Determining the blockage status of a roof gutter

Abstract

Method for determining the blockage status of a gutter (1), comprising the following process steps: S1) in a training phase while ensuring a completely unblocked gutter (1): S1a) successive recording of a fill level value describing the fill level of the gutter (1) and storing each recorded fill level value, resulting in a fill level value reference time series, and S1b) successive recording of a precipitation value indicating the amount of precipitation per unit time and area at the location of the gutter (1), and storing each recorded precipitation value, resulting in a precipitation value reference time series, and S2) in a monitoring phase following the training phase without ensuring a completely unblocked gutter (1): S2a) successive recording of the fill level value and storing each recorded fill level value,so that a series of actual fill levels is obtained, S2b) recording the precipitation value successively and storing each recorded precipitation value, so that a series of actual precipitation values ​​is obtained, and S3) comparing the series of actual fill levels and the series of actual precipitation values ​​with the reference series of fill levels and the reference series of precipitation values ​​and determining the blockage status of the gutter (1) on the basis of this comparison.

Description

The invention relates to a method for determining a blockage state of a roof channel.Gutters are usually arranged on a roof edge of a building and serve to collect and discharge rainwater in a targeted manner. A roof channel is usually made of weather-resistant materials such as metal or plastic. The main function is to collect the rainwater accumulating on the roof and to conduct it safely to downpipes or gullys. This enables a controlled discharge of the water and contributes to avoiding water damage to construction substance, foundations and masonry of the building. Furthermore, gutters minimize the risk of floor erosion, since the water is discharged in a targeted and controlled manner. For optimum operation, periodic maintenance is important to ensure that the gutter is free of debris such as leaves and branches so that the effluent of rainwater is not obstructed by blockages in the gutter.CN 111640284 A describes a wireless roof drainage monitoring system and an early warning method thereof. The monitoring system includes a roof precipitation sensor, a gutter water level sensor, a downpipe water flow sensor, a wireless communication module, and a server. The roof precipitation sensor is disposed on an outer roof, the gutter water level sensor is disposed in a roof drainage gutter, and the downcomer water flow sensor is disposed on the surface or an outlet of a gutter downcomer. The roof precipitation sensor, the gutter water level sensor, and the downpipe water flow sensor are all connected to the wireless communication module, and the wireless communication module is connected to the server via a wireless network. Precipitation, water level and drain rate data can be obtained in real time. Thus, a blockage of the roof channel is reported in real time during a rain event. However, immediately during a rain event, it is often not possible to detect an actual blockage of the roof gutter. An immediate real-time report of roof gutter plugging therefore frequently leads to false alarms.US 2020 / 0048910 A1 relates to a monitoring system for gutter systems which detects various operating states with the aid of sensors. For example, water levels, temperature, amount of precipitation, UV radiation or the degree of soiling of the gutter are detected. The sensors can operate optically, capacitively, acoustically or on the basis of electromagnetic waves. The data are transmitted wirelessly and can be used to predict the need for maintenance or to control further building functions. The system can be integrated into new as well as existing gutters.CN 115168520 A describes a system and method for the long-term monitoring of wastewater networks on the basis of the detection of liquid levels and rain intensity. Sensors are used to detect anomalies at sampling points, for example by comparison with historical data or thresholds. An early warning system gives indications of disturbances, such as blockages, and takes into account different operating modes (sun day / rain day). The data is mapped and analyzed by AI models (e.g., LSTM) to improve the stability of the sewage network and allow continuous monitoring.Proceeding from this, it is the object of the present invention to provide a reliable and efficient method for determining a state of blockage of a roof channel.This object is achieved by the subject matter of claim 1. Preferred refinements are found in the dependent claims.According to the invention, a method for determining a state of blockage of a roof channel is thus provided, having the following method steps: (1) in a training phase while ensuring a completely unfloded roof channel: temporally consecutive recording of a fill level value describing the fill level of the roof channel and storage of the respectively recorded fill level value, so that a fill level value reference time series results, and temporally consecutive recording of a precipitation value which specifies the quantity of precipitation per time and area at the location of the roof channel, and storage of the respectively recorded precipitation value, so that a precipitation value reference time series results, and (2) in a monitoring phase temporally downstream of the training phase without ensuring a completely unflopped roof channel: temporally consecutive recording of the fill level value and storage of the respectively recorded fill level value, so that a fill level value actual time series results, successively in time detecting the precipitation value and storing the respective detected precipitation value, so that a precipitation value actual time series results, and matching the fill level value actual time series and the precipitation value actual time series with the fill level value reference time series and the precipitation value reference time series and determining the blocking state of the roof gutter on the basis of this matching.The invention provides that a certain number of precipitation events have taken place in the training phase before the monitoring phase can be entered. As a result, the fill level value reference time series contains a certain number of possible fill level values of the completely non-plugged roof channel, which can then be correspondingly assigned to precipitation values of the completely non-plugged roof channel in the precipitation value reference time series. The precipitation values are contained in terms of the amount of precipitation per time and area at the location of the roof gutter. The time series thus generally show the filling levels of the roof gutter or precipitation values in a specific temporal sequence. In this case, periods of the training phases for the fill level value reference time series and the precipitation value reference time series are preferably identical, such that the values can be directly correlated with one another. In this context, it is preferred that the time series values are successively acquired at intervals of 5 minutes during a precipitation event and thereafter at intervals of 15 minutes.It is further preferred that the fill level values of the fill level value reference time series are determined taking into account dimensions of the gutter, such as the gutter width, length, height and / or inclination. For this purpose, the fill level values can be detected with a fill level sensor, which has, for example, a water level sensor in a riser pipe. Alternatively, sensors can also be used which use capacitive pads for measuring the fill level. The current precipitation values can be obtained accordingly with a precipitation sensor and / or with weather data on site via GPS localization.It is essential to the invention that the fill level value reference time series detected in the training phase and the precipitation value reference time series are matched to the fill level value actual time series and the precipitation value actual time series in the monitoring phase. The reference time series provide accurate knowledge of the behavior of the gutter at and shortly after a precipitation event, such that a plugging condition of the gutter is enabled based on matching the reference time series with the actual time series. If the comparison of the fill level values of the fill level value actual time series with the fill level values of the fill level value reference time series in the case of a comparable precipitation event reveals that the fill level values of the fill level values actual time series during and after the precipitation event are generally higher or above a predetermined limit value than those of the fill level value reference time series during and after the comparable precipitation event, the roof channel is blocked. In this case, the comparable precipitation event is determined by the comparison of the precipitation value reference time series with the precipitation value actual time series. It is therefore possible only on the basis of the reference time series determined in the training phase to determine a blocking state for the specific roof channel. Advantageously, the accuracy of the determination of the blockage state can be increased as required by correspondingly detailed training phases.Generally, the state of clogging may be specified in various ways. According to a preferred development of the invention, however, it is provided that the blocking state is a state from a finite number of states which specify a measure of the blocking of the roof channel with respect to the outflow of precipitation water. For example, the blockage of a gutter can be measured with four different states: "unfloded", "lightly plugged", "heavily plugged", and "fully plugged". It can thus be established immediately after the state has become known in a uniform and reliable manner whether and when there is a need for action or the roof channel should be cleaned. In this context, however, the blocked state according to a further preferred development of the invention can also indicate the extent to which the outflow of precipitation water from the roof channel is inhibited. This data can then be given, for example, as a percentage: 0%=no inhibition until 100%=no outflow any longer. The percentage data here enables a more accurate metering of the maintenance intervals or cleaning intervals and gives the user a more accurate insight into the development of the blockage states.According to a preferred development of the invention, the method for determining the state of blockage of the roof channel comprises the following further step: in the training phase, wherein the roof channel is brought into different states of blockage: temporally consecutive recording and storage of the fill level values in the fill level value reference time series and the precipitation values in the precipitation value reference time series for the respective state of blockage of the roof channel. It is provided that the individual blocking states in the roof channel are intentionally brought about in order to obtain corresponding fill level values and precipitation values. Because these blocking states are already contained in the fill level value reference time series and in the precipitation value reference time series, the corresponding blocking states can be detected with higher accuracy when comparing the actual time series with the reference time series.According to a further preferred development of the invention, it is provided that the method for determining the state of blockage of the roof channel has the following further steps: in the training phase, ensuring the completely unblocked roof channel: temporally consecutive acquisition of at least one meteorological value, wherein the at least one meteorological value is an air humidity and / or a temperature and / or a wind direction and / or a wind intensity, and storing the respectively acquired meteorological value, so that a meteorological value reference time series results, and in the monitoring phase, without ensuring the completely unblocked roof channel: temporally consecutive acquisition of the meteorological value, and storing the respectively acquired meteorological value, so that a meteorological value actual time series results, and matching the fill level value actual time series, the precipitation value actual time series and the meteorological value actual time series with the fill level value reference time series, the precipitation value reference time series and the meteorological value reference time series and determining the blocking state of the roof gutter on the basis of this matching. By including the meteorological values matched to the fill level values, accuracy in determining plugging conditions may be further increased. Simultaneously, false alarms such as, for example, blockages due to snow or the like can also be detected. In addition, additional meteorological values such as wind speed or generally meteorological data available from a weather service may be included. In this way, the precipitation event can be better described. These values can originate from an external service, the unit itself, but also from a further technical system, such as a smart card weather station.In principle, the blocking state can be reported at regular intervals. According to a preferred development of the invention, however, it is provided that a message is automatically carried out in the monitoring phase when a predetermined blocking state is exceeded. This message can then be received by a service, e.g. a rooftop company, and / or the user, e.g. a home owner, who can then remove the blockage corresponding to the message. Because the report is made only when a predefined blocking state is exceeded, unnecessary maintenance work on the roof channel is avoided. The gutter is efficiently maintained.In principle, the training phase and the monitoring phase can be of any length. According to a preferred development of the invention, however, it is provided that in the training phase and in the monitoring phase, the detection of the fill level values is not ended before a fill level of zero has been detected. As a result, it is also possible to detect a blockage which occurs only immediately after a precipitation event. For example, if leaves are flushed into a downcomer fluidly connected to the gutter and clog the downcomer, small residual amounts of precipitate remain in the gutter. These are detected and can be reported accordingly.It is possible to determine the clogging state in various ways. According to a preferred development, however, it is provided that the blocking state of the roof gutter is determined with the aid of an artificial neural network. The neural network is preferably a recurrent neural network. According to a particularly preferred development of the invention, it is further provided that the determination of the blocking state of the roof gutter is carried out with the aid of a time series prediction model. As a result, it is then possible to predict future blockages in the form of the blockage states with a certain probability and correspondingly to output a message even before a blockage state occurs. It is still further preferred that the time-series prediction model is a long short-term memory model.Long short-term memory (LSTM) is a type of recurrent neural network (RNN) that has been developed to solve the disappearance gradient problem in conventional recurrent neural networks. LSTMs are less sensitive to the length of the gaps between relevant information compared to other RNNs and sequence learning methods such as Hidden Markov Models. They are used in various applications such as handwriting recognition, speech recognition, machine translation and healthcare. A typical LSTM consists of a cell and three ports: the input, output and the forget port. These ports regulate the flow of information into and out of the cell. The visitor decides which information is discarded from the previous state, while the entrance gate determines which new information is stored. The output port in turn controls which information is output from the current state. By these mechanisms, the LSTM network can maintain useful long term dependencies for predictions in both current and future time steps.The LSTM model is particularly suitable for creating time-series predictions due to its ability to capture long dependencies and store information over a longer period of time for processing time-series values.Accordingly, the reference time series of the training phase are provided as input to the LSTM model, whereupon the LSTM model recognizes patterns and relationships on the input, which, in the monitoring phase, together with the comparison with the actual time series, make a blockage or an imminent blockage recognized. If filling level value reference time series, precipitation value reference time series and meteorological value reference time series containing filling level values, precipitation values and meteorological values for the blockage states are additionally provided to the LSTM model in the training phase, a high accuracy can be achieved with the LSTM model in the prediction and determination of the blockage state of the roof channel. In this context, in a further preferred development, it is provided that information about interfering objects such as photovoltaic installations, gums and satellite dishes is inserted in the long short-term memory model. Together with the dimensions of the roof channel, the LSTM model is thus customized to the roof channel. The determination of the state of blockage of the roof channel can be effected periodically and / or on demand. As a result, it is possible to decide flexibly, if necessary, how often the determination of the blockage state should be carried out either taking into account an energy saving aspect or a safety aspect.According to the invention, there is also a non-transitory computer-readable storage medium having instructions stored thereon which, when executed on a processor, effect the method described above.The invention further relates to a system having a fill level sensor, a communication interface connected to the fill level sensor for signal transmission, and a computing unit, wherein the communication interface is designed to transmit a fill level value measured with the fill level sensor to the computing unit, and the computing unit is designed to carry out the method for determining the blocking state of the roof gutter, as described above. The signal transmission can be cable-less or cable-bound. The fill level sensor is arranged magnetically or by means of a snap-in device in or on the roof channel. For communication with the computing unit, the communication interface has a microcontroller with a transmitting unit. As already mentioned, the precipitation values can be measured on site with a precipitation value sensor and / or can be called up via a GPS localization of the roof gutter. The same applies to the meteorological values. The corresponding sensors are then likewise connected to the communication interface for signal transmission. The sensors and the communication interface can be operated with a battery, rechargeable battery, photovoltaics or a connection to a domestic power network.The invention is described in more detail below with reference to the drawings on the basis of preferred exemplary embodiments.The drawing shows FIG. 1 schematically shows a flow chart for a method for determining a blocking state of a roof channel according to a preferred exemplary embodiment of the invention, and FIG. 2 schematically illustrates a system for determining a blockage condition in accordance with a preferred embodiment of the invention.FIG. 1 schematically shows a flow chart for a method for determining a blocking state of a roof channel 1 according to a preferred exemplary embodiment of the invention. In a first step S 1, it is provided that in a training phase, ensuring a completely uncopped roof channel 1 in step S 1 a, a fill level value describing the fill level of the roof channel 1 is detected successively in time and the fill level value detected in each case is stored in a fill level value reference time series. The fill level values are measured with a fill level sensor 2 that can be seen in FIG. 2. The fill level sensor 2 has a water level sensor in a riser pipe. The fill level values are linearized taking into account dimensions of the roof channel 1.In parallel, in step S 1 b, the temporally consecutive acquisition of a precipitation value, which indicates the amount of precipitation per time and area at the location of the roof channel 1, and the storage of the respectively acquired precipitation value in a precipitation value reference time series are provided. In addition, in step S 1 d, the temporally consecutive detection of meteorological values takes place. The meteorological values are an air humidity, a temperature, a wind intensity and a wind direction at the location of the roof channel. These meteorological values are stored in a meteorological value reference time series.The training phase also includes introducing various plugging conditions into the roof gutter. The blocked state indicates as a percentage by which extent the discharge of precipitation water from the roof channel 1 is inhibited. The fill level values and the precipitation values are recorded and stored in step S 1 cfor the respective blocking state of the roof channel in the fill level value reference time series and the precipitation value reference time series. It is ensured that several precipitation events have taken place during the training phase.The fill level value reference time series, the precipitation value reference time series and the meteorological value reference time series are passed as input values for training to a long short-term memory (LSTM) model. With the trained LSTM model, in a monitoring phase in step S 2 temporally downstream of the training phase, without ensuring a completely uncopped roof gutter 1 in steps S 2 ato S 2 c, the temporally consecutive acquisition of the fill level value and storage of the respectively acquired fill level value are effected, so that an actual fill level value time series is obtained, the temporally consecutive acquisition and storage of the precipitation value are effected, so that an actual precipitation value time series is obtained, and the temporally consecutive acquisition of the meteorological value, and storage of the respectively acquired meteorological value, so that an actual meteorological value time series is obtained.In a third step S 3, the fill level value actual time series, the precipitation value actual time series and the meteorological value actual time series are matched with the fill level value reference time series, the precipitation value reference time series and the meteorological value reference time series with the aid of the LSTM model. On the basis of this comparison, the LSTM model is then used to determine the blocking state of the roof gutter. At the same time, the LSTM model can be used to predict a future blockage in the form of a future blockage state of the roof gutter, in response to the prediction of the time series values. The state of blockage of the roof channel 1 or the prediction of a future blockage of the roof channel 1 is carried out at regular intervals. These intervals can be defined by a user.Finally, FIG. 2 schematically shows a system for determining a blockage state according to a preferred exemplary embodiment of the invention. A filling level sensor 2 is magnetically fastened in the roof channel 1. The fill level sensor 2 is connected to a communication interface 3 for transmitting the fill level value. A precipitation sensor 7 for detecting a precipitation value, a thermometer 8 for detecting a temperature, a hygrometer 9 for detecting a relative humidity, and a wind meter 10 for detecting wind direction and wind intensity are also connected to the communication interface 3 for transmitting respective values. These values are transmitted to a receiving unit 6 of a computing unit 4 by a transmitting unit 5 connected to the communication interface. Within the computing unit, the LSTM model can then be used to determine the blocking state of the roof channel in response to the transmitted values. The computing unit can be present locally or can be located on an external server. Once a blockage condition of 50% or more is detected, a message is sent to a service service to cause the gutter to clean.List of reference characters1 Roof channel 2 fill level sensor 3 communication interface 4 computing unit 5 transmitting unit 6 receiving unit 7 precipitation sensor 8 thermometer 9 hygrometer 10 wind meter

Claims

Method for determining a state of blockage of a roof channel (1), having the following method steps: S1) in a training phase, ensuring a completely unblocked roof channel (1): S1a), acquiring a fill level value describing the fill level of the roof channel (1) in chronological succession and storing the fill level value acquired in each case, so that a fill level value reference time series results, and S1b), acquiring a precipitation value, which indicates the amount of precipitation per time and area at the location of the roof channel (1), in chronological succession, and storing the precipitation value acquired in each case, so that a precipitation value reference time series results, and S 2) in a monitoring phase temporally downstream of the training phase without ensuring a completely unflopped roof gutter ( 1): S 2 a) temporally consecutive detection of the fill level value and storage of the respectively detected fill level value, so that a fill level value actual time series results, S 2 b) temporally consecutive detection of the precipitate value and storage of the respectively detected precipitate value, so that a precipitate value actual time series results, and S 3) matching the fill level value actual time series and the precipitate value actual time series with the fill level value reference time series and the precipitate value reference time series and determining the state of clogging of the roof gutter ( 1) on the basis of this matching.The method according to claim 1, wherein the plugging state is one of a finite number of states indicating a measure of the plugging of the roof gutter (1) with respect to the discharge of precipitation water.The method according to claim 1, wherein the plugging state indicates to what extent the discharge of rainwater from the gutter (1) is inhibited.Method according to one of claims 2 or 3, comprising the following further step: S1c) in the training phase, wherein the roof channel (1) is brought into different blocking states: temporally consecutive recording and storage of the fill level values in the fill level value reference time series and the precipitation values in the precipitation value reference time series for the respective blocking state of the roof channel (1).Method according to one of the preceding claims, comprising the following further steps: in the training phase, ensuring the completely uncopped roof gutter (1): S1d), successively in time, acquiring at least one meteorological value, wherein the at least one meteorological value is an air humidity and / or a temperature and / or a wind direction and / or a wind strength, and storing the respectively acquired meteorological value, so that a meteorological value reference time series results, and in the monitoring phase, without ensuring the completely uncopped roof gutter (1): S2c), successively in time, acquiring the meteorological value, and storing the respectively acquired meteorological value, so that an actual meteorological value time series results, and S 3) matching the actual level value time series, the actual precipitation value time series, and the actual meteorological value time series with the reference level value time series, the reference precipitation value time series, and the reference meteorological value time series, and determining the clogging state of the gutter ( 1) based on this matching.Method according to one of the preceding claims, wherein a message is automatically carried out in the monitoring phase when a predetermined blocking state is exceeded.Method according to one of the preceding claims, wherein in the training phase and in the monitoring phase the detection of the fill level values is not ended before a fill level of zero has been detected.Method according to one of the preceding claims, wherein the blocking state of the roof gutter (1) is determined with the aid of an artificial neural network.Method according to one of the preceding claims, wherein the determination of the state of blockage of the roof channel (1) is carried out with the aid of a time series prediction model.The method of claim 9, wherein the time-series prediction model is a long short-term memory model.Method according to one of the preceding claims, wherein the determination of the state of blockage of the roof channel (1) takes place periodically and / or on demand.A non-transitory computer readable storage medium having instructions stored thereon that when executed on a processor cause a method according to any preceding claim.System having a fill level sensor (2), a communication interface (3) connected to the fill level sensor (2) for signal transmission, and a computing unit (4), wherein the communication interface (3) is designed to transmit a fill level value measured with the fill level sensor (2) to the computing unit (4), and the computing unit (4) is designed to carry out the method for determining the state of blockage of the roof channel (1) according to one of Claims 1 to 11.

Images