]. The study then used the values from the measurements as inputs to a passenger modelling system to decide the effect of these acceleration values on passengers (Pittard & Swift, 1995). It divided the passengers into the categories “attentive” and “inattentive”, and found that the inattentive passenger could fall and cause another passenger to fall under the most severe acceleration conditions. The study does not discuss jerk as a factor, but emphasisis the importance of acceleration (or deceleration in this case), and its duration as the important factors in deciding passenger falls [“..the mean acceleration and the time for which it lasts are likely to be the most important factors” (Swift & Brennan, 1995)].

Fruin points to the importance of both acceleration and jerk: “For escalators, the primary concern is deceleration and rate of change deceleration (“jerk”) associated with a sudden unexpected stop” (Fruin, 1988).

Work carried out in the railway sector suggests suitable limits to prevent passenger falls on trains. The following “acceleration and jerk levels are suggested as thresholds above which a typical standing passenger with a good hand hold would fall over and risk injury [12]:

These sources above show that the most two important criteria for evaluating the quality of a stop are: deceleration and jerk. However, the studies do not provide:

1. Recommended limits for the values of acceleration and jerk.

2. The relative importance of the two variables in affecting the quality of a stop.

3. The quantitative relationship between the value of acceleration and jerk and the probability of a passenger fall.

In order to be able to establish a relationship between the risk of a passenger fall and the kinematics of the stopping escalator , two approaches are possible, described below:

1. Analytical method: This method relies on theoretical modelling, whereby the model of the human body is linked to the kinematics of the stopping escalator to assess the probabliy of falling under different kinematic scenarios. DYNAMAN is a model that can be used to carry out such modelling (“….is a finite element model of a human based on the DYNA3D code. Any size of adult or child can be modelled and the anatomical joints can be assigned any characteristic. Hence the model can represent the various levels of human attentiveness and strength, such as a fully braced or unconscious person..” [10]). This has not been used in this paper , but it is hoped that it can be used in future work to corroborate the work in this paper.

2. Empirical method: This method is based on the use of human subjects who would ride the escalators during the stop and provide a subjective assessement of the quality of the stop and their assessment of the risk of falling. This method has been used in this piece of work. An example of this empirical approach can be found in the work shown in [11] where experimental tests on subjects were used to find there perception threshold of movement in relation to age and other factors. Another example is shown in [14] where subjects were asked to assess the quality of a stop on an escalator, but the number of tests was limited and the study drew no conclusions from the results.

The Empirical method used in this piece work used the following:

1. An escalator: An escalator on London Underground on Hyde Park Corner Station was used for this (always travelling in the down direction). An adjacent escalator was also available for use.

2. A method by which the kinematics of the stopping escalator can be varied (especially the acceleration and jerk). In the case of this escalator was fitted with a closed loop hydraulic braking system. It allowed the user to vary the value of the deceleration for each stop. The jerk was not varied intentionally, but varied randomly in response to the escalator stop and the variation in the value of the deceleration. It is important to note that it was possible to stop the escalator via a variety of methods as follows:

a. VVVF (variable voltage variable frequency stop): This uses the electrical drive that drives the escalator to stop it in an electrical manner.

b. The escalator has two mechanical (hydraulically lifted spring applied) brakes, an operational brake and an auxiliary brake. It is possible to stop the escalator using one of the brakes, the other brake or both brakes. This provides three different qualities of stop.

c. It is also possible to stop the escalator using a controlled release of the hydraulic pressure in the operational mechanical brake (hydraulically lifted spring applied). This is referred to as ‘intelligent hydraulic braking’, as the amount of hydraulic pressure is reduced or increased to achieve a smooth stopping of the escalator in response to closed loop speed feedback.

3. Passengers willing to ride the escalator during the stop and provide a subjective assessment of the quality of the ride. Around 10 subjects were used for this.

4. A method of recording the speed-time profile of the stop and extracting the deceleration and jerk from it. A measurement unit was used (eva625 with a handheld tacho-wheel).

Fifteen tests were repeated as follows:

1. The escalator was started up in the down direction.

2. Once the escalator attained its full speed, the subjects were asked to ride the escalator (the escalator was run in the down direction as this is the more critical direction for falling).

3. The subjects were told when a stop would take place, so they would be classified as attentive passengers, as they would be expecting the stop and would thus be holding the handrail and facing the correct direction of travel.

4. Once the escalator was stopped, the subjects were asked to rate the stop on a scale from 1 (very bad) to 10 (very good). The average of all ratings from the subjects was taken.

5. The speed-time profile of the stopping escalator was recorded using a handheld tacho-wheel measurement device.

6. The subjects would then get off, and steps 1 to 5 above would be repeated. This was 15 times.

7. In order to ‘calibrate’ the subjects’ assessment, the escalator was first stopped via a VVVF stop. The subjects were asked to rate this as 10 (best stop) on a scale from 1 to 10.

8. The escalator was then stopped using the two mechanical brakes acting together as this was expected to be the worst stop. The subjects were asked to rate this as a 1 (worst stop) on a scale from 1 to 10.

9. Another 13 stops then followed with different setting (but not in any specific order so that the subjects are not ‘guided ’ to the answer) and the subjects were asked to rate the stops.

One of the tests was carried out on the adjacent escalator (HPC2). QBC is an acronym for Quick Braking Contact which is a function in the intelligent braking system . The main parameter that was changed in the intelligent braking hydraulic system was the target stopping time (1.7 seconds in some cases; 2 seconds in others).

The results of all the 15 tests are shown in Table below whereby the data has been baesd on tests carried out in [13]. For each stop three kinematics parameters were extracted: maximum jerk value (pre-ac-celeration-peak); maximum jerk value (post-acceleration peak) and maximum deceleration value. In addition, for each stop, the average of the subjects’ stopping comfort parameters is shown (1 to 10).

These three parameters can be better understood a plot of the stopping deceleration curve as well as the jerk curve on the same time axis. The convention is to assume the speed is negative in the down direction, and hence the acceleration is positive in this case.

The speed is sampled at 64 samples per second. The acceleration is derived from the speed curve by differentiation and then low pass filtering the result at 4 Hz. The jerk is then derived from the acceleration curve by differentiation and no filtering (one sample every 15.625 milliseconds).

The two values of jerk are the maximum values of jerk that are attained before the deceleration attains its peak; and after the acceleration attains its peak (hence the terms pre-acceleration-peak and postacceleration-peak).

It is worth emphasizing that it is the ‘maximum’ value of deceleration and jerk rather than the ‘average’ value that are used in this case.

Single variable regression using MS Excel was used to find the correlation coefficients between the maximum deceleration during the stop, the maximum jerk (pre-peak) and maximum value of jerk (post-peak) in relation to the subjective stopping comfort index (1 to 10).

The regression results for maximum deceleration and subjective stopping comfort index are shown in Figure 2, that shows an exponential relationship and a value of R² of around 91.3 % (i.e., the value of maximum deceleration explains 91.3 % in the variation of the subjective stopping comfort). Figure 3 shows the exponential regression between the preacc-peak jerk value and the value of subjective stopping comfort index. The square of the coefficient of correlation in this case (R²) is around 87.5 %. Figure 4 shows the linear correlation between the post-acc-peak jerk value and the subjective stopping comfort index, with an R² value of 66.3 %.

The decision to use linear or non-linear regression was based on the function that provided the best correlation.

It is clear form these results that the strongest indicators of stopping comfort are the values of the maximum deceleration during the stop and the maximum jerk value (pre-acceleration-peak). The correlation with the post-accelerationpeak jerk value is poor (66.3 %).

The physical significance of this can be explained as follows. The most important factor that will cause a person to fall is the force pushing him/her forward in the direction of travel. This is best represented by the maximum (not average) value of the deceleration. However, when someone has more time to prepare himself/ herself for this value of deceleration, he/ she stands a better chance of preventing the fall. So with a low value of pre-acceleration- peak jerk, the maximum deceleration takes a longer time to be attained and thus gives the passenger more time to prepare. A high value of pre-acceleration- peak jerk provides little time for the maximum deceleration value to be attained and hence presents the person with less time to prepare and prevent a fall.

The post-acceleration-peak jerk value is less relevant in this case, as it happens after the maximum deceleration value has been attained and the passenger is already fully holding onto the handrail.

In order to further understand the relative importance of these two variables in predicting the subjective stopping comfort index, multiple variable linear regression was carried out (using the subjective stopping comfort index as the dependent variable, and the maximum value of deceleration and the maximum value of jerk [pre-acc-peak] as independent variables). This resulted in the following relationship with a value of R² of 91.2 %:

Although the coefficient of correlation in this case is not different from the one arrived at for the relationship between the maximum deceleration value and the subjective stopping comfort index, it leads us to a very important result. The relative importance of the maximum value of deceleration is much higher than the maximum value of jerk (pre-acc-peak) (by around a factor of 10). This shows that the efforts for improving the quality of the stop should be concentrated on reducing the value of the maximum deceleration, more that the efforts directed at the value of jerk (although some improvement will be gained from controlling the value of maximum jerk).

This formula also gives us a tool to be able to predict the quality of a stop by measuring the maximum deceleration and the maximum value of the jerk (preacc-peak).

A number of trails have been carried out on a stopping escalator to understand the relationship between the stopping comfort experienced by the passengers and the kinematics of the stopping escalator. The subjects were asked to rate the quality of each stop on a scale from 1 to 10 (1 being very bad; 10 being very good).

It has been found that the escalator stopping comfort experienced by passengers can be predicted from the maximum values of deceleration and jerk of the stop. It is the pre-acceleration-peak jerk value that is more important that the post-acceleration- peak jerk value. The maximum value of deceleration is more significant that the maximum value of jerk (pre-acceleration-peak).

Using single variable regression it has been found that the maximum value of deceleration against the subjective stopping comfort index exhibits an R² value of around 91 %; the maximum value of jerk (pre-acceleration-peak) exhibits an R² value of around 87 %.

Using two variable linear regression with the subjective stopping comfort index as the dependent variable and the maximum value of deceleration and jerk (preacceleration-peak) as the independent variables, exhibits an R² value of around 91 %, but more importantly, confirms the relative importance of deceleration against jerk (coefficients of linear equation 10:1).

Further work is needed in the following areas:

1. The number of tests carried out is limited (15 tests). More tests are needed to further reinforce the results.

2. The tests have asked the subjects about the stopping comfort, but these do not necessarily translate into probability of passenger falls. Ideally, the data should be based on actual passenger falls on escalators. However, this requires historical data, as well as the kinematics of the escalator on which the fall occurred at that specific time. This is impractical.

3. An alternative to historical data would be to carry out modeling on a human body model standing on an escalator and assess the probability of a fall based on various values of maximum deceleration, jerk and body conditions (attentive/inattentive; holding onto the handrail…).

4. Non-linear multiple variable regression could be carried out to see if this provides a better value of R².

5. With the use of an electrical braking system, more precise control is possible on both deceleration and jerk independently. This will allow a more accurate analysis of the effect of each variable independently.