DE102019118669A1 - Method and system for determining a position signal in an industrial truck - Google Patents

Abstract

The invention relates to a method for determining a position signal (56) in an industrial truck, which reproduces the time profile of a deflection (54) of a mechanical component of the industrial truck, with at least one discontinuous sensor signal by means of at least one sensor (40) by recording discrete measured values of a measured variable is generated, which is converted into a step-shaped position signal (50). The invention also relates to a system (2) for determining a position signal (56) and an industrial truck. The method is further developed in that, by means of a neural network (60), the course of the deflection (54) of the mechanical component approximate constant position signal (56) is calculated.

Description

The invention relates to a method for determining a position signal in an industrial truck, which reproduces the temporal course of a deflection of a mechanical component of the industrial truck, at least one discontinuous sensor signal being generated by means of at least one sensor by recording discrete measured values of a measured variable, which is converted into a stepped position signal becomes. The invention also relates to a system for determining a position signal and an industrial truck including such a system.

Industrial trucks are used to pick up loads in one place, to move them to another place and to unload them there. In many cases, loading and unloading requires high precision and is increasingly being carried out with the aid of machines. Industrial trucks include a large number of movable mechanical components, including the steering including the steering wheel, or electrically or hydraulically driven lifting devices, as well as many fluid cylinders, e.g. hydraulic cylinders, by means of which various components of the industrial truck can be controlled. Examples of this are hydraulic steering cylinders for adjusting the wheels of the industrial truck, with industrial trucks with a lifting mast also often providing high-lift cylinders, side-shift cylinders and tilt cylinders, by means of which the alignment of a load-carrying means can be adjusted.

Such fluid cylinders comprise a piston rod which can be deflected along a cylinder longitudinal direction. In order to enable precise control, it is advantageous if the current state, e.g. the deflection, of the piston rod can be determined as precisely as possible. For example, by precisely determining the lifting height of a load fork when loading and unloading, time can be saved and accidents avoided.

Out DE 10 2016 117 800 A1 a fluid cylinder is known with a piston rod guided in a cylinder jacket, the deflection of which can be determined by means of an optical sensor unit. For this purpose, the piston rod has a large number of markings that are applied to the piston rod at regular intervals. The sensor unit detects the transitions between marked and unmarked sections, so that the deflection of the piston rod can be estimated by counting the transitions taking into account a known distance between the markings.

Since this estimation of the deflection of the piston rod is based on the measurement of discrete measured values of an optical measured variable, i.e. the detection of differences in brightness between the markings and the mark-free sections of the piston rod, the deflection of the piston rod is only at the exact positions of the transitions between light and dark areas exactly determinable. The resulting position signal of the deflection of the piston rod is therefore step-shaped.

In order to estimate the deflection of the piston rod between the transitions as well, it is known to place a moving average value through the stepped position signal.

An example of discrete deflection measurement signals on vehicles is in DE 100 40 372 A1 illustrated contactless, in particular optical, measurement of the steering angle of a vehicle steering system, in which a coding element connected to a steering shaft with a code track is moved past a stationary sensor that detects the code of the code track.

Another exemplary deflection measurement relates to the measurement of the lifting height of a load handling device of an industrial truck by means of a roller body, as shown in FIG DE 197 31 687 A1 is revealed. When the load suspension device moves, the roller body rolls on a moving surface, the roller body interacting with a digital angle sensor designed as an incremental angle sensor.

A lifting height measurement for a lifting device of an industrial truck using a cable is shown in EP 1 203 743 A1 disclosed. The cable pull comprises a cable that can be wound onto a cable pulley. The pulley drives two independent transducers, one of which is designed as an incremental encoder.

Against the background of this prior art, the object of the invention is to provide an improved method, an improved system and an improved industrial truck, by means of which or in which a current state of a mechanical component, in particular its deflection, can be determined as precisely as possible.

This object is achieved by a method for determining a position signal in an industrial truck, which reproduces the time profile of a deflection of a mechanical component of the industrial truck, at least one discontinuous sensor signal being generated by means of at least one sensor by recording discrete measured values of a measured variable, which is converted into a step-shaped position signal is converted, the method being further developed in that, by means of a neural network, the course of the deflection of the mechanical component approximate constant position signal is calculated.

In the context of the invention, a neural network is understood to mean an artificial neural network. The invention is based on the basic idea that by calculating a constant position signal approximating the course of the deflection of the piston rod by means of a neural network, the deflection of the piston rod can be determined better than before and extremely precisely. The constant position signal determined in this way reproduces the deflection of the piston rod much more precisely than is conventionally possible, for example, using a sliding mean value. The neural network allows the deflection to be determined particularly precisely not only at the positions in which the deflection can be precisely calculated by recording and evaluating the discrete measured values, but also in the sections between these positions. This achieves greater precision than with other methods.

In addition, not only the current and past deflections of the mechanical component can be determined from the calculated continuous position signal, but also the future deflection can be estimated.

In embodiments, the mechanical component is a piston rod of a fluid cylinder or a steering wheel, without being limited to these components.

The sensor is preferably aligned directly with the mechanical component, in particular the piston rod. In this way, precise detection of the discrete measured values of the, in particular optical, measured variable is achieved. In the case of optical measurement, a section of the mechanical component, in particular a piston rod, in which a detection area of the at least one sensor is located, is preferably illuminated by means of a light source. In this way, differences in the optical measured variable can be proven particularly well. The optical measured variable is advantageously an intensity.

At least one marking is preferably applied to the mechanical component, the at least one sensor detecting a transition between a marking and an unmarked section of the mechanical component as a sudden change in an optical measured variable and converting it into a jump in the stepped position signal.

The markings allow changes in the optical size, that is to say in particular the intensity or brightness, to be recorded particularly clearly. For this purpose, the markings are in particular darker or lighter than the sections in between. The at least one sensor advantageously recognizes the transition as a change from light to dark and vice versa. A multiplicity of markings are preferably applied to the mechanical component, which markings are arranged at regular intervals from one another. The markings are in particular arranged one behind the other in the longitudinal direction of the mechanical component, or one behind the other for a rotary movement in the circumferential direction.

According to one embodiment, two optical sensors are provided, the distance between which differs from the distance between the markings on the mechanical component. In this way, a direction of movement of the mechanical component can be determined. This makes use of the fact that when the mechanical component moves, the two sensors detect the transitions with a time interval between them. The direction of movement can be determined by determining which sensor detected a transition first. In addition, there is a finer temporal gradation of the combined output signal or a combination of the output signals, which leads to an increased approximation accuracy.

The at least one discontinuous sensor signal is preferably converted into the stepped position signal by counting the transitions detected by the sensor and calculating a stepped approximation of the deflection of the mechanical component using known distances between the transitions.

Directly behind the edges of the step-shaped position signal, which correspond to a transition between a marking and an unmarked section, the step-shaped position signal essentially reproduces the deflection of the mechanical component exactly. The neural network also enables the deflection of the mechanical component to be precisely determined in the area between the edges.

According to one embodiment, a speed signal is calculated from the constant position signal, which reproduces a time profile of a speed of the mechanical component, with an acceleration signal being calculated in particular from the speed signal, which shows a time profile of an acceleration of the mechanical component.

The current state of the mechanical component includes its deflection in addition, its current speed and acceleration. By determining these variables, the mechanical component can be controlled even more precisely. The calculation of the speed signal and / or the acceleration signal is preferably done numerically by determining the change in the position signal over time. Since the deflection of the mechanical component can be determined very precisely by the method according to the invention, the speed and the acceleration of the mechanical component are also calculated very precisely. A more precise calculation of the speed and acceleration of the mechanical component, for example, advantageously improves the functionality of control systems to which these variables are supplied as input variables.

When calculating the constant position signal by means of the neural network, a temporal course of at least one parameter value is preferably taken into account, the temporal course of the parameter value having an influence on the constant position signal, the at least one parameter value in particular being transmitted to an input layer of the neural network.

By taking into account the at least one parameter value, the accuracy of the calculation of the deflection of the piston rod is improved. In particular, the at least one parameter value is transmitted via a CAN bus (Controller Area Network bus). For example, control variables of the industrial truck can be transmitted to the neural network via the CAN bus. In one embodiment, the parameter value is the position of a control stick via which the deflection of the piston rod of the fluid cylinder is controlled. In a further embodiment, the parameter value is a control current to a hydraulic valve. An application example for the latter case is lifting or lowering a load. Further parameter values can be other variables preceding the sequence of action of the measured variable, which announce a change in the measured variable through their own change. It can also be a variable that makes a change in the measured variable probable, for example a stroke which, by approaching a stop, makes the change, e.g. of a speed, probable.

In order to take into account the time profile of the parameter value, delay neurons are used in the neural network in embodiments or parameter values from past time steps are introduced into one or more additional neurons of an input layer of the neural network.

Input and output signals are preferably used to train the neural network. The training output signals specified for training are preferably position signals which reproduce the actual course of the deflection of a mechanical component as precisely as possible. Step-like training input signals are preferably used as input signals, which result when the discrete measured values of the optical measured variable are measured by means of the at least one sensor by converting the predetermined training output signals.

The training output signals are alternatively determined from predetermined step-shaped training input signals, for example by drawing a curve manually through a predetermined step-shaped training input signal, or by forming an average value signal.

To train the neural network, both training signals that result from a constant speed of the mechanical component and training signals that result from a speed of the mechanical component that changes over time are advantageously used. During the training, training signals with acceleration that change over time are also preferably taken into account, such as can occur, for example, when starting and braking the mechanical component.

One advantage of calculating the constant position signal by means of the neural network is that the neural network achieves a much more precise determination of the deflection of the mechanical component in time segments with a constant speed of the mechanical component than is the case with a moving average. This is due to the fact that the sliding mean value from the resulting step-shaped position signal determines an approximation of the deflection of the piston rod, which at least approximately follows the step shape as a wave-shaped signal. In contrast, the neural network, which was trained with a position signal with an, in particular linear, course in this time segment, calculates an essentially linear position signal. As a result, the position signal calculated by the neural network reproduces the actual deflection of the mechanical component much more precisely.

The object is further achieved by a system for determining a position signal, in particular by means of a method according to one of the aforementioned embodiments, the position signal reproducing the time profile of a deflection of a mechanical component of an industrial truck, the system at least comprises an optical sensor which is set up to generate at least one discontinuous sensor signal by detecting discrete measured values of an optical measured variable, the system having an evaluation unit which is set up to convert the discontinuous sensor signal into a stepped position signal, the system being further developed as a result, that the evaluation unit is further set up to calculate a constant position signal approximating the course of the deflection of the mechanical component from the stepped position signal by means of a neural network.

The system for determining a position signal realizes the same features as the method according to the invention and on the basis of this shares the features and advantages as well as properties of the method.

The evaluation unit is in particular a computer.

According to one embodiment, the system comprises an output unit, the evaluation unit being set up to transmit the constant position signal to the output unit, the output unit being set up to display the constant position signal and / or a current deflection of the mechanical component determined from the constant position signal .

The output unit is, for example, a screen or another display device. The current, past and / or future deflection of the mechanical component can advantageously be output to a user of the system by means of the output unit. Alternatively or additionally, the output unit is set up to forward the constant position signal to a data processing unit and / or to process it further itself. For example, it is provided that input values for a control system are determined from the continuous position signal.

Finally, the object is achieved by an industrial truck comprising a system according to one of the aforementioned embodiments and at least one fluid cylinder with a mechanical component, the deflection of which is detected or can be detected by the sensor of the system.

The industrial truck also realizes the same properties, features and advantages as the system according to the invention described above and the method according to the invention described above.

At least one marking is preferably applied to the mechanical component, the at least one sensor being designed to detect a transition between a marking and an unmarked section of the mechanical component as a sudden change in the optical measured variable, the evaluation unit being designed to detect the to convert sudden changes in the optical measured variable into a jump in the stepped position signal.

The evaluation unit is preferably set up to convert the discontinuous sensor signal into the stepped position signal by counting the transitions detected by the at least one sensor and calculating a stepped approximation of the deflection of the mechanical component using known distances between the transitions.

Further features of the invention will become apparent from the description of embodiments according to the invention together with the claims and the accompanying drawings. Embodiments according to the invention can fulfill individual features or a combination of several features.

In the context of the invention, features that are identified with “in particular”, “advantageously” or “preferably” are to be understood as optional features.

• 1 eine schematisch vereinfachte Seitenansicht eines Flurförderzeugs mit mehreren Fluidzylindern und einem System zur Bestimmung eines Positionssignals,
• 2 eine schematisch vereinfachte Darstellung eines Fluidzylinders mit einem Sensor,
• 3a eine schematisch vereinfachte Darstellung eines stufenförmigen Positionssignals sowie eines aus dem stufenförmigen Positionssignal mittels eines gleitenden Mittelwerts berechneten Positionssignals,
• 3b eine schematisch vereinfachte Darstellung eines stufenförmigen Positionssignals sowie eines aus dem stufenförmigen Positionssignal mittels eines neuronalen Netzwerks berechneten stetigen Positionssignals sowie einem Mittelwertsignal,
• 4 eine schematisch vereinfachte Darstellung eines neuronalen Netzwerks zum Berechnen eines stetigen Positionssignals aus einem stufenförmigen Positionssignal.

The invention is described below without restricting the general inventive concept on the basis of exemplary embodiments with reference to the drawings, with express reference being made to the drawings with regard to all inventive details not explained in more detail in the text. Show it:

• 1 a schematically simplified side view of an industrial truck with several fluid cylinders and a system for determining a position signal,
• 2 a schematically simplified representation of a fluid cylinder with a sensor,
• 3a a schematically simplified representation of a stepped position signal as well as a position signal calculated from the stepped position signal by means of a moving average,
• 3b a schematically simplified representation of a stepped position signal and a continuous position signal calculated from the stepped position signal by means of a neural network and a mean value signal,
• 4th a schematically simplified representation of a neural network for calculating a continuous position signal from a stepped position signal.

In the drawings, the same or similar elements and / or parts are provided with the same reference numerals, so that they are not presented again.

1 shows an industrial truck in a schematically simplified side view 10 , which is a high-lift truck in the exemplary embodiment shown. The industrial truck 10 comprises a drive part 12 and a load part 14th . It also includes the industrial truck 10 multiple fluid cylinders 15th , 18th , 24 , 26th , which are, for example, hydraulic cylinders. By means of a steering cylinder 15th let the rear wheels 16 of the truck. For setting the alignment of a lifting mast 22nd is a high-lift cylinder 24 provided, by means of which a fork 28 is adjustable in the vertical direction. There is also a sideshift cylinder 26th for lateral displacement of the load carrier 28 and a tilt cylinder 18th , to the inclination of the lift mast 22nd via a pivot point 20th , intended. These components are examples of deflectable mechanical components of the industrial truck 10 .

It also includes the industrial truck 10 a system 2 for determining a position signal. The system 2 has an evaluation unit 80 , for example a computer, and an output unit 82 , for example a screen. On the output unit 82 For example, a deflection of a piston rod of the fluid cylinder can be achieved 15th , 18th , 24 , 26th and / or display a variable calculated therefrom, such as a lifting height.

In 2 Figure 3 is a schematically simplified view of a hydraulically actuated piston rod 30th of the fluid cylinder 24 shown. The piston rod 30th is in a cylinder jacket 32 guided on which a cylinder head 34 is arranged with a seal. In the cylinder jacket 32 is the piston rod 30th can be moved in the vertical direction, i.e. from top to bottom in the drawing.

On the piston rod 30th are multiple marks 36 upset. The markings 36 reflect incident light less strongly than the other areas of the piston rod 30th . In the embodiment shown, the distance between each two adjacent markings is 36 constant.

The system 2 includes a piston rod sensor 40 pointing to the area of the piston rod 30th is aligned on which the markings 36 are upset. According to one embodiment, an in 2 includes illuminant, not shown, with which the area on the piston rod 30th that the piston rod sensor 40 is directed, is illuminated. The markings 36 are in the longitudinal direction of the piston rod 30th applied so that they are with a deflection of the piston rod 30th one after the other on the piston rod sensor 40 be led past.

The piston rod sensor 40 detects an intensity or brightness of the section of the piston rod 30th that the piston rod sensor 40 is aligned. If the piston rod is deflected 30th the piston rod sensor detects 40 a transition 38 between a mark 36 and an unmarked portion of the piston rod 30th as a jump in intensity or brightness. For the sake of clarity, in 2 just one of the transitions 38 provided with a reference number.

By recording the jumps in intensity at the transitions 38 as discrete measured values, and plotting these measured values against time, a discontinuous sensor signal is generated. From the discontinuous sensor signal, taking into account the known distance between the transitions 38 a stepped position signal 50 generated in the 3a and 3b is shown schematically simplified.

In the 3a and 3b is the deflection 54 the piston rod 30th against the time 72 shown. The stepped position signal 50 is only an approximation of the actual deflection 54 the piston rod 30th because only at the transitions 38 the deflection 54 is exactly determinable. In the areas between the transitions 38 , so in the middle of the markings 36 or on the areas between the marks 36 , the deflection 54 from the measured values of the piston rod sensor 40 do not determine directly. To reduce the deflection in these areas too 54 To calculate, it is known from the prior art to use a moving average 57 from the stepped position signal 50 to calculate as in 3a shown schematically. The resulting position signal 57 however, it is not an exact representation of the actual course of the deflection over time 54 the piston rod 34 , as the moving average 57 at least to some extent the stepped position signal 50 follows, resulting in a wavy profile.

To improve the precision of the calculation of the deflection 54 becomes the position signal from the stepped position signal 50 by means of a neural network 60 calculated. The resulting steady position signal 56 is in 3b shown schematically as a dashed line. The continuous position signal 56 is a much more precise approximation of the actual course of the deflection 54 the piston rod 30th . So are in the 3a and 3b the edges of the stepped position signal 50 arranged at the same time interval from one another. Such a stepped one Position signal 50 results from a smooth movement of the piston rod 30th so that the actual course of the deflection 54 is a linear signal. This is made possible by the steady position signal 56 in 3b much better than the moving average 57 in 3a .

For training the neural network 60 For example, the position signals that indicate the actual deflection are used as training output signals 54 the piston rod 30th reproduce. The resulting stepped position signals, for example, are used as training input signals 50 the neural network 60 fed. In 3b a training output signal is shown schematically in which, due to the uniform speed, it is a mean value curve shown as a dotted line 58 acts.

4th shows a schematically simplified representation of the neural network 60 . The neural network 60 includes multiple nodes 61 , of which only one is provided with a reference number for better clarity. The knots 61 are each an input layer 62 , a hidden layer 64 or an output layer 66 assigned. In the input layer 62 the stepped position signal becomes the neural network 50 fed. In the in 4th The exemplary embodiment shown is also a parameter value 68 supplied, which is, for example, the position of a lever for operating the fluid cylinder 24 or a control current to a hydraulic valve. The hidden layer 64 determined from the values in the input layer 62 the continuous position signal 56 which is in the output layer 66 is issued.

All of the features mentioned, including those that can be taken from the drawings alone and also individual features that are disclosed in combination with other features, are considered to be essential to the invention alone and in combination. Embodiments according to the invention can be fulfilled by individual features or a combination of several features.

List of reference symbols

2

system

10

Industrial truck

12

Drive part

14th

Load part

15th

Fluid cylinder

16

wheel

18th

Fluid cylinder

20th

Pivot point

22nd

Lifting mast

24

Fluid cylinder

26th

Fluid cylinder

28

Load fork

30th

Piston rod

32

Cylinder jacket

34

Cylinder head

36

mark

38

crossing

40

Piston rod sensor

50

stepped position signal

54

Deflection

56

constant position signal

57

moving average

58

Mean value curve

60

neural network

61

node

62

Input layer

64

hidden layer

66

Output layer

68

Parameter value

72

time

80

Evaluation unit

82

Output unit

QUOTES INCLUDED IN THE DESCRIPTION

This list of the documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent literature cited

• DE 102016117800 A1 [0004]
• DE 10040372 A1 [0007]
• DE 19731687 A1 [0008]
• EP 1203743 A1 [0009]

Claims (14)

Method for determining a position signal (56) in an industrial truck, which reproduces the temporal course of a deflection (54) of a mechanical component of the industrial truck, with at least one discontinuous sensor signal being generated by means of at least one sensor (40) by detecting discrete measured values of a measured variable is converted into a stepped position signal (50), characterized in that by means of a neural network (60) a constant position signal (56) approximating the course of the deflection (54) of the mechanical component is calculated from the stepped position signal (50). Procedure according to Claim 1 , characterized in that the mechanical component is a piston rod (30) of a fluid cylinder (15, 18, 24, 26) or a steering wheel. Procedure according to Claim 1 or 2 , characterized in that at least one marking (36) is applied to the mechanical component, the at least one sensor (40) having a transition (38) between a marking (36) and an unmarked section of the mechanical component as a sudden change in an optical The measured variable is detected and converted into a jump in the stepped position signal (50). Procedure according to Claim 3 , characterized in that the at least one discontinuous sensor signal is converted into the stepped position signal (50) by counting the transitions (38) detected by the sensor (40) and by means of known distances between the transitions (38) a stepped approximation of the deflection (54) of the mechanical component is calculated. Method according to one of the Claims 1 to 4th , characterized in that a speed signal is calculated from the steady position signal (56) which reproduces a time profile of a speed of the mechanical component, wherein in particular an acceleration signal is calculated from the speed signal which shows a time profile of an acceleration of the mechanical component. Method according to one of the Claims 1 to 5 , characterized in that, when calculating the constant position signal (56) by means of the neural network (60), a time profile of at least one parameter value (68) is taken into account, the time profile of the parameter value (68) having an influence on the continuous position signal ( 56), in particular the at least one parameter value (68) being transmitted to an input layer (62) of the neural network (60). Procedure according to Claim 6 , characterized in that the parameter value is a position of a control stick via which the deflection of a piston rod of a fluid cylinder is controlled, or a control current to a hydraulic valve. Procedure according to Claim 6 or 7th , characterized in that delay neurons are used in the neural network (60) to take into account the time course of the parameter value or parameter values from past time steps are introduced into one or more additional neurons of an input layer of the neural network (60). System (2) for determining a position signal (56), in particular by means of a method according to one of the Claims 1 to 8th , the position signal (56) reproducing the temporal course of a deflection (54) of a mechanical component of an industrial truck, the system (2) comprising at least one sensor (40) which is set up to generate at least one discontinuous sensor signal by capturing discrete measured values of a measured variable The system (2) has an evaluation unit (80) which is set up to convert the discontinuous sensor signal into a stepped position signal (50), characterized in that the evaluation unit (80) is also set up by means of a neural network (60) to calculate a constant position signal (56) approximating the course of the deflection (54) of the mechanical component from the stepped position signal (50). System (2) according to Claim 9 , characterized in that the system (2) comprises an output unit (82), the evaluation unit (80) being set up to transmit the continuous position signal (56) to the output unit (82), the output unit (82) being set up is to display the continuous position signal (56) and / or a current deflection (54) of the mechanical component determined from the continuous position signal (56). Industrial truck (10), comprising a system according to Claim 9 or 10 and at least one mechanical component whose deflection is or can be detected by the sensor (40) of the system (2). Industrial truck (10) Claim 11 , characterized in that the mechanical component is a piston rod (30) of a fluid cylinder (15, 18, 24, 26) or a steering wheel. Industrial truck (10) Claim 11 or 12 , characterized in that at least one marking (36) is applied to the mechanical component, the at least one sensor (40) being set up to define a transition (38) between a marking (36) and an unmarked section of the mechanical component as to detect sudden change in the optical measured variable, the evaluation unit (80) being set up to convert the sudden change in the optical measured variable into a jump in the stepped position signal (50). Industrial truck (10) Claim 13 , characterized in that the evaluation unit (80) is set up to convert the discontinuous sensor signal into the stepped position signal (50) by counting the transitions (38) detected by the at least one sensor (40) and using known distances between the transitions (38) a step-shaped approximation of the deflection (54) of the mechanical component is calculated.

Images