DEVICE AND METHOD FOR DETECTING LEAKAGE OF A HYDRAULIC CYLINDER

Abstract

The present invention relates to a device for detecting leaks in a hydraulic cylinder, comprising: a first pressure sensor for detecting a pressure value in a first pressure chamber of a hydraulic cylinder, a second pressure sensor for detecting a pressure value in a second pressure chamber of the hydraulic cylinder, an evaluation unit for continuously detecting the pressure values of the first pressure sensor and the second pressure sensor, the evaluation unit being designed to detect a leak, preferably an internal leak, in the hydraulic cylinder that deviates from the norm based on the pressure values recorded by the first pressure sensor and the second pressure sensor.

Claims

1. A device for leak detection in a hydraulic cylinder, the device comprising: a first pressure sensor for acquiring a first pressure value in a first pressure chamber of a hydraulic cylinder; a second pressure sensor for acquiring a second pressure value in a second pressure chamber of the hydraulic cylinder; and an evaluation unit for repeated acquisition of the first and second pressure values of the first pressure sensor and the second pressure sensor, wherein the evaluation unit is configured to identify a leak of the hydraulic cylinder that deviates from a norm based on the first and second pressure values that were acquired from the first pressure sensor and the second pressure sensor.

2. The device according to claim 1, wherein the evaluation unit for evaluating the first and second pressure values that were acquired from the first pressure sensor and the second pressure sensor uses a neural network or is a neural network.

3. The device according to claim 1, wherein the evaluation unit is configured to classify combinations of the first and second pressure values from the first pressure sensor and the second pressure sensor as being within the norm or outside of the norm using machine learning.

4. The device according to claim 1, wherein the evaluation unit is configured to perform evaluation of the first and second pressure values during ongoing operation of the hydraulic cylinder.

5. The device according to claim 1, wherein the evaluation unit is configured to form classification parameters for identifying the leak of the hydraulic cylinder which deviates from the norm using unsupervised machine learning.

6. The device according to claim 5, wherein the evaluation unit is configured to be trained using the unsupervised machine learning using data of the first pressure sensor and the second pressure sensor for a faulty hydraulic cylinder and using data of a non-faulty hydraulic cylinder.

7. The device according to claim 5, wherein the evaluation unit is configured to subject a combination of the first and second pressure values of the first and second pressure sensors to a plausibility check based on a principle of supervised learning after training using the unsupervised machine learning.

8. A method for leak detection in a hydraulic cylinder, the method comprising: repeatedly obtaining first pressure values from a first pressure sensor that measures first pressure in a first chamber of the hydraulic cylinder and second pressure values from a second pressure sensor that measures second pressure in a second chamber of the hydraulic cylinder; and identifying a leak that deviates from a norm based on the first and second pressure values that are obtained.

9. The method according to claim 8, wherein the leak is identified by classification of the first and second pressure values of the first and second chambers measured at a same time or a series of the first and second pressure values of the first and second chambers measured at the same time.

10. The method according to claim 8, wherein machine learning is used for identifying the leak.

11. The method according to claim 10, further comprising: assessing whether there is a deviation from the norm; and using output obtained on account of machine learning as training data for supervised learning to verify whether assumptions used for the supervised learning are correct.

12. The method according to claim 8, wherein the leak is identified during operation of the hydraulic cylinder.

13. The method according to claim 8, wherein identification of the leak is achieved by classifying the first and second pressure values of the first and second chambers measured at a same time or a temporal sequence of the first and second pressure values of the first and second chambers measured at the same time.

14. The method according to claim 8, wherein the leak is identified also using a state of travel of the hydraulic cylinder.

15. The method according to claim 14, wherein the first pressure value of the first pressure sensor, the second pressure value of the second pressure sensor, and the state of travel of the hydraulic cylinder for a common timepoint form a dataset, and identification of the leak is based on the dataset or a temporal sequence of a plurality of the datasets.

16. The device according to claim 1, wherein the evaluation unit is configured to identify the leak as an internal leak of the hydraulic cylinder.

17. The device according to claim 1, wherein the first pressure sensor is configured to continuously measure the first pressures and the second pressure sensor is configured to continuously measure the second pressures.

18. The method according to claim 8, wherein the leak is identified as an internal leak of the hydraulic cylinder.

19. The method according to claim 10, wherein the machine learning identifies the leak using the first and second pressures measured by the first and second pressure sensors for a faulty hydraulic cylinder and using the first and second pressures measured by the first and second pressure sensors for a non-faulty hydraulic cylinder.

20. The method according to claim 8, wherein the first pressures are continuously measured and the second pressures are continuously measured.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Further features, details and advantages of the invention are evident from the following description of the drawings, in which:

[0040] FIG. 1: is a schematic view of the device according to the invention on a hydraulic cylinder, and

[0041] FIG. 2: is a flow diagram for the method for leak detection. FIG. 1 is a schematic view of a differential cylinder which is divided, by a piston, into a first pressure chamber 3 and a second pressure chamber 4. In this case, the piston comprises a seal 6, such that the two pressure chambers are fluidically separated from one another. If the piston is moved, this is achieved by the supply and corresponding discharge of a hydraulic fluid via the two fluid connections 7.

DETAILED DESCRIPTION

[0042] However, if the seal 6 is faulty or already worn beyond the allowable extent, a leak 5 occurs, such that hydraulic fluid flows from a chamber 3 that is under a high pressure to a chamber 4 that is under a lower pressure.

[0043] Since a leak of this kind is not visible from the outside, but leads to increased pumping effort for the hydraulic fluid, and at worst can even lead to damage to the hydraulic cylinder, it is advantageous to promptly identify this leak state.

[0044] For this purpose, one pressure sensor 2, in each case, is provided on each of the two pressure chambers 3, 4, in order to transmit pressure values to an arithmetic unit 1. An evaluation then takes place in said arithmetic unit, which evaluation can identify a leak between the two chambers, beyond the piston, on the basis of the identified pressure values of the two pressure chambers. In this case, the piston seal is typically worn, such that there is a fluid passage between the two pressure chambers which are actually separated by the piston.

[0045] In this case, the pressure states, which clearly identify the leak state of the hydraulic cylinder, are classified in the arithmetic unit by means of machine learning. During the operation of the hydraulic cylinder, the pressure signals from the two chambers are evaluated, and the operating state of the piston seal is deduced.

[0046] In this case, the evaluation is explained on the basis of a flow chart, which is shown in FIG. 2.

[0047] S1 and S2 specify that both a completely intact hydraulic cylinder as a reference, and a faulty hydraulic cylinder comprising a worn piston seal, are required, in order to generate corresponding measuring data of the two hydraulic cylinders in S3. Thus, in S3, the pressure of a differential cylinder is acquired on the rod side and the piston side. In this case, the acquisition can take place over a specified travel path at different compressive loads of the cylinder, and can furthermore also draw on values from the path measurement system of the hydraulic cylinder.

[0048] After the measuring data from the intact hydraulic cylinder have been complied, these form reference measuring data (S4). Analogously thereto, the measuring data of the faulty hydraulic cylinder form measuring data of the faulty cylinder (S5).

[0049] In S6, the measuring data of the intact and of the faulty hydraulic cylinder are combined to form a common dataset, in order to be used as a basis for unsupervised machine learning. In this case, in the present case it is advantageous for the underlying duster method to be a density-based method, which draws on algorithms for data density and distance functions.

[0050] The labeled data (S8), which are used in S9 as training data for supervised machine learning (S10), are then obtained therefrom.

[0051] The neural network, thus trained, forms, on the basis thereof, a model (S11) which can be tested using test data (S12) which are obtained from the labeled data (S8), such that a tested model (S13) is obtained as a result.

[0052] This tested model (S13) is applied by the arithmetic unit 1 such that, in the case of corresponding pressure values of the two pressure sensors, it is possible to reliably deduce a fault in the piston seal. For this purpose, the generated pressure values are simply continuously forwarded to the arithmetic unit 1, which can identify, on the basis of the tested model, the presence of a leak which deviates from the norm and which may be caused by a faulty piston seat.