Are you reaping the benefits from your monitoring system?

A paper published for the EPRI International Maintenance Conference - Chicago August 19-20 2003 by R.G. Herbert & D.J. Wallis, RWE Power International

Abstract

Many power stations do not benefit from their sophisticated monitoring systems due to lack of expertise. Further, systems are generally over sold by their suppliers, using words such "failure predictions" and "estimating times to failure". This is the Holy Grail of monitoring but rarely achievable. However, the monitoring of power stations using the correct level of expertise and equipment can reap significant benefits. Examples of efficient fault detection, limitation of secondary damage, economic use of maintenance effort and allowing plant to run with known fault conditions are real and have been achieved. Conversely, insufficiently trained staff can wrongly diagnose convincing, but benign changes, resulting in a net cost.

Unfortunately, such expertise is rare. It does not come from a short course and many organisations with one or a small number of power stations cannot afford to employ this level of staff for the small number of failures they are likely to experience. This paper shows that, with new technology and communications, such organisations can economically apply the best expertise, gaining the maximum from their monitoring systems and achieving real pay back.

Introduction

Insurance companies often refer to monitoring systems, particular vibration monitoring systems, as wallpaper on the control room wall. They are fitted, look pretty, but are not used. It is a sad reflection on the monitoring industry and those companies that sell the equipment. Equipment manufactures often oversell their products, promising unachievable benefits, and power plant managers, in general, may not have managed their staff and equipment to gain the benefits that such systems can truly offer.

Monitoring equipment manufacturers have been and continue investing large sums of money to improve the level of monitoring, automate the process and downgrade the necessary levels of expertise. In all but a few exceptions there has been little progress over the last 20 years, except for the obvious improvements in both computing power and communications. With the current level of instrumentation installed on large rotating plant in the power industry, there are few techniques available today that were not available 20 years ago.

This paper seeks to show and demonstrate that monitoring can begin to provide some of the benefits expounded by the monitoring equipment manufactures. Using standard techniques primarily in the field of vibration monitoring together with the advances in computer technology, it is now possible to monitor many GW of plant with few staff at one central location. Monitoring systems today are relatively cheap; the manpower to use them is expensive and experienced diagnostic engineers are scarce.

In this field, which has grown rapidly and where the language used has become distorted, it is useful to restate the objective of such monitoring and define some of the common terms. This is covered in the first sections.

Subsequent sections deal with the monitoring process to achieve these goals, such as how and who is using the system, bringing together all levels of staff, from technicians to senior engineers, so that their time is efficiently used. Further sections discuss automated detection, the concept of remote monitoring from a central location and an example is presented to demonstrate these features. A wide range of examples is presented showing the benefits obtained and finally conclusions are draw from our extensive experience of remote monitoring.

The objective of the condition monitoring process

The primary objective of the condition monitoring process is to reduce costs.

In most cases cost savings are achieved by giving plant managers the best possible assessment of their plant condition to estimate the availability, plan outages and minimise consequential damage.

Further indirect secondary savings come from this information, such as a better knowledge of plant behaviour, identification of benign conditions, provision of plant status information, and supporting contractual and warranty claims. All of these issues are becoming increasingly important in the modern power industry.

This objective should not be confused with that of the standard turbovisory system, which protects the plant and may not generate data from which the plant condition can be assessed.

Definitions

In the wide field of condition monitoring, the words detection, diagnosis and prognosis are often used. It is telling that the sector rarely mentions words such as cure, remedy, correction or remedial actions, since it is not until one of these activities has been performed that benefits will be seen. For the purpose of this paper the definition of these words are reiterated below.

Detection The discovery of something hidden or not easily observed

Diagnosis The identification of diseases from their symptoms

Prognosis The forecasting of the course of a disease

Remedial action The action tending or intended to cure a disease

Further, to gain benefits from the "condition monitoring process", and for the purpose of this paper, the process does not just involve the computer sitting in the control room. It must include the management structure and people around it that provide the detection, diagnosis, and prognoses service and the staff that specify the remedial actions.

The monitoring process

Many power plants purchase the latest condition monitoring system but forget that condition monitoring is an ongoing process that requires manpower with the correct skills. Further, these staff must have sufficient credibility such that their judgements are listened to and their suggested remedial actions are acted upon. The cost of the system is relatively low. The on-going manpower to reap the benefits from the process is not.

The condition monitoring process can be considered to consist of three major components:

First line analysis is the routine, often daily, checks. The objective of this work is to ensure that the monitoring equipment is operational and abnormal behaviour is quickly highlighted.

Second line analysis is a regular review that will pick up problems identified from the first line analysis and identify longer-term trends. It will highlight problems before they become critical and provide information for planned maintenance activities.

Third line insurance is the provision of expert advice when abnormalities have been identified. This is non-routine and requires the highest level of expertise.

By recognising that the monitoring process is a multi-skilled, multifaceted activity, and requires a range of qualities and skill levels, the on-going cost can be minimised. The routine checks and reviews can be increasingly automated and allocated to logical, methodical, and lower skilled individuals, while rotor dynamic and plant specialists can provide the diagnostics, prognosis services and suggest the remedial actions.

Automation

To aid the monitoring process and reduce the manpower requirements for the routine fault detection, there is an increasing array of solutions available. For vibration, simple overall level alarms are not sufficient and there is an increasing number of more sophisticated parameters available, some of which are locked into a single or multiple plant parameters. However, if these are not set up correctly, they can lead to either no events or too many events being detected. The former creates a false sense of well being and the latter generates detection overload. Although, these tools are being increasingly supplied with monitoring systems, as they are the simple application of mathematics, the real skill comes with their application.

Setting of these increasingly complicated detection routines uses highly skilled staff and can take considerable time. Further, this is an on-going process often needing to be reset for changes in operating regime and after outages. This can be expensive and thought may be given to using less skilled staff to visually detect pattern changes with simpler alarms.

Remote monitoring

Staff

Fortunately, power plants rarely develop failure conditions and most of the monitoring is routine, falling into the first and second line activities. With the help of some automation this process can be significantly de-skilled and require staff with the ability to reliably, logically and routinely look at, and report on data. This may be suited to plant operators but the responsibility for condition monitoring is rarely placed in their hands. Station engineers are usually given the task but unfortunately engineers have a whole range of immediate problems that need to be solved and take precedence over such mundane activities. Further, due to the infrequent nature of most failures, station based engineers, responsible for a small number of machines, will not develop sufficient specialist skills to reliably diagnose faults. Most of the time the monitoring will become routine and tedious, gradually declining in importance and quality, such that when a rare failure occurs it is missed.

Therefore, driven by the scarcity of staff skills, condition monitoring is gradually becoming a centralised activity. This has been made possible by the introduction of low cost high speed computer communications. Staff with the correct temperament for routine activities undertake the day-to-day role and expensive specialists can quickly learn from faults coming from a fleet of machines. A power station or even group of stations may wish to dedicate and train staff to the necessary level. However, to provide cover for holidays, sickness, succession, etc. there are few organisations that can justify the necessary resources. Although the higher skill levels generally need to be centralised, depending on the available manpower and skill level at the power station, some of the activities can be carried out locally. Thus, with training and support from an Engineering centre, the monitoring activity can be tailored to meet individual power station needs, Figure 1.

Figure 1 Engineering centre to station training

Many of the major plant manufacturers are now operating, or setting up, such centralised systems, which are usually provided to the station as part of a long term service agreement. However, plant operators should assess whose interests are best being served by this service and whether the correct flows of information are being maintained.

Ultimately the success of the remote monitoring process will depend on the ability of staff to provide accurate and timely plant related information. However, it will also depend on the ability of the centre to work closely with the power station, removing the sense of remoteness and generating an atmosphere of confidence and trust.

The monitoring process at a remote location

If routine monitoring becomes a centralised activity remote from the power station site, using a team of analysis staff with a range of expertise, the whole process has to be organised. The plant manager needs to know what information is being extracted from his data and be confident it is performed regularly and thoroughly. The analysis centre needs to schedule the analysis activity to meet the requirements of the monitoring contract and store the information for compiling reports and passing between the team, such that the whole team is aware of current and suspected problems. Information can no longer be held in an individual's head. Table 1 highlights just some of the important features that need to be considered when a centralised condition monitoring process is adopted.

 

Table 1: Essential features needed for the control of a centralised monitoring service

An example of the monitoring process

Good on-site monitoring systems can now automatically communicate to the remote centre facility giving the plant status, Figure 2. Details of each machine's running regime and alarm status are presented for the previous day. Figure 3 takes an example for the unit 3 IP/LP running regime and Figure 4 gives an extract from the alarm log.

Figure 2: Total plant running status

Figure 3: Sample running regime for the Unit 3 IP/LP during the previous day

Figure 4: Sample from the alarm log for the unit 3 IP/LP

The site status highlighted that the Unit 3 IP/LP had triggered alarms, Figure 2, the alarm log records the particular alarms being triggered, Figure 4, and the running status shows these occurred around 26 hours after the machine had reach full operating speed, Figure 3. The particular alarm sequence is given in Table 2.

Table 2: Alarm sequence

It is only at this point that the analyst needs to make contact with the site to investigate the alarms at the time and on the transducer location indicated, Figure 5. Most days no significant alarm is triggered and no action is required, thus reducing the technical support and senior engineer manpower input and consequently the overall cost of monitoring.

Figure 5 Trend displays of the first shaft order vibration data from the transducer in alarm

It was noted yesterday that the clutch bearing vibration exceeded 0.5 in/s rms (12.5 mm/s rms). This is above the limit of 11.8 mm/s rms and into zone D of ISO 10816-2, which is a level normally considered being of sufficient severity to cause damage. A level has been occasionally high on this bearing but yesterday represents a worse case. The high levels are mainly concentrated at this bearing with no significant effect at the adjacent generator or IP/LP bearings. Shaft vibration, although high at about 5 mil pk-pk, did not reflect this high pedestal level.

It is suggested that such behaviour is isolated to the clutch bearing and may be associated with inadequate loading of the support feet under the bearing. A similar problem occurred on a unit at XXX Power Station and was corrected by returning these loadings to their design values.

Further to my e-mail yesterday I attach a display of the latest large change in vibration on the clutch bearing as the load is slightly increased. This is associated with a change in the axial load on the IP/LP thrust bearing, which suggests that the clutch teeth are not sliding correctly and exacerbating the previously discussed support loadings.

It is noted that the station has not had an outage of sufficient length to inspect the pedestal support loads since May and at this time of the year you may not get an opportunity. In view of this we will continue to monitor and inform you of deterioration.

Figure 6 Sample correspondence following detection of an alarm condition

The data, figure 5, confirms that a significant event has occurred and at this stage the senior engineer is requested to investigate. A sample statement is presented, Figure 6, of the initial correspondence, plus additional information following this further analysis. It highlights that an event has been detected and a possible diagnosis has been made. Suggestions of extended monitoring are also proposed to understand the prognosis of the fault. More importantly, it shows that a full investigation has been undertaken, giving both possible causes and indicating further actions, appreciating the commercial running regime of the plant. 
  

Figure 7 Sample monitoring log

The example presented represents the relatively rare occasion when a fault is detected. However, running in parallel with this high level process is the mundane, but immensely important, monitoring scheduling process. This is used to record all aspects of the monitoring process raised in Table 1, programming the analyst to monitor the correct plant at the correct times and report their findings. In most cases these reports will record a benign null state of the machine and documentate both system and instrumentation defects, always ensuring the monitoring system is fully functional. Figure 7 shows an extract from a typical report, which highlights alarms being triggered and issues that should be kept under review, in this case the clutch.

Typical examples where benefits have been accrued

Using the above approach, an impressive array of defects has been detected, diagnosed and ameliorated, Table 3. This has been achieved in some cases by simple remedial action, but on other occasions machines have been allowed to safely continue operation with a known fault condition.