• Aucun résultat trouvé

Cătălin Iancu, PhD

University “Constantin Brâncuşi” Târgu-Jiu, Romania ciancu@utgjiu.ro

Abstract - In this paperwork is presented an experimental method for determination of vibration state of solid, three-dimensional structures, applied for mechanical press PMCR 63.

First is presented the experimental equipment needed, then the real experiment and, in the end the interpretation of results. This kind of experiments are made in order to seek eventually resonance phenomenon and the results will be used as a comparison criterium for validation of analitic model built for further FEA analysis.

I. INTRODUCTION [1]

In the large category of conventional machines for plastic deformation, the mechanical presses have the largest use.

Designing such equipment, with high productivity, and using automation, leads to a large utilization of presses. Notice that the mechanical presses are widely used among this type of equipment.

For the mentioned reasons have been developed modern analysis methods, finite element method being hardly used lately. So far the structural calculus using FEA method was very little or even not used for structures like mechanical presses, being eventually applied on static ground, existing the possibility to perform dynamic analysis.

In order to observ possible danger working regime and to identify possible excitement frequency nearby natural frequencies of press, and also to determine the bed deformation under loads developed on normal working conditions, have been conducted experiments on real structures, because deformation, as characteristic frequencies, will be used as a comparison criterium for validation of analitic model built for further FEA dynamic analysis.

II. REQUIRED EXPERIMENTAL EQUIPMENT [3]

- Piezoelectric transducer for acceleration Bruel & Kjær, type 4391;

- Load amplifier Bruel & Kjær, type 2635;

- Resistance transducerfor linear stroke, type LK 15.

- Notebook equipped with analogue-numerical interface Keithley, type DAS 1602;

-Software for data acquisition and processing, wrote under TestPoint.

In fig.1 is a picture of the needed experimental equipment, presented before.

Fig.1. Experimental equipment A. Accelerometer

The accelerometer is an electromechanical transducer, which gives an electrical signal according with the acceleration which is submitted to.

One of the most used accelerometer is the piezoelectric type, which in principle consist of two piezoelectric discs, with a seismic mass on top, pre-compressed trough a lamellas system. The whole system is mounted inside a metallic case, which serves for the take up of mechanical motion.

When the accelerometer is subject to a mechanical vibration, the seismic mass will induce on piezoelectric discs a variable force, according to mass acceleration.

Based on piezoelectric effect, discs will load with an electrical charge according to the force applied on them, so according to mass acceleration.

The system mass-elastic lamella is practically an elastic mechanical system with one degree of freedom, characterized by a special resonance frequency. This is a particular characteristic of each accelerometer.

T. Sobh et al. (eds.), Innovative Algorithms and Techniques in Automation, Industrial Electronics and Telecommunications, 85–88.

© 2007 Springer.

85

Fig.2. Diagram of a piezoelectric accelerometer For frequencies lower then resonance frequency, the acceleration of the seismic mass will be approximately the same with the acceleration of the base of transducer, and so the electric charge output of transducer will be according to the acceleration, which the transducer is submitted to.

Accelerometer type 4391, made by Bruel & Kjær, used on experimentation, has the main characteristics:

- current sensitivity: 1pC/ms-2±2%;

- frequency domain: 0,1Hz…12 kHz;

- resonance frequency: 40 kHz;

- mass: 16 g.

B. Load a m p l i f i e r

The load amplifier is introduced in the circuit for two reasons:

1°.-amplify the electric signal quite small, given by the piezoelectric transducer (which can be acceleration or force transducer);

2°.-transform the high impedance of the piezoelectric transducer to much lower impedance, in order for coupling to the other measurement and processing devices.

Load amplifier made by Bruel & Kjær, type 2635, used on experimentation, has the main characteristics:

-current sensitivity, step selection from 0.01 mV/pC to 10 mV/pC;

- frequency domain: 0,1Hz…200 kHz;

C. Resistance transducerfor linear stroke.

This is a potentiometric transducer of a strong design, in order to resist heavy working conditions.

Resistance transducerfor linear stroke type LK- 15, made by Novotechnik, used on experimentation, has the main characteristics:

- resistance: 5 KΩ;

- maximum supply voltage: 42 V;

- active stroke: 150 mm.

- maximum speed: 10 m/s;

- maximum acceleration: 200 m/s2; - resistance insulation (at 500 Vcc): ≥10MΩ.

D.Numerical acquisition interface

This interface makes the analogue-numerical conversion of the input electric signals, and transfers these signals to the computer in order to achieve numeric processing.

In principal an acquisition interface is characterized by:

number of input channels, global frequency sampling, and resolution.

Numerical acquisition interface type DAS 1602, made by Keithley used on experimentation has the main characteristics:

- number of input channels: 8 differential / 16 Single End;

- global frequency sampling: 100 kHz;

- resolution: 12 bit.

E. Computer notebook 586DX type 3005

F. Softwarefor data acquisition and processingTestPoint–

Keithley

TestPoint is a programming environment of high performance, object oriented, integrating data acquisition control, numeric analysis, matrix calculus, signal processing and graphic representation, in an easy way to handle [5].

The limitations that may occur using Test Point programming environment are subsequent to performances of computer used.

For graphic representation is used a specific graphics, with two cursors, being able to represent up to 6 curves in Cartesian coordinates, associated to one of 4 ordinates. The graphic representation has two slider-objects for selecting the upper and lower limits of representation domain and other two slider-objects for horizontal displacement of two cursors, data associated to curves in the cursors position being listed in a corresponding numbers of displays, having the same color as the curves.

Fig.3. Example of TestPoint-Keithley software display IANCU

86

III. EXPERIMENTS ON NORMAL WORKING CONDITIONS

Using the experimental equipment described previously, have been conducted experiments for determination of vibration state of a mechanical press type PMCR-63, mechanical press with open bed, with nominal force 63 tf, in order to observ possible danger working regime and to identify possible excitement frequency nearby natural frequencies of press [1][2].

Also has been determined the bed deformation under loads developed on normal working conditions. Deformation, as characteristic frequencies, will be used as a comparison criterium for validation of analitic model built by FEA analysis.

Experiments have been conducted on a mechanical press type PMCR-63, made in 1999 la S.C. MIRFO S.A. Tg-Jiu, Romania, and working on normal condition at “AUTO SPARE PARTS Inc.”, Topoloveni, jud.Argeş, Romania.

Fig. 4. Aspect of conducting experiments (Disposition place of measurement points)

In fig. 4 is presented a picture taken while measurements had been conducted. By points 1, 2, 3, and 4 are represented the disposition place for the acceleration transducers.

For all the measurements bed deformation has been determined in single place, in the “C” zone.

On picture are represented:

1 - Measurement point nr.1- (disposition place);

2 - Measurement point nr.2- (disposition place);

3 - Measurement point nr.3- (disposition place);

4 - Measurement point nr.4- (disposition place);

It have been determined the following characteristics:

-vibration acceleration on horizontal-transversal direction;

-vibration acceleration on horizontal-longitudinal direction;

-vibration acceleration on vertical direction;

-bed deformation in the “C” zone.

Measurements have been conducted while press executes a combined operation of sheet metal stamping with punching on an auto item of steel sheet of 3-mm thickness.

IV. PROCESSING DATA AND NUMERIC RESULTS

In fig.5 are represented the characteristics of vibration accelerations and bed deformation, determined in measurement point nr. 1. The observation made hereby are equally the same for the other measurement points (2-4).

The order of curves is:

-Green curve, over the whole display- bed deformation;

-Blue curve, higher- vibration acceleration on horizontal-longitudinal direction;

-Red curve, middle- vibration acceleration on vertical direction;

-Black curve, lower- vibration acceleration on horizontal-transversal direction;

Fig.5. Characteristics determined in measurement point nr. 1

MECHANICAL PRESS PMCR 63 87

It can be observed that the bed deformation, measured in the maximum opening zone (green curve), has a maximum while effective cutting of sheet, a slight inertia on tempering, due to friction between the dies, and then go back to initial state, whole phenomenon during less then 0.5 sec.

For the real working conditions (sheet metal stamping, of 3-mm thickness), maximum deformation is about 0,225 3-mm.

The blue curve represents the vibration acceleration on horizontal-longitudinal direction. It can be observed two small vibratory zones in the extremities of measurement interval (t=0, 5 sec.), due to coupling-decoupling of press clutch.

Also it can be observed two vibratory zones of high amplitude, which boundary the jump of green curve (bed deformation), and represents the beginning of cutting process and respective the ending and dies spalling.

The same shape is remarked for vibration acceleration on vertical direction (red curve) and for vibration acceleration on horizontal-transversal direction (black curve), the correspondent amplitudes being slightly different, that proves the good rigidity of press.

In order to determine the spectral composition of functioning generated vibrations [2], it was done the Fourier transform of recorded accelerations.

In fig.6 is represented the modulus of Fourier transform of vibration acceleration on vertical direction, registered in measurement point nr.1.

Fig.6 Modulus of Fourier transform of vibration acceleration on vertical direction

Fig.7 Modulus of Fourier transform of vibration acceleration on horizontal-longitudinal direction

In fig.7, shown above, is represented the modulus of Fourier transform of vibration acceleration on horizontal-longitudinal direction, registered in measurement point nr.1.

V. CONCLUSIONS

By examining whole characteristics [4] [6], presented in fig. 6 and 7 (and also for the other measurement points) it can be observed that the frequency spectrum has a relative uniform distribution in frequency range 1÷550Hz, so in normal working conditions there aren’t any resonance phenomenon.

Also it can be determined the natural frequencies, which will be used as a comparison criterium for validation of analitic model built for further FEA analysis.

REFERENCES

[1]. Iancu C., “Contributions to dimensional optimization of mechanical presses in dynamic regime”, Ph.D. Thesis, University of Pitesti, Romania, 2001.

[2]. Manea, I., “Introduction to modal analysis by practice and theory”, Ed. Scorilo, Craiova, Romania, 2000.

[3]. User’s guide, DAS 1600/1400 Data Acquisition, 1997.

[4]. Sound & vibration catalogue, Bruel & Kjaer, 1997.

[5]. ”TestPoint” Techniques & Reference, Capital Equipment Corporation, Massachusetts, USA, 1996.

[6]. Randal, R.B., “Frequency Analysis”, Bruel & Kjaer, 1987.

IANCU 88

Quality of Uni- and Multicast Services in a Middleware. LabMap Study Case

Cecil Bruce-Boye University of Applied Science Lübeck

3 Stephensonstrasse Lübeck, 23562 Germany

Dmitry A. Kazakov cbb software GmbH

1 Charlottenstrasse Lübeck, 23560 Germany Abstract- The quality of service (QoS) is essential for a

distributed data acquisition and control system. QoS depends on numerous factors, which are difficult to predict in advance. In this paper we present an empirical comparison of uni- and multicast based implementations of distributed middleware services. The LabMap® middleware offers both uni- and multicast data distribution services. Unicast is based on TCP, multicast on the PGM streams. We measured the performance of both transport layers on the typical 1-n case where multicast deployment would be possible. Our study shows that PGM is a useful complement to TCP transport, though its use should be carefully planned in advance.

I. INTRODUCTION

A typical distributed data acquisition and control system is a loosely coupled network of nodes capable to publish and subscribe system variables. For the applications running on the nodes the network is abstracted away through the middleware.

The middleware provides: naming and identity services; data distribution and transport services; information services such as browsing and time services. Quality of these services (QoS) plays a decisive role for applications to enjoy advantages of the middleware.

Within the middleware the network is abstracted as a device with some network protocol supporting the services. Thus the concrete transport protocol is abstracted as well, to allow reuse of the middleware core implementation.

In the recent past there was little choice for a middleware working over the Ethernet. It was either TCP sockets for unicast or else UDP datagrams for broadcast connections. The choice was difficult, in particular, because of unreliability and lack of traffic control of UDP. Modern multicast technology presents an answer to UDP problems. It provides efficient filtering to protect an outside network from potentially massive traffic between distinct nodes. Traffic separation is achieved physically by switches. The corresponding network protocols are available for network traffic management control from the application side. Is the multicast technology mature to meet the requirements typical for automation and control application area of middleware technology?

We carried out an extensive empirical study of QoS based on uni- and multicast transport layers on the example of the

LabMap® middleware1 [1]. One of the goals of the study was to justify empirical results of LabMap® deployment for hardware-in-the-loop scenario. [2]

II. RELATED WORKS

To the present time multicast in middleware, if implemented, was exclusively on the broadcast basis. For the CORBA (Common Object Request Brocker) [3] there exist proposals for deployment of multicast [4], but no known implementations of. For unicast services QoS measures can be found in [16].

OPC (an initiative for open data connectivity) [5] also does not provide multicast layers.

The iBus middleware [6] provides multicast services, however its transport layer is not natively multicasting and any figures about performance are unknown to us.

Seppo Sierla conducted a QoS study of NDDS (Network Data Delivery Service) implementation of RTPS (Real-time Publish-Subscribe) middleware interface specification for unicast services [7].

Spread is a message distribution toolkit which supports multicast messages [8]. It is not a middleware, but it can serve as a reliable multicast transport layer for a middleware.

Performance data on messages services are available for spread [9].

III. MIDDLEWARE ARCHITECTURE OVERVIEW

A. Networking

The middleware abstracts connections between nodes and represents them to the applications as publisher / subscriber relations. However, the efficiency of this abstraction highly depends on the nature of the underlying connections. In general to consider are:

Peer-to-peer connections, like TCP/IP sockets. The advantage of a peer-to-peer connection is that it allows a straightforward packet filtering based on MAC (Media Access Control) addresses. An error correction mechanism is usually easy to implement, for example, by resending, because both sides are aware of each other´s state. The disadvantage of peer-to-peer connections is that in the case

1 LabMap® is applied for testbed automation by automotive vendors like AVL, Daimler-Chrysler AG, Opel, VW, MAN, Bosch AG.

T. Sobh et al. (eds.), Innovative Algorithms and Techniques in Automation, Industrial Electronics and Telecommunications, 89–94.

© 2007 Springer.

89

of 1-n and n-1 connections the overall overhead is a multiplicative of n.

Multicast connections, like UDP datagram sent at the broadcast address. The advantage of multicasting lies in fixed overhead for 1-n connections. However using UDP might expose difficult QoS problems and middleware management problems. This problem, which is not specific to middleware, was recognized the by IETF (Internet Engineering Task Force) and a series of standards was designed to provide more efficient transport protocols suitable for 1-n connections. In particular IGMP (Internet Group Management Protocol) [10] and PGM [11]

(Pragmatic General Multicast) are of especial interest for middleware.

B. Remote services

The policy of QoS is the requirements imposed by a subscriber on the published data. It is the expectations of the subscriber on QoS. The following policies are important:

On demand - the subscriber explicitly requires data from the publisher. This type of policy is essential for implementation of events, commands, client-server queries, higher order services such as browsing.

Periodic – the subscriber receives a data flow from the publisher. The subscriber specifies the data period. This type of policy is used for physical state data known to be defined at each moment of time. Usually the subscriber asks the publisher for the “native” data period because it is the physical limit and more frequent polling makes no sense.

This policy is widely used, but exposed to various problems.

The period should be twice as long as the “native” period, otherwise the subscriber will experience “oversampled”

data. This sufficiently limits the system performance. When timestamps are supported, the subscriber can filter out repetitive data, however this would mean an additional burden for it.

Periodic on change – the subscriber receives data only upon state change. Usually the state change is value or timestamp change. This type of policy does not suffer the problems typical for the periodic policy. Yet it sometimes makes application design more difficult, because the subscriber should synchronize on the data, which depending on the data sources might stay unchanged for a long period of time. In such circumstances, for the subscriber it might become difficult to detect data losses. Another problem is that the system load is less predictable under such event-controlled scenario. For mission critical applications that could be unacceptable due to possible time constraints violation.

However, in some cases time constraints can still be satisfied due to physical / logical constraints a priory known for the system.

C. Local services

QoS is also influenced by the middleware notification services available to the subscribers. The services can be synchronous or asynchronous to the subscriber’s execution threads. For a service to be synchronous implies that the data

transfer may occur only on demand and also the subscriber is blocked until I/O completion. Such architecture is obviously flawed, so in all known implementations the synchronous services are only interfaces to the underlying middleware services, which themselves are natively asynchronous. The notification services can be:

Callbacks – this notification service is usually performed from a separate execution thread, which requires interlocking and data exchange with the notified thread.

Further callback can be blocking or non-blocking. A blocking callback prevents data loss. That is - if a next notification needs to happen during callback processing it is postponed until callback completion. Blocking callbacks is a great danger for the whole system because they may violate time constraints, strain system resources and deadlock. Non-blocking callback may suffer data losses.

Synchronization objects – this notification service is based on a waitable resource.

There are many types of synchronization objects:

Event is the most simple synchronization object. An event gets signaled upon notification. The subscriber can wait for the event. This solution may also suffer data losses and data corruptions.

Semaphore is a synchronization object that can be used to protect shared resources, such as data. The most used variant of semaphore is mutex. Mutex represents a very low-level mechanism exposed to various problems, from deadlock to priority inversion. Usually mutex and event are used as building blocks for higher-level synchronization objects, which are more reliable and safer to use. This notification service is based on a waitable resource.

Queue is a more elaborated synchronization object. As with the callbacks the queue may be blocking or not. Thus it is again a trade-off between time constraints and data consistency. Queue represents a 1-1 synchronization object.

Blackboard is a 1-n synchronization object. The notifications get published on the blackboard and interested threads may inspect the blackboard for the notification and enter waiting for a new notification.

Protected object is a higher-level primitive, which can be specialized into each of mentioned above objects. Protected objects are language supported and can be used only in interfaces written in higher-level languages providing concurrency primitives. Protected objects are known to be very efficient in terms of context switches [12]. The disadvantage is that they require language support, which a middleware interfacing to lower-level languages like C++

would lack.

IV. MULTICAST TECHNOLOGY OVERVIEW

Two issues are essential for middleware to take an advantage

Two issues are essential for middleware to take an advantage

Outline

Documents relatifs