MAINTENANCE

Ronan ALLIBERT

Iain King

25/10/99

Every machine is concerned with "failure". A failure is a non-satisfying condition for a given equipment. There is partial or complete failures. Consequences of complete failure (functional failure) are :

• Consequences about security, with the possible loss of the equipment and its operators;
• Operational consequences, indirect economical loss for the non-production, and other costs than the one to repair;
• Non-operational consequences, direct economical loss, due to the repair cost itself;
• Consequences about hidden failures (of an hidden function), that exposes the machine to a risk of multiple failures.

Security and economical consequences like those need to be avoided, or at least minimised.

Maintenance is an ensemble of operations realised on a system to keep it or bring it back to its required function. Such a way to:

• Ensure the client a high availability of a production unit; ensure a high security level and equipment inherent reliability;
• Establish security and availability on equipment when a fault occur;
• Follow the life cycle of each part of the production system, and ensure its appropriate renewing;
• Get necessary information to upgrade design when components of an equipment are few reliable;
• Minimise the faults and the cost that comes with; realise all those aims at the best total cost.

There is three main ways to look at maintenance on an industrial system: (maintenance policies)

Unscheduled maintenance activities which are performed to satisfy any routine requests for action are classified as demand or corrective maintenance. This maintenance on-demand category includes emergency request for work, daily or weekly maintenance requests, and any other unscheduled repair demands that are placed on the maintenance department.

Preventive maintenance is a policy of periodically scheduling and performing work activities on facility assets and equipment items to minimise the chance of failures and downtime. The primary objective is to reduce the costs associated with the owning and operating facility and fleet assets while prolonging asset life cycles and equipment productivity. The periodically scheduling is based on the equipment MTBF/MTTF.

MTBF: mean time between failures

MTTF: mean time to Failure

This maintenance is based on condition, an on-time checking and analysis is performed on equipment to detect any appearing degradation and then it enables to make the most effective decision. This a really trustful technique but it is for sure the most expensive one, so it is used on big systems.

In those cases, information about interventions done, and its frequency should be recorded in a data base, and at least in a CAMM -Computer Assisted Managing Maintenance-.

Another way of looking at the those different policies is given by the next chart (fig. 1). Policies are classified differently. Other policies are mentioned, but an express definition is given.

Fig. 1 Maintenance Policies

Techniques

Working out those policies is quite different from one to the other. Corrective maintenance doesn’t need any specify structure to be managed, but preventive maintenance needs at least a data base with all the MTBF of the equipment included in the maintenance program.

In the case of predictive maintenance, other stages are necessary. We need first to analyse, very specifically every parts of the system, to optimise the required number of captor, and their places. We need to complete this stage really specific technical and theoretic knowledge. It’s most of the time done in research labs specialised in a peculiar domain. Then all the captors are fixed and adapted to the spares to monitor, connected to a computer designed for the need. The computer system will be the interface between the machine and the man (maintenance operator) and it will sometimes need to make decision on its own (stop the monitored system), when human interventions are to slow to ensure security.

Nowadays about fifty percent of systems uses vibration analysis. (mechanical, acoustic vibrations…). Others devices are used, like opto-electronic systems, IR camcorders, thermograph, …

Implementation

A periodical maintenance program is required for every item which its loss of function would have a consequence about security. If preventive maintenance doesn’t ensure the required level of security, the equipment must be re-designed.

A periodical maintenance program is required for every item which its loss of function isn’t detected easily by the machine operator, and then unreported to the maintenance services.

In all the others cases consequences of failure are economical, preventive maintenance should then be justified economically.

Every consequences of failures (even economical ones) are due to design characteristics, and by then could only be reduced by design changes:

• Security consequences could be reduced in most of the case by redundancy;
• Hidden functions generally could be detected by instrumentation (test) or other design characteristics;
• Implementation and economical efficiency of preventive maintenance depends on easiness of equipment inspection;
• Cost of corrective maintenance depends on failures modes and inherent reliability of the concerned system.

Periodic maintenance can ensure to obtain the inherent level of reliability of an equipment, but no maintenance policy can ensure a greater reliability than the inherent reliability of an equipment due to its design.

Choosing a maintenance policy (maintenance program process) depends on its applicability and its efficiency, it should satisfy both criteria. Applicability depends on article design, efficiency, on consequences that the program should prevent.

Economical environment

Maintenance has now become essential to the design and exploitation of industrial system, looking at improving a good level of security, reliability and looking at earning money. For example the system down on one of the production line in Peugeot cost them about £100000 a day… But a maintenance that doesn’t match with the system for sure is a way to waste money, and can bring critical situation, dangerous for the humans and the system itself.

That is why different maintenance policies are used in different industrial system. Not so long ago, curative and preventive maintenance where the only policies. But the first one of the two was just applied on systems physically risk-less when fault occurred, and the second one would bring exploitation over-cost.

In response to those problems, conditional or predictive maintenance appeared to monitor systems, in a way to anticipate faults and act consequently (immediate stop, or stop planned for the coming days).

Still predictive maintenance is very expensive, (mathematical model, captors, computer, network,…), it is only used in big systems. But we can see that new cheaper technologies monitoring appears like new mathematical models of spares, so it brings out new horizons for smaller production units.

Technological environment

Others technologies

There is no other technology facing the maintenance in its domain. Even if equipment quality increases, there is always degradations, so maintenance is still useful.

Technical evolutions

Maintenance evolution is to be able not to stop a machine during its production cycle even if non-critical failures occur, keeping still a high level of quality, until equipment maintenance or replacement can be done, in a way to increase productivity gain and security.

Other evolutions are needed for captors, to get easier installations, and more accurate results. Same needs are for interpreting those captors, to increase monitor’s reactivity and synthesis possibilities, with a better relationship with data bases.

Those evolutions are necessary to give smaller firms the possibility to use predictive maintenance which is still an expensive policy, very specific to machine types, and quite long to install (1 to 4 years).

There is a lot of research programs in different domains, like mathematics (models), physics (vibration analysis –e.g. Lab in IUT de Roanne !-, thermal emission… to indicate failure), computer science (for monitoring system to be able to make decision when failures are probable, but not certain; more accurate results are needed, not a binary system–operational spare part 1 or not-, but gradual results –e.g. 67% of the optimal estate of the spare part 1-)…