The Importance Of Effective Maintenance

Since the maintenance has significant impacts in industry, it has received a deep attention from the expert and practical maintenance. According to study, maintenance costs can represent from 15% to 40% of the costs of goods produced [x2]. Maintenance of process equipment is one of the inescapable tasks associated with the operation of industrial plants and until relatively recently, it was implemented either on a routine basis or after the failure of equipment. Attitudes are changing and now many organisations are adopting methods for identifying incipient faults, so that maintenance can be scheduled before there has been a failure, which would lead to loss of production and spoilage of raw material. Plants and different machine centres are assembled from a wide variety of mechanical and electrical equipment, which due to their very nature are subject to wear, corrosion, erosion and other forms of degradation [1].

It is widely accepted that intelligent, sensor based manufacturing is vital to achieve a high availability level of the sophisticated manufacturing systems in conjunction with high quality levels of manufactured components. Advanced sensor design, coupled with signal processing technologies, permits improved information about the process condition enabling process optimisation and control. It is not adequate to have information relating solely to the tool condition. Additional capabilities such as in process quality control and machine tool diagnostics are a requirement of the future, i.e. a shift from monitoring the tool condition to monitoring the process condition and the resulting part quality. [2] The use of mechanical vibration and acoustic emission signals for fault diagnosis in rotating machinery has grown significantly due to advances in the progress of digital signal processing algorithms and implementation techniques. The conventional diagnosis technology using acoustic and vibration signals already exists in the form of techniques applying the time and frequency domain of signals, and analysing the difference of signals in the spectrum. Unfortunately, in some applications the performance is limited, such as when the signals caused by a damaged element are buried in background noise. [x3]

In recent years, the possibility of obtaining more detailed information regarding the operation of process equipment by applying signal analysis techniques to conventional process measurements has, hitherto, been largely overlooked. This is because the principal purpose of the majority of process measurement systems is to provide a ‘smooth signal’ for process control purposes. Therefore most process measurement systems are designed to produce a steady output signal, which is achieved by restricting the frequency response of the measurement system and thereby suppressing the ‘noise’ component. [1]

Signal transients are generally characterised by a short duration in time as compared to the observation interval. Research has shown that very important measurement information is often associated with the transients [1,2]. For example the vibration signals generated on a gearbox, the transients usually correspond to the local fault of the gear teeth, such as deformation, breakage and fracture. Thus analysing the transients of gearbox vibrations is useful for representing the characteristics of the gearbox health. However, it is not an easy task because the vibration measurements often carry heavy noise in the working environment, which could bury the essential vibration information of gears. Therefore, it is important and necessary to detect the transients with the machine health information in the very low signal-to-noise ratio conditions [3,4].

1.2. Previous Work:

This section contains information about the previous work that has been carried out with regards to condition monitoring and signal processing techniques. It also provides an outline of the different modern signal processing techniques used by researchers to identify faults with various machine components.

1.2.1 Machine tool

A machine management system was developed at Cardiff University as a part of the MIRAM project (Machine Management for Increasing Reliability, Availability and Maintainability) that provides rapid fault diagnosis and allows the implementation of a predictive maintenance strategy [4]. This was used to extract data from the machine components by the sensors attached to it. The data processing methods implemented in the system allowed the characterization of normal and faulty behaviour of the machine component under investigation.

An analysis has demonstrated the application of vibration monitoring using signal processing techniques such as Skewness, Standard deviation, RMS (Root Mean Square) and Kurtosis on the spindle system of machine tools. The results obtained from a signal after implementing kurtosis as a signal processing technique show a relatively high kurtosis value compared to other signals, which indicates a fault, associated with the spindle system. This analysis also describes the variations in the transients obtained from the coolant system of the machine tool. [5]

Water treatment plant

The research work conducted by a PhD student on the WRAP (Water Resources and Action Programme) test rig set-up at Cardiff University focuses on analysing the flow, pressure and pump speed request characteristics with respect to blockages induced on different valves. This analysis also studies the performance of the plant in relation to any leakages occurring in the system.

Another analysis, which studies the performance of the same water filtration plant considers inducing 25 %, 50%, 75 % and 100 % blockages on the valves pre and post the pump. A fault dictionary created using a mathematical model approach was used to differentiate between the pre and post pump blockages and their degree of severities. [6]

Use of modern signal processing techniques

A review of modern signal processing techniques applicable to the analysis of signals whose frequency content is non-stationary has been studied in (R. Burnett, J.F. Watson*, S. Elder) [7]. Techniques like FFT (Fast Fourier Transform), STFT (Short-Time Fourier Transform) and Spectrogram have been discussed and compared using both test and actual data. Results are presented which identify the technique most appropriate for the task of fault detection in machine components under transient conditions [7]. A continuous wavelet transform (CWT) technique used for fault signal diagnosis in an internal combustion engine and its coolant system by using vibration signals has been studied by Wu & Chen (2005) [8]. The advantages of using this technique for the characterization of transitory features of non-stationary signals have also been identified.

Another journal contains a detailed study of the raw data obtained from physical machine parameters. It reviews the implementation of various modern signal processing and analysis techniques in time and frequency domain. Analysing signals in time domain for an initial analysis does not necessarily provide meaningful information due to background noise but transforming the signals into frequency domain provides detailed information. Further analysis shows how the vibration harmonics are linked to the identification of different types of machine faults. The harmonics within the signal categorising a component as faulty can be compared with the amplitude of those corresponding to components in a healthy condition to identify machine’s condition. [9]

Aims of the project:

The aims of the project are enumerated below:

Study the characteristics and relevance of different modern signal processing techniques for the task of early fault detection in industrial machine components.

Apply these modern signal processing techniques using Matlab to an archive of machine tool and water treatment plant signals obtained from the ‘data acquisition system’ (DAS) and from a ‘knowledge transfer program’ (KTP).

Analyse these Matlab processed signals and identify the characteristics of the signal indicating a normal or faulty behaviour of the component under analysis.

Chapter 2: Applications / Systems under study

Wadkin V4-6 FMC:

The Wadkin vertical machining centre was designed as a flexible manufacturing cell to perform operations on small to medium sized components. It comprises a fixed base, a single column, three axes of movements, an automatic tool changer, eight pallet loading stations and pallet transporter. Figure 2 shows a schematic of the machine tool centre.

The Y-axis of the machine is formed by 4 hardened and ground steel slide-ways, which are fixed to the cast iron base of the machine. A table, which slides on low-friction strips can be moved to the left or right to form the X axis, by a motor. The Z-axis of the machine is formed by a cast iron column, which is bolted to the rear and upper surface of the base. This axis is moved and controlled in the same way as the X and Y axes.

The head of the machine comprises the tool spindle, which is driven via a pair of pulleys, a belt and an electric motor. A tacho-generator attached to the motor is used to control the spindle speed. An encoder driven by a belt around the spindle is used to control the spindle position and orientation. Inside the spindle is a draw-bar with a spring (a stack of Belleville washers), which holds a tool in the spindle. A tool can be ejected from the spindle by activating a hydraulic cylinder above the spindle.

This machining centre also consists of a tool transfer system and a 150 tool store is located to the right of the machine cell. The tool comprises 5 disks, attached to a vertical shaft, from which 30 tools can be hung. This shaft can be indexed to present a tool on each of the disk to the tool and horizontally between the spindle and the disks. On this mechanism is a pair of grippers, which holds a tool when being transport between the tool store and the spindle. [11]

Label as on diagram

Component

Description

Reservoir

2 tanks each of 39 litres capacity. Grating prevents ingress of large swarf

Filter

Wire gauze filter attached to pump inlet

Pump / Motor

Graham Precision pump, CPP 76. 3 phase AC motor, 2900 rpm, 1.1kw. Pump single stage peripheral impeller type. Maximum flow rate 11.6 lit/min.

Pressure relief valve

Mechanical operated set at 4.8*10^5 N/m^2

Manual shut-off tap

Manual tap

Flow control valves

Solenoid operated

Reason for analysing coolant pump signals:

A large amount of data obtained from the Wadkin machine tool has already been grouped and processed into various charts and histograms for the ease of the analysis in previous research. The aim was to facilitate the research into machine tool condition monitoring system. Among the 160 recorded failures between a certain period of time, 84% fall within 18 pre-classified machine locations. Among these recorded failures coolant system being one of the 18 locations, ranks no. 3 in the failure frequency by machine location histogram [10]. Therefore applying modern signal processing techniques to these particular signals was considered to be a good approach to identify any faults within the component.

Device(s) for acquiring signals:

The raw data extracted from the coolant system were the pressure and flow signals. These signals were acquired by using flow and pressure transducers connected to the system and a data acquisition system was used to store these signals in a database for further analysis. The Kobold flow meters and switches types VKM have a spring-loaded float, which slides within a cylindrical measuring tube and has an integral orifice. This type of flow meter is a low cost flow meter and switch, which fully compensates for viscosity and to large extent for density even with very low flows.

Wadkin V4/6 Spindle system:

He hardened steel spindle is supported in the head casting by two pairs of pre-loaded angular contact bearings. The spindle is driven by a D.C. electric motor via fixed pulleys and a wide poly-vee belt, and has a speed range of 40-5000 rev / min. Figure 3 shows the schematic diagram of the spindle drive system. The analyses and modelling of this is very similar to the feed drive mechanism, however it does not include any linear movement and is subjected to a distributing torque proportional to cutting force components. [12] The spindle, which is fastened into the head of the machine tool, is shown in figure 9.1. It has two bearings at both the top and bottom of the shaft, together with seals to retain the grease either side of each set of bearings. On the top of the spindle is a pulley, which is driven via poly-vee belt ad a pulley attached to the spindle motor, which is shown in figure 9.2.

To measure the speed of the spindle and its orientation at low speeds, an incremental encoder is positioned towards the side of the spindle. This is driven from a pulley fixed to the spindle, toothed belt and a pulley attached to the encoder as shown in figure 9.3. In figure 9.4 and exploded view of the drawbar, which is fitted into the spindle, is shown. It uses a stack of Billeville washers, which are used to draw a tool into the spindles taper.

Reason for analysing spindle vibration signals:

The bearings in the spindle system can fail fue to overload, misalignment, insufficient lubrication and assembly defects. Condition monitoring of these bearing is therefore very critical to the machining process.

Spindle system’s component specification:

Device(s) for acquiring signals:

The vibrations from main spindle bearings were measured by using two accelerometers mounted on the head of the machine tool as shown in figure 4. Two ENDEVCO model 5216-100 series piezoelectric accelerometers were used, which are designed specifically for vibration measurement in rugged environments of industrial machinery. These type of accelerometers have very wide dynamic range and an extremely low noise floor. [12]

Data Acquisition System (DAS):

A flexible and generically applicable data acquisition system has been developed at Cardiff University for the machine tool condition monitoring. This system can be viewed as a data logger associated to one or more sensors. Its flexibility allows the addition of any data logging software, hardware or sensor(s) to be linked for the analysis of different machine components when required. The DAS is controlled by a relational database management system (DBMS). The DMBS provides a user interface to specify the details of the condition monitoring tests to be performed [13].

2.2 Water treatment plant (WRAP rig):

2.2.1 Background:

The WRAP (Water and Resources Action Program) rig is a test rig set-up at Cardiff University used for research purposes. It is housed in a small container for it to be transported easily to plants or sites for demonstration purposes. The rig is designed to allow a direct performance comparison between two different types of filtration media. This rig was specifically used to demonstrate the performance of Recycled Glass Media (RGM) compared to the conventional filtration media such as sand. In previous research this rig has been used to process wash down water from a dairy production line while its sister rigs were used for different applications such as the removal of bio fouling and other wastes. The figure below shows the WRAP rig along with the individual constituents that make up the rig.

Functionality of the rig:

This explanation of the rig is based on the detailed piping and instrumentation diagram of the rig shown in figure x. The rig consists of four 6-staged impeller centrifugal pumps P1, P2, P3 and P4 as indicated in the diagram with P1 being disconnected as there was no external open reservoir in operation. Pumps P2 and P3 transfer unprocessed liquor to filters F1 and F2 from tank T1. This relationship / process can be altered using valves V8, V9 and V10 for operational flexibility if required.

The two pneumatically actuated diaphragm valves V5A and V5B, which are controlled by the programmable logic controller (PLC) enables the flow to be selected from either the unprocessed effluent stream entering F1 or F2. The PID (Proportional Integral Derivative) loop also maintains the flow rate into each filter at a pre-programmed set point. The flow transmitters FT1 and FT2 were controlled to provide a 0-10V output to the invertors, which controlled pumps P2 and P3. This was achieved via the PID loop algorithm in the PLC.

The analysis of this report only studies a small part of the WRAP rig, which is shown in figure x below. Two Cylon UC32.24 PLC’s networked using a proprietary serial protocol controlled the system. These in turn were networked to a communications module, which contained a GSM modem and features two RS232 ports for connection to external devices. Chemwatch, Cylon’s proprietary SCADA system was used for three original operation of the rig. The system was configured to display two process mimic screens that showed real time values of the various sensor readings. A number of alarm and control screens were also present allowing a privileged operator to alter alarm thresholds, control parameters and manually override system components such as pumps.

WRAP rig system components:

Pump P3:

Data acquisition system (KTP)

Ball valves & Flow transmitters

Order Now