Method for establishing the dynamic position of a mechanical translation actuator and related encoder

Field of application

The present invention concerns the field of magnetic positional encoders and, more precisely, a method for establishing the dynamic position of a mechanical actuator for translation purposes, and related encoder. Review of the known art

In the field of positional encoders, subject of the present invention, a rough distinction may be made between two types of operation, as illustrated in Figures 1 and 2. Figure 1 shows an example in which a first type of encoder is used to establish a position. On looking at the figure it will be seen that an actuator 1 comprises a cylinder 2 inside which a piston 3 translates, said piston consisting of a head 4 joined to a shank 5 that projects from the cylinder for a defined maximum length. A permanently magnetized disk 6 is fixed to the head 4 of the piston 3. The free end of he shank 5 is threaded to receive a tool or some other device requiring translation. To establish the position of the head 4 of the piston 3, use is made of an encoder 7 comprising N resistances R of equal value and connected in series where each intermediate point in the series is connected to a switch 8 arranged in series to a current generator (not shown). The switch 8 may be of various kinds, for example one using Reed type relays or on/off Hall-effect sensors.

Operatively speaking, when during translation the magnetic disk passes close to one of the switches 8, the magnetic field temporarily switches it on so that voltage V, measured at the ends of the series, will depend on the position of the disk 6, or rather, on the head 4 of the piston 3. The main drawback to this first type of encoder is its lack of precision in establishing the position (about 3 mm for the Reed-type relays and about 2 mm for the Hall on/off sensors). Further, the minimum measurable increase in movement is low in the on/off form since it depends on the number of relays or sensors, per unit of length, that can be placed there, this in turn depending on the bulk of the device.

Figure 2 represents an example of what happens when a second type of encoder is used to establish a position. In relation to the preceding figure, and numbered as before, it will be seen that there is no magnetic disk at the head of the piston 3, and that the encoder, marked number 8, is made differently and is connected to a processing modulus 9. The figure shows that, outside the actuator 1, there is a mechanical connector 10 one end of which is joined to the encoder 8 and the other end to the shank 5 of the piston 3. The encoder can be made in various ways, for example: a) using a linear potentiometer or an extensometer, the slider being fixed to a mechanical connector 10 so that the voltage measured is proportional to the distance from its origin. This gives excellent precision. b) suitably placing parallel bands consisting of short dark lengths along the direction of translation of an optic sensor fixed to the mechanical connector 10 and, diversely coded, in binary mode, for the various positions that can be reached; c) placing a magnetic band along the direction of translation of a head for reading positional encoding registered on the magnetic band.

One drawback to the second type of encoder is that, to establish the position, additional mechanical connectors are required for translating the actuator, a factor which limits its possibilities of use.

The conclusions drawn for the actuators and related positional encoders are repeated in like manner in devices for establishing the level of liquids in receptacles, the same type of encoder being used for these as well. Summary of the invention Purpose of the present invention is to overcome the drawbacks described above, its subject being an encoder for establishing the position of a permanent magnet fixed to means sliding within a container transparent to the magnetic field, wherein more than two sensors of magnetic field are placed along the path followed by the translating magnet, each one generating a response signal that varies linearly with the movements of approach and withdrawal of the magnet, said encoder comprising:

- means for cyclic acquisition of samples of all said response signals; first means for memorizing maps of calibrated samples of all said response signals at several absolute positions of the magnet; - second means for storing present samples of all said response signals; processing means suitable for minimizing the cumulative distance between said response signals and each of said stored calibration maps;

- data interfacing means controlled by processing means to go out the value of the absolute position stored in the calibration map associated to the minimum distance, as described in claim 1.

Further characteristics of the present invention considered innovative are described in the dependent claims. According to one aspect of the invention, the sensors of magnetic field are Hall-effect sensors (giving a linear response in the length concerned).

According to another aspect of the invention, the first means of storage include a digital memory previously written by an external processor interfaced with the encoder for receiving the corresponding positional values of the magnet, as measured by the gauge. This confirms that the stored maps uni vocal correspond to the encoder hosting them, . the maps depending on the responses given by the Hall sensors actually used,

allowing for a slight variation between one sensor and another, so forming a sort of "signature" for the encoder.

According to another aspect of the invention, the encoder is interfaced with an external calibrating apparatus that measures and conveys the positional values to said processing means, programmed to produce the calibrating maps and enter them in a digital memory in the first means of storage. Having done this the calibrating apparatus is disconnected. According to another aspect of the invention, inside the encoder a calibrating apparatus is activated off line to produce and store calibration maps.

According to another aspect of the invention, the sliding means consist of a piston, and the container of a cylinder from which the piston shank projects to function as the actuator of translation; in this case the magnet is discoid. In fluid dynamic and pneumatic applications, the piston and its shank are preferably hermetically sealed inside the cylinder.

According to other aspects of the invention, the Hall sensors are placed at different points along the path of translation; for example, at the same distance one from another, at varying distances, aligned or unaligned. According to the application, sensors are glued to the external, or internal (where the application does not require a hermetic seal), surfaces of the container, maintaining a suitable distance between magnet and sensor, for example a distance of 5 mm.

The number of sensors per unit of length, compatible with the space needed by the device used, may affect maximum precision when estimating the position of the magnet. Space required for commercial devices is not less than 3 mm but that of the devices most often used is slightly less than 5 mm. A space of 10 mm between Hall sensors may be considered acceptable, a distance irrespective of that between the points along the path of translation chosen for purposes of calibration. A further object of the invention is a method for finding the position of a permanent magnet fixed to sliding means inside a container transparent to the magnetic field, wherein more than two sensors in the magnetic field

are placed along the magnet's path of translation, each one generating a response signal variable linearly with the approach and withdrawal movements of the magnet, said method comprising the following steps: a) cyclic acquisition, out of line, of samples of all response signals obtained at an absolute position of the magnet accurately measured by a gauge, and storage of said samples on a calibration map together with said position; b) repetition of the preceding step at a number of absolute positions, previously fixed in accordance with the desired degree of accuracy, so obtaining an equivalent number of maps; c) cyclic acquisition of present samples of all said response signals and their storage; d) calculation of the cumulative distance between present samples of all said response signals and corresponding samples present in each calibration map ; e) selection of the minimum cumulative distance and index of the associated map where the absolute position of the magnet can be read, as described in an independent claim on method.

The cumulative distance calculated at step (d) is preferably limited to a subset of responses whose distances are the only ones of real significance, typically those of the first four samples in the cycle. According to one aspect of the method used in the invention, steps from c) to e) are cyclically repeated when the magnet is moving, so obtaining in real time a series of positional details In that case, the first thing to do is define the interval of sampling (cycle of steps c), d), e) allowing the desired degree of positional accuracy to be obtained, and then suit the . capacity of calculating the processing means to the number of mathematical operations to be carried out in said interval. The temporal interval of sampling δT and the interval of positional sampling δS are linked to a speed V of magnet translation according to the following formula: δT = δS / V. For example, a speed of 0.1 m/s and δS = lmm

gives δT = 10 m/s. Clearly, the required capacity of calculation will depend on the number of Hall sensors in use.

According to another aspect of the method used, each calibration map must show interpolation between samples of sensor responses in relation to the adjacent positional values of the magnet using an arbitrary number of interpolation points chosen according to the degree of accuracy required. Use my be made of any known method of interpolation; for example: linear, polynomial, spline, etc: The simplest is linear interpolation and is the one to recommend in this case, in view of the linearity of the sensor responses and will preferably be used in the forthcoming description. The interpolated responses from the sensors associated to corresponding positional values of the magnet, also interpolated, produce a greater concentration of interpolation maps that can help to make the position found more accurate, as in these examples: • Example 1 : assuming that calibration has been carried out at a resolution of 1 mm, and that a factor of 10 has been interpolated among the mapped values, an estimate of the magnet's position is obtained at a resolution of 0.1 mm.

• Example 2: assuming that calibration has been carried out at a resolution of 0.1 mm and that a factor of 10 has been interpolated among the mapped values, an estimate of the piston's position is obtained at a resolution of 0.01 mm, and so on.

As interpolation results in larger calibration maps, when it is intended to trace positional information in real time while the magnet is moving, a check should be made to see if it is worthwhile in view of the extra capacity of calculation needed. For example, with V = 0.1 m/s a resolution of 0.1 mm is obtained sampling at δT = 1 ms. In that case, for effective resolution the calibration map must be the interpolated map for δS = 0.1 mm; in other words it must be ten times larger than the map for δS = 1 mm. so as to have a capacity of calculation ten times greater. If using 16 sensors spaced at 10 mm one from another, a processor would be

needed with a capacity of calculation of about 10 Mflop which is less than that of the processors used in present-day personal computers. According to another aspect of the method of the invention, the sampled value of sensor responses stored in the calibration maps becomes altered in accordance with the encoder's working temperature to make up for variations in sensor responses compared with calibrated values. With the rise in working temperature calibrated values are increased of amount contrary to the reduction in sensor responses due to the smaller magnetic field generated by the permanent magnet. According to an optional aspect of the invention, since the absolute position established for the magnet is based on minimization of an objective function (distance), as described in steps c), d), e), counting on the presence of inevitable, if only small, differences between the various responses, it would be useful to enhance these differences (for calculation purposes only) to avoid any uncertainty on the minimum point that might occur by making rough calculations with whole numbers of limited length.

For this purpose the digitalized samples, of each single response given by the sensors, are first multiplied by a fraction of the value acquired, this value being then added to the initial value. These altered values are used both in calibration and in subsequent comparisons with calibrated values. An example explains the result: let us suppose Vl = 3, V2 = 2.8, V3 = 2.5, and V4 = 2.4, the values of four successive samples of the response Fl; now, when Fl is aligned with the map, it appears in the module δl = δ2 = δ3 = δ4 = 0 for all the samples; contrariwise, if Fl is moved from its position on the map, we have module δl - 0.1; δ2 = 0.3; δ3 = 0.1 and σδ = 0.5: If Vi' = 1.1 Vi we have: Vl ' = 3.9; V2' = 3.08; V3' = 2.75 and V4 = 2.64, which gives: δl ' = 0.892; δ2' = 0.33; δ3 = 0.1 1 and σδ = 1.332 > 0.5 q.e.d. Compared with the known art, the advantages of the invention are evident.

- • The encoder needs no mechanical parts to support movement of the actuator which in time is subject to wear and non -alignment.

• The invention offers a far greater degree of accuracy than on/off systems due to the resolution in measuring the position of the piston. Interpolation makes it possible to increase this level of accuracy. For example, if sensors are set apart at 10 mm, this gives a resolution of 0.1 mm, but if set at 3 mm, resolution is 0.01 mm.

• The invention can trace the position of the piston when moving at the usual speeds of actuators, an acoustic and/or visual signal being given when approaching the desired position.

• The invention can be used for fluid-dynamic actuators, whether using oil or compressed air, or a mechanical type.

• The invention can serve as a device for establishing the depth of liquids in tanks where access presents difficulties. Short description of the figures

Further purposes and advantages of the present invention will be made still clearer by the following detailed description of an example of its realization, and by the attached drawings given for explanatory purposes only and in no way limitative, wherein: - figures 1 and 2 show positional encoders made in accordance with the known art;

- figure 3 shows a positional encoder made according to the present invention; figures 4 and 5 illustrate different layouts of sensors used in the encoder in figure 3 along the path of translation followed by a mechanical actuator; figure 6 shows a typical response curve of voltage measured at the output of a Hall-effect sensor; figure 7 shows the response curves of all the sensors used; - figure 8 gives a diagram of a modulus for processing signals from sensors used by the encoder in figure 3;

figure 9 illustrates a calibrating apparatus on the encoder in figure 3;

- figure 10 illustrates the flow diagram for the initial calibration program operated by the processing modulus of the encoder in figure 3, connected to the calibrating apparatus in figure 9; - figure 11 illustrates the map of two memories, MSENS and MCAL respectively, for the output voltages of sensors read by the processing modulus with corresponding positions of the actuator given by the calibrating device;

- figure 12 briefly explains how the maps of memories MSENS and MICAL would be altered by applying the linear interpolation indicated by the formulae provided;

- figure 13 shows the flow diagram of the program for dynamic measurement of the actuator's position executed by the encoder's processing modulus in figure 3; - figures 14, 15 and 16 correspond to a table showing digitalized voltage values of the response curves in figure 7.

A detailed description of some preferred forms of realizing the invention. Illustrated in Figure 3 is an actuator 11 comprising a cylinder 12 made of some non-ferromagnetic material, plastic for example, inside which a piston 13 translates, said piston consisting of a head 14 joined to a shank 15 projecting to some extent from the cylinder 12. A permanently magnetized disk is fixed to the head 14 of the piston 13, for example inside the head 14. The free end of the shank 15 is threaded to receive a tool or some other device requiring translation. In order to establish the position of the head 14 of the piston 13, use is made of an encoder comprising a sensor module 17 equipped with sixteen linear Hall-effect sensors 18 and with a multiplexer 19 connected to a processing modulus 20. The sensors 18 are aligned along a generating line of the cylinder 12 spaced at a constant distance of 10 mm one from another. The figure shows that the 5 ^th sensor S5 is aligned with the magnetic disk 16. This latter can be procured on the market at the required size (for example, one of those marked FLEXO 150-180 in the Magnetic Italiana S.r.l

catalogue). The Hall-effect sensors are linear devices, also available on the market, such as those marked MLX90242 in the Microelectronic Integrated Systems' catalogue.

Figure 6 shows the response (Vout x 100) given by the 14 ^th sensor S 14 (dotted line) measured along 151 mm between the centre of disk 16 and that of the sensor varying by 1 mm at each step. The curve relates to a magnetic north pole's left-hand approach to the marked face of the sensor; were it otherwise the curve would be reversed in relation to the long horizontal section. Perfect alignment is achieved at the central point of the decline by VxIOO = 250 (2.5 V).

Figure 7 shows the response given by all sixteen sensors 18 in which, for the sake of simplicity, the markers of responses in millimetres, measured for each sensor, are not given. Table 1 (figures 14, 15, 16) shows all the digitalized voltages for the above responses. As appears from the figure, as well as from the tabled values, the curves are not perfectly identical, slightly diverging one from another, and this is always true in actual cases, a factor which the invention cleverly exploits to establish the position of the magnet 16 of the piston 13 as will be explained further on. Figure 4 shows another possibility of using the sensors 18 where the distance between sensors is not constant but varies according to a pre- arrangement. In this case too, positional establishment works well, without any alteration, showing indifference to tolerances when positioning the sensors. Figure 5 shows yet another possibility of using the sensors 18 arranged in helical form along the surface of the cylinder 12. Circular symmetry of the actuator 11 makes this possible.

Figure 8 illustrates a diagram of blocks in the . processing modulus 20 completed by traced outlines of the external components 17 and 19. Modulus 20 includes the following components: a microprocessor 21 with its own bus 22 to which all the other components are interconnected: an analog-to-digital converter 23, a memory 2 for the data read by the sensors; a calibrating memory 25; an interface 26 for input/output data; an interface 27 to external devices (alarms, displays, etc;) and, finally, a

DC feeder 28 that supplies all voltages needed by all the blocks shown. Operation of the processing modulus 20 and, more generally, of the encoder, will be explained in subsequent figures of which Figure 9 shows the calibrating apparatus used only in an initial stage (out of line) while Figures 11 to 13 show the firmware controlling microprocessor 21 and the state of memories 24 and 25. Figure 9 shows a calibrating table to which the actuator 11 and a calibrating apparatus 30 are firmly fixed. This latter is diagrammatically represented by a graded bar mounted on the threaded end of shank 15 so as to translate, together with them, through a display showing the position reached by head 14. The figure shows the initial zero position of the shank 15 entirely inside the cylinder 12, the head 14 against the terminal wall and the magnetic disk 16 opposite the first Hall sensor at a point in the median circle between the two bases of opposing polarities. The display 33 is connected to an interface 34 towards the processing modulus 20. On the gauge 30 are two buttons, Pl and P2, both connected to the interface 34. Button Pl serves to confirm the zero position at the microprocessor while button P2 is an optional if calibration is completely manual. This schematization shown in the figure is not limitative, as any known measuring system can be adopted for calibration purposes so long as it is able to distinguish the smallest positional interval at the required degree of accuracy, for example 0.1 mm. Apart from the initial configuration, all the other measurements can be made automatically through the interface 34 using a calibrating apparatus controllable by means of a computer. The length of the gauge can vary as the situation requires; experiments using pistons up to 500 mm long have been made with the encoder.

When in operation and with reference to Figure 10, calibration begins at stage CO with the piston 13 and gauge 30 positioned at zero. The end of stage CO is decided by the operator on pressing button Pl. In Cl an index k is zeroed, said index k allowing for the forward movement of the piston 13 (automatic, or manual with button P2) for previously established lengths according to the resolution required. Index k assumes M= 153

values (from O to M-I). In C2 the positional reading made by gauge 30 is transferred to the processor 21. In C3 an index i is zeroed, its purpose being to direct the individual sensors. Index I assumes N = 16 values (from 0 to N-I). At this point the initial writing address in calibrating memory 25, indicated by MCAL, is established at stage C4. The program enters a cycle of stages C5, C6, C7, C8 and C9 in which a sample VS(i) of the response from each sensor is repeatedly acquired and written, in sequence, in memory 24, indicated by MSENS and, similarly, the position POS (k), measured by gauge 30 for that cycle, and passed by its own interface 34 to the processor's interface 26, is written in memory MCAL. When the present cycle has been completed, the positional index k is increased by one unit to stage ClO, the preceding cycle is repeated, and so on, with repetition of M calibrating cycles, the last of which appears at stage CI l. The next stage, C 12, tests for the presence of interpolation: if this is not required, it terminates at C 13; otherwise interpolation proceeds at C 14.

Figure 11 shows the content of memories MSENS and MCAL on completion of calibration. Each memory is written in sequence starting from the address given by the combination of indices i, k = 0.0. Figure 12 again shows the content of memory MSENS as in the preceding figure, but only to the first two values of the k index to show how linear interpolation actually works. Interpolation is made between samples VS(i,k+l) and VS(i,k) in order to generate F-I new samples in each interval, the interpolation factor being F which, in the example given, is F = IO. The same applies to positional values and therefore to the content of memory MCAL. The calibrating memories with interpolated data are called MSENS-int and MCAL-int and these replace MSENS and MCAL. The interpolated VS samples are expressed as VSI(J ,k) and the interpolated POS positions of the piston are expressed as POSI(J ,k) in which the j index varies between 0 and F-I. This having been established, the interpolated values are determined by the following expressions:

VSl(j,k) = VS(i,k) ₊ t^δ _; (J = O,...,F - I) , r wherein A(i,k) = \VS(i,k + I) - VS (i,k)\

POSI ( j, k) = POS {k) + ^{j ' P} ° - (y = 0,...,F - l) .

Jr

The figure shows images of memories MSENS-int and MCAL-int for interval δ(0.0)

Figure 13 shows the flow diagram of the program that controls the encoder, when the piston 13 moves forward (automatic or manual movement with button P2), to trace their positions as it reaches them. In order to simplify the flow chart, no indication is given of the optional steps that could be taken to signal, acoustically and/or visually, a previously set position at which the piston could stop, but the explanation already given would enable any expert to handle this feature. The cyclic stage of establishing a position begins at step RO with acquisition of N samples of the responses from all sensors, indicated by VSW(i) to distinguish them from the VS(i) samples present in the MSENS calibrating memory (or from the VSI(i) samples in the MSENS-int memory if interpolation has been made). Subsequent steps are ideally divisible in the following two parts:

• Calculation of all objective OBJ functions at the M=I 53 positions of piston 13 previously calibrated (steps Rl to R5) and

• Search for the minimum ofOBJ(k) steps R6 to Rl 1) repeated on establishing each new position (signalled) at step Rl 1. The first part begins at step Rl indicating the field of variability of index k and the k-nth objective OBJ(k) function. R2 gives the field of variability of index i indicating the acquired VSW(i) samples compared with the calibrated VS(i) samples. Index I covers its entire field at increase in index k. In R3 a calibrated VS(i) sample is read by memory MSENS and passed into records under the same name. In R4 the following partial objective function is calculated for the VSW(i) sample: OBJ(i) =

VSW(i) - VS (i) I that corresponds to the "distance" between compared samples: The OBJ(i) function is not limitative and other functions, suitable for the purpose, can be used, for example OBJ(i)={VSW(i) - VS(iJ), or OBJ(I) =(VSW(i)-VS(i)) ² , or others still that are not explicitly linked to the difference between values. In the same way all suitable objective functions can be considered for establishing a "distance" between compared values. At the next step R5, the OBJ(i) value is added to the preceding ones to obtain a cumulative function of

OBJ (k J = 2 ,OBJ\ i\ representing the k-nth position. z=0 From the first part, the second continues on to step R6 wherein an initial value of a MIN variable is set indicating the increasingly lower value of the OBJ(k) functions during minimization. The initial MIN value is the maximum in order to be certain of the result in subsequent comparisons. R7 indicates variability of the k index. In R8 a comparison is made between OBJ(k) and the variable MIN value, this comparison being repeated by increasing k till a value of OBJ(k) < MIN is reached. In that case, in step R9, the MIN variable assumes a value of OB J(k) and, at step RlO, the k value is stored as mink and comparison is resumed until completed. Finally, at step RI l, the POS(mink) element w ill indicate the absolute minimum of objective function OBJ(k) corresponding to the absolute position of the piston 13. This position will appear on the display (not shown) o be transferred to the operator by interface 26, after which, with the piston moving, another cycle will begin to establish the piston's position. Based on the description given for a preferred example of the invention, some changes can clearly be made by an expert in the field without thereby departing from its sphere, as will be seen from the following claims.