Patent Application
20230058286
This application claims priority to European Patent Application No. 21192474.1 filed Aug. 20, 2021, the entire contents of which is incorporated herein by reference.
The present disclosure relates generally to ice detection systems and methods, and more specifically to an ice detector system, and method for detecting and distinguishing different icing conditions.
There are many areas where it is useful or important to detect an icing or freezing condition so that action can be taken to avoid or mitigate the effects of such conditions. Icing on aircraft and other vehicles, structures, and systems, can be a hazardous safety concern. For example, governmental agencies, such as the Federal Aviation Administration (“FAA”) in the United States of America and the European Aviation Safety Agency (“EASA”) in Europe, often establish rules and safety requirements that relate to icing on aircraft and aircraft are often fitted with sensors to detect icing and/or de-icing systems. In the aircraft industry, standards exist for aircraft in relation to them being permitted to operate in different icing or freezing conditions. Aircraft may need to be certified to operate in conditions that are classified by the FAA as “Appendix C icing conditions” (see FAA CFR 14 Part 25). More recently, concerns have been raised about different icing conditions e.g. so-called ‘Supercooled Large Droplets’ or SLDs which have, over the last three decades, caused a number of accidents and safety events. Aircraft that are certified to operate in Appendix C conditions may still not be sufficiently robust against SLD conditions etc. The FAA has recently issued heightened rules, at least for certain types of aircraft, pertaining to conditions in which specific water content and altitudes in the atmosphere result in the formation of SLDs. These conditions, which are not covered by Appendix C icing conditions, are referred to as ‘Appendix O’ icing conditions. Only aircraft that are certified for Appendix O conditions may operate in such conditions. For aircraft, there is, therefore, a need to be able to distinguish between dry, or not freezing conditions, Appendix C conditions or Appendix O conditions, since the aircraft may need to remove itself from conditions that may arise but for which the aircraft is not certified. The ability to detect and distinguish between different icing or freezing conditions is also important in other areas e.g. other vehicles, such as trains, automobiles and ships, but also in other fields of industry e.g. wind turbines, machinery etc.
Traditionally, icing condition detection has been performed by optical detection using a reverse polarized laser, magneto-restrictive detection e.g. by monitoring the frequency change on magnetic probes, temperature detection based on water content estimation or by monitoring electric field variation in capacitive elements. Sensing using dual-layer heating and temperature measurements can be used for such detection and the sensor parameters e.g. sampling frequency, temperature, etc. can be controlled to provide faster or more reliable detection based on the confidence levels of the detection as well as, for example, how close the conditions are to the boundary between Appendix C and Appendix O conditions. With such systems and methods, though, the sensor data needs to be analysed on board the vehicle in real time. Conventional ice detection and/or de-icing systems and methods are unable to differentiate the Appendix C icing conditions from the Appendix O icing conditions, and thus aircraft may have insufficient de-icing protocols to handle Appendix O icing conditions, thus resulting in potential safety concerns.
The present disclosure provides a method of determining and differencing icing conditions comprising: creating a statistical model from test data in which set of data from n data inputs are each plotted as a point in n-dimensional space, and wherein the n-dimensional space is divided into a plurality of regions each representative of a different icing condition according to the data in that region; for a set of current data from n data inputs, using the model to classify the icing condition indicated by the current data, by: obtaining current data from n data inputs; providing the current data as input data to the model; determining the region of the model in which the current data set is located; and identifying the icing condition indicated by the current data set according to the determined region.
To reduce the amount of computation required, the statistical model may comprise a hierarchical model comprising two or more levels, and wherein using the model to classify the icing condition comprises: providing the current data as input data to a first level of the model; determining a region of the first level of the model in which the current data is located; identifying a first level of icing condition according to the determined region of the first level; providing the current data to a subsequent level based on the determined region of the first level: determining a region of the subsequent level in which the current data is located; identifying the icing condition for the current data set according to the determined region of the subsequent level.
The first region may be indicative of a dry condition and a second region indicative of a freezing condition. In one example, the first level defines different regions indicative of a dry condition and a freezing condition, and a second level defines different regions indicative of a freezing condition of a first type e.g. Appendix C, and of a freezing condition of a second type, e.g. Appendix O.
Also provided is a system for detecting a classifying icing conditions, comprising: means for storing a statistical model from test data in which set of data from n data inputs are each plotted as a point in n-dimensional space, and wherein the n-dimensional space is divided into a plurality of regions each representative of a different icing condition according to the data in that region; means for obtaining current data inputs and providing them to the statistical model, means for performing an algorithm to determine the region of the model in which the current data set is located; and identify the icing condition indicated by the current data set according to the determined region.
Examples of a system and method according to the disclosure will now be described with reference to the drawings. The description is given by way of example only. The scope of the invention is as defined by the claims.
The detailed description of exemplary embodiments herein makes reference to the accompanying drawings, which show exemplary embodiments by way of illustration. While these exemplary embodiments are described in sufficient detail to enable those skilled in the art to practice the disclosure, it should be understood that other embodiments may be realized and that logical changes and adaptations in design and construction may be made in accordance with this disclosure and the teachings herein without departing from the scope of the invention as defined by the claims. Thus, the detailed description herein is presented for purposes of illustration only and not of limitation.
Disclosed herein, according to various embodiments, are systems and methods for ice detection. More specifically, the present disclosure provides systems and methods for detecting icing conditions and for facilitating distinguishing between different icing conditions. While examples and details herein pertain to implementing the disclosed ice detection apparatus and system in conjunction with an aircraft (e.g., for improving flight safety by detecting and distinguishing between different icing conditions during flight), the ice detection system and method may be implemented for other structures and systems, such as other vehicles, buildings, bridges, wind turbines, power cables, etc.
According to the present disclosure, icing conditions are detected and distinguished by hierarchical classification using statistical models. The first level is a classifier between dry (non-freezing) and freezing conditions. If a freezing condition is identified, a second level is a classifier between different icing conditions e.g. between Appendix C and Appendix O conditions. This is just one example, and the second level could distinguish between other icing conditions. Further levels could conceivably also be added to further classify conditions.
The classifiers can be linear or quadratic, and use parameters that are pre-programmed using test data e.g. data from simulation, wind tunnel tests, flight tests or the like.
The sensor inputs may be compared to known statistics to remove outliers. The initial statistics are taken from sensor specifications and are updated using the non-outlier measurements.
According to the disclosure, with reference to
The algorithm 10 is a multi-class classification algorithm designed using statistical models. Each data set—i.e. set of data from the various sensors and inputs—is plotted as a point in n-dimensional space, where n is the number of data inputs, which includes data, 1, from the sensors plus other data, 2, e.g. from internal measurements e.g. power, data from the flight deck such as air speed, static air temperature, angle of attack, angle of side-slip, etc. All the data points corresponding to a single condition class are considered to be samples from a multi-variate distribution. As seen in
To minimize processing, a first level of classification is performed on the input data 1,2. Based on the pre-programmed parameters, it is determined in which region the condition is in the n-dimensional space. At the first level, in this example, the n-dimensional space is divided into regions where, based on the probability distribution, it is more likely that the condition is dry, and regions where, based on the probability distribution, it is more likely that the conditions are freezing. In the example of
The data inputs 1,2 are compared to known statistics, initially e.g. from sensor specifications, to avoid outliers being used in the classification. The initial statistics can be taken from sensor specifications as shown in
By determining icing conditions using direct sensing of atmospheric conditions and/or sensor inputs e.g. indicating ice accretion of the sensed surface such as an airframe, an accurate detection and discrimination of icing conditions can be achieved in real time. This can result in fewer failure or safety events. The overall effect will also be to decrease costs for all concerned by allowing improved and internationally accepted certification, standards and compliance for a wide range of ice hazards. Ice hazards can be detected quickly and addressed or avoided more quickly thus leading to less damage that needs to be inspected and repaired—e.g. an aircraft is able to quickly detect an unsuitable condition and remove itself from that condition before damage occurs.
By using a statistical model based on test data, it is possible to detect and classify current ice conditions based on sensed data input to the model, thus allowing an accurate detection in real time.
Whilst embodiments have been described above, variations are possible within the scope of the invention as defined by the claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 21192474.1 | Aug 2021 | EP | regional |
