Gully malfunction detection

Abstract

A gully malfunction is detected when there is a deviation from a detected water level variation inside the gully and an expected water level variation associated with obtained rain information. The rain information may be obtained e.g. from rain sensors, or be deduced from other gullies, located near the gully of interest.

Description

Field of the invention

[0001] The present invention relates to level detection in storm drains, also referred to as drain gullies or gully wells. Specifically, the invention relates to determining if a gully is not working properly, e.g. because its cover has been blocked, or it has been filled with silt and debris.Background of the invention

[0002] Gully wells (storm drains) take care of rainwater (storm water) on the streets. Such wells or drains typically have a sand trap where sand and silt can be collected rather than following the flow of water into the drain pipes. If the sand trap, gully cover, or other part of the gully becomes clogged with leaves, gravel and debris, the water cannot be drained. This rubble needs to be cleaned up to create a free flow again.

[0003] When the silt traps are not emptied in time they block the outflow of the rainwater. This causes the water levels to increase in the gully to the point where the water level overflows above the gully cover and leads to a flooding. This can cause inconvenience and danger to passing traffic. Furthermore, when the silt level increases part of the silt moves from the silt trap and into the connecting pipes, causing blockages downstream in the system and damages and erosion of the pipelines.

[0004] The applicant, AXsensor, has technology for continuously measuring the water level in the gully well. When the sand trap becomes full and blocks the outflow, the sensor will warn of elevated levels and trigger an alarm to empty the sand trap. AXsensor's main method of identifying when the silt trap is to be emptied is to continuously monitor the water levels inside the gully.

[0005] However, a partly filled silt trap may be difficult to detect based only on level information. Also, if the gully cover is blocked by debris, the storm water cannot run off into the gully and a flooding will occur. This will not be detected by a level meter inside the gully. Therefore, there also needs to be a method of identifying when the gully cover is blocked.General disclosure of the invention

[0006] Aspects of the present invention are defined by the independent claims. Embodiments of the invention are further defined by the dependent claims.

[0007] According to the invention, a gully malfunction is detected when there is a deviation from a detected water level variation inside the gully and an expected water level variation associated with obtained rain information. The rain information may be obtained e.g. from rain sensors, or be deduced from other gullies, located near the gully of interest.

[0008] The detected deviations may be of various types. For example, they can relate to a baseline level which is the equilibrium for the water level. In a normal flow, after every rainfall, the water level should come back to the same equilibrium. If this baseline level is increased, this may be an indication of malfunction. As will be discussed below, a malfunction may also be detected based one the rate of change in the water level during the rainfall.Brief description of the drawings

[0009] The present invention will be described in more detail with reference to the appended drawings, showing currently preferred embodiments of the invention. Figures 1a-1c show storm drains. Figures 2-4 show rainfall and expected and actual water level variation in a gully, according to an embodiment of the invention. Figure 5 is a satellite view of a town with gullies marked. Figures 6-8 show rainfall and expected and actual water level variation in a gully, according to another embodiment of the invention. Figure 9 illustrates modelling of a gully. Figure 10 illustrates a wake-up function of a system according to an embodiment of the invention. Detailed description of preferred embodiments

[0010] Figure 1a shows a gully (storm drain) 1 (gully) with unrestricted flow through the outlet 2. The gully has a water lock 3, arranged above a silt trap 4, containing some silt (sand, debris, etc.). A level meter 6 (e.g. from AXsensor) is arranged in the drain 1 to allow measurement of the water level inside the gully 1.

[0011] In figure 1b, the silt trap 3 is filled with silt to such an extent that it blocks the water lock 3, thereby blocking the flow of water. As a result, the water level inside the gully 1 will increase.

[0012] Figure 1c shows a different type of gully. In this case, the water lock is formed by a drain pipe which forms a knee with a high point. Silt, sand and debris will collect in the entrance of the drain pipe gradually blocking the flow of water.

[0013] In the present disclosure, the basic principle is to recognize deviations from the normal course of level changes that occur in gullies during rainfall.

[0014] According to a first principle, deviations from the normal course of level change that occurs during rain in gullies are recognized by studying rain data and comparing with water level results.

[0015] In a first example, illustrated in figure 2, it is identified when the sand level starts to choke the outlet of the storm water to optimize the triggering time of emptying the silt trap.

[0016] The first curve 21 shows three types of detected rain fall to illustrate that different rain patterns give rise to different flow in the gully. The second curve 22 shows level change as a function of time for a "healthy" well in connection with three different rain scenarios. The third curve 23 shows level change in an obstructed gully, relative to normal level change for a "healthy" well in connection with the same three rain scenarios. Given information about the type and quantity of rain fall, the system can identify different peak levels and fall times relative to "healthy" peak levels and fall times. One relevant measure is the fall time from the peak level to a lower level such as a given ratio of the peak level, e.g., 75% of the peak level. As shown in figure 2, there is a difference h between expected peak levels and the detected peak levels, and rate of change from the peak level to a predefined level is different. These detected deviations may trigger an alarm.

[0017] In a second example, illustrated in figure 3, it is detected that gully covers have become clogged by correlating with rain data.

[0018] Again, the first curve 31 shows detected rain fall to illustrate that different rain patterns give rise to different flow in the gully. The second curve 32 shows level change as a function of time for a "healthy" well in connection with two different rain scenarios. The third curve 33 shows an obstructed gully, where there is no level increase. Based on an indication of rain fall in the area, a level increase is exapted. Since there is no level activity a deviation alarm is triggered. The fourth curve 34 shows a similar scenario, but this time with a partially clogged gully cover that results in level increase but as high as expected, which therefore triggers a deviation.

[0019] In a third example, illustrated in figure 4, drift of the equilibrium level over time is detected.

[0020] Again, the first curve 41 shows detected rain fall to illustrate that different rain patterns give rise to different flow in the gully. The second curve 42 shows level change as function of time in a "healthy" gully. Over time, without rain, the level comes down to the original equilibrium position, or "no-rain" level, which is the normal procedure. The "no-rain" level can be defined as the level a given time after the end of the rain fall. In the third curve 43 a deviation has occurred which causes the equilibrium position, or no-rain level" 44a, 44b to drift from the original value over time. An alarm can then be triggered to check cause for this.

[0021] Malfunction detection according to the first principle requires information about rain fall. Such information can be obtained from a rain sensor in the vicinity of the gully. However, it may be difficult or overly costly to provide rain sensors at each gully.

[0022] According to a second principle, detection of rain is not required. Instead, deviations from the normal level change that occurs during rain in gullies are recognized by studying and comparing with the level change behavior of neighboring gullies. One way to solve this is through Machine Learning and Artificial Intelligence where, with the help of key parameters, the system is "taught" to register deviations. Another way is to use angles and "runoff times" in a similar manner as discussed above. The behavior in neighboring gullies can be seen as an indication of rain fall, in the absence of actual rain fall detection.

[0023] The system will compare the level change behavior relative to several neighbors. These neighbors are defined by specific distances relative to each other and each well is given a unique radius of neighbors with which it is compared in a geographic catchment area. Furthermore, each sensor / neighbor is tested for criteria parameters to determine if a neighbor is approved to be compared to. For example, a maximum distance to the nearest neighbor is set at x meters. If the neighbor is too far away, the system will not compare with this neighbor. We also reproduce rainfall by using sensor data to mimic rain pattern (as it dynamically travels over time) and to add knowledge if there is rain in the area. When reproducing rainfall by using sensor data to mimic rain pattern (as it dynamically travels over time) the criteria parameters to determine if a neighbor is approved can become dynamic and instead of a circle with a radius a "radar" shaped chart can instead determine the neighboring boundaries. This is indicated in figure 5.

[0024] In a first example, illustrated in figure 6, it is identified when the silt level starts to choke the outlet of the storm water to optimize the triggering time of emptying the silt trap by comparing levels to neighboring gullies.

[0025] The first curve 61 shows normal level change as a function of time for a "healthy" gully (Gully II) in connection with three different rain scenarios. The second curve 62 shows normal level change for a second "healthy" gully (Gully I) in connection with the same three rain scenarios. This second gully has different runoff times even if silt trap is empty due to different dimensions etc. The third curve 63 shows detected level change in the second gully, which now has been filled with silt. Even though there is no information about actual rain fall, the level variations in other gullies, such as in the first gully, can be used to estimate an expected level variation for the second gully. The detected level variation can be compared to this estimated "normal" level change, and deviations can be detected in a similar manner as described in relation to figure 2.

[0026] In a second example, illustrated in figure 7, it is identified when gully covers become clogged by comparing levels to neighboring gullies.

[0027] Curves 71 and 72 show normal level change for two "healthy" gullies in connection with two different rain scenarios. Again, there are slightly different drain times despite an empty sand trap due to differences in design. The third curve 73 shows detected level change in the second gully (Gully I), which now has an obstructed cover, e.g. by leaves and debris. Comparison is made against neighboring gullies such as the first gully, which shows that neighboring gullies detect a rainfall. The absence of any level rise in the second gully indicates a deviation and possible malfunction. Curve 74 shows a similar scenario, but this time with a partially clogged gully cover that results in level increase but not high enough. The deviation in average level is detected and an alarm can be triggered.

[0028] In a third example, illustrated in figure 8, drift of the equilibrium level over time is identified by comparing levels to neighboring gullies.

[0029] The first curve 81 shows level variation in a healthy gully. Overtime, the level comes down to the original equilibrium position, or "no-rain" level, which is the normal procedure. In the second curve 82, a deviation has occurred, causing the equilibrium position (no-rain level) to drift from the original value over time. As discussed above, with reference to figure 4, an alarm may be triggered to check the cause for this.

[0030] In the illustrated embodiment, a check can be made to neighbor gully, where the equilibrium level (no-rain level) is stable. Curve 83 shows level variation for such a gully (Gully II). Furthermore, with this comparison to a neighbor gully it is also possible to assess that it has rained.

[0031] Regarding time constants, more info / data and / or setting up models is required. More generally, you should be able to approximate a well as a 2nd order system. Probably a higher order system which is also nonlinear system.

[0032] In a system based on a well that is filled with water and has an outlet a little way up the well. When the outlet is fully open, the normal time constant is obtained. When the outlet begins to be choked (by sand clogging), a new time constant is obtained. When completely plugged without outflow, the time constant approaches infinity and the well overflows. The angles are a measure of the time constant (can also be defined as when different threshold values are reached).

[0033] Figure 9 shows an example of a well outlet with different diameters. The x-axis is visualized as flow in (m 3< / s). In our models the liquid level measured height (h) is displayed in meters.

[0034] In some embodiments, a wake-up function is useful to obtain the necessary level data from the gullies. To take decisions based on the graphs, the graphs need to be reliable. Within the IoT technology there is a balancing between saving battery consumption and the frequency of data points created. With our wake-up function technology, illustrated in figure 10, we can have the sensor in a low sampling frequency state when there is no activity in the gully and then activate a higher sampling frequency of level measurement when there is an activity. The higher sampling frequency is required to make conclusions and predictions based on that data. The wake-up functionality ensures long battery life and sufficient data sampling frequency.

[0035] The person skilled in the art realizes that the present invention by no means is limited to the preferred embodiments described above. On the contrary, many modifications and variations are possible within the scope of the appended claims.

Claims

1. A method for determining malfunction of a gully, comprising: obtaining rain information relating to rainfall over a given time period, determining an expected water level variation inside the gully based on the rain information, detecting a water level variation inside the gully during the time period, comparing the detected water level variation with the expected water level variation, and identifying a gully malfunction based on a deviation between the detected water level variation and the expected water level variation.

2. The method according to claim 1, wherein the rain information is a measured rainfall over time obtained from a rain sensor in a vicinity of the gully.

3. The method according to claim 1, wherein the rain information is predicted rainfall from a weather forecast.

4. The method according to claim 1, wherein the rain information is obtained by: identifying a set of neighbor gullies located in a neighborhood surrounding the gully, for each neighbor gully, obtaining a relationship between rainfall and water level variation, detecting a water level variation in the neighbor gullies during the time period, and obtaining the rain information based on the detected water level variation and said relationship for each neighbor gully.

5. The method according to claim 1, wherein the rain information is obtained by: identifying a set of neighbor gullies located in a neighborhood surrounding the gully, detecting a water level variation in the neighbor gullies during the time period, and obtaining the rain information using an appropriately trained neural network and the detected water level variation in the neighboring gullies.

6. The method according to claim 4 or 5, wherein the rain information is an estimated rainfall over time during the time period.

7. The method according to any one of the preceding claims, wherein the expected water level variation is determined using a mathematical model of water flow through the gully.

8. The method according to claim 6, wherein the mathematical model includes a transfer function, such as a second order or higher transfer function.

9. The method according to any one of the preceding claims, wherein said deviation is a deviation between peak water levels or average water level.

10. The method according to any one of the preceding claims, wherein said deviation is a deviation between a fall time from a peak water level to a lower water level.

11. The method according to claim 10, wherein the lower water level is a predefined ratio of the peak water level.

12. The method according to any one of the preceding claims, wherein said deviation is a deviation between a no-rain level, reached a predefined time after an end of the rainfall.

13. The method according to any one of the preceding claims, wherein the water level variation inside the gully is measured using an acoustic sensor or a radar level sensor.

14. The method according to any one of the preceding claims, wherein the step of detecting water level variation is performed with a first sampling frequency during an idle period, and with a second sampling frequency during a wake period, and wherein the second sampling frequency is higher than the first sampling frequency.

15. A gully malfunction detection system comprising: a water level sensor arranged to detect a water level variation inside the gully, means for obtaining rain information relating to rainfall over a given time period, processing circuitry connected to the water level sensor and configured to: determine an expected water level variation inside the gully based on the rain information, compare the detected water level variation with the expected water level variation, and identify a gully malfunction based on a deviation between the detected water level variation and the expected water level variation

Images