Total cost of ownership: a study of automation systems

It has often been said that we live in an age of solutions. Today, many vendors offer competing platforms for automation which accomplish much the same task, namely the automation of plant floor equipment and machinery. However, it is rare for manufacturing automation system users to properly investigate the total cost of ownership (TCO) for the range of automation platforms on offer.

The true cost of control systems is not just the initial purchase cost, but the sum of costs over the life of the system. This includes the three major phases of the system life cycle:

• Purchase and commissioning
• Operation
• Decommissioning

However, it is the initial purchase cost which is often fixated on, and decisions made only on this up-front price. But the true cost throughout the life of the system can be evaluated and a more relevant value attributed to the different options available. Of these three phases of a control system life, it is by far the operation phase which distinguishes one vendor from the next, in terms of TCO.

This is especially relevant at a time when our country is struggling to compete globally; when costs of labour, taxes and many other direct input costs here are higher than our global competitors. We need significantly more automation in order to reduce costs of production. Those companies which have made a significant investment in automation are more able to compete in the global marketplace than those which have fallen behind, and still rely on significant manual labour.

Replacement complexity, MTBF and MTTR

Over the life of a system, components will need replacement due to fatigue, damage or obsolescence. It is also of value to consider what the anticipated lifetime of the machine will be. This will help to frame what considerations are of interest in the context of the analysed system. So it is important to consider not only the mean time between failures (MTBF) but also the cost and difficulty of replacement for the various components which make up the system, since some components will be simpler to upgrade than others. The reason for the second consideration is that at some point during the life cycle of a machine, a component will need to be replaced with a different model due to obsolescence. The effort required for this substitution will dictate extra measures or considerations which should be made at the design stage.

To illustrate this point, an example of PLC will be used as these devices are ubiquitously used throughout machinery. Since it is possible that the PLC will become superseded over the life of the machine, it would be worth considering the ability to re-use the program code in a new controller.

This ability to re-use the program code in a new-model PLC should be of high importance; the cost to rewrite the machine program will far exceed the cost of the hardware replacement. Look for upward compatibility and a consistent use of the programming software across most models of PLC as well as simple model conversions. Table 1 illustrates some typical MTBF data for various control system components.

Type	MTBF (hours)	MTBF (production years)
Medium PLC CPU	279,102	44.73
Servo driver	100,000	16.03
Micro PLC	105,118	16.85
HMI 7″ model	895,255	143.47
AC motor inverter	82,645	13.24

Table 1: Typical MTBF data for various control system components.

It should be noted that one production year (for Table 1) has been calculated as 24 hours per day, 7 days per week, 52 weeks of the year (100% uptime). Note the actual lifetime of system components will be less than the above table due to power quality and other variables including operating temperature.

Direct costs during operation of the system can be considered as the most visible costs incurred. These are typically:

• Initial purchase cost
• Support or maintenance agreements
• Software licensing
• Training
• Maintenance costs such as replacements
• Electrical efficiency of the system

Note that since all systems will consume electrical power, a relative comparison of electrical efficiency can be more useful than an absolute cost comparison.

While the initial cost of hardware may not be greatly dissimilar between different vendors, there is often a large divergence between solutions on the remaining direct input costs. Programming software (including upgrades) and technical support agreements can add significantly to the direct costs and are highly variable between vendors. Often over a period of five years or less, the cost of service agreements and software licensing can easily exceed the initial investment of the control system. This is especially important to smaller manufacturers, who still need software but have a smaller amount of hardware to amortise the costs over. How readily a vendor’s solution is supported is also worth considering, with obscure brands available at a lower price but which will likely end up costing more over time, due to limited support for the product, and limited support from system integrators and programmers.

Indirect costs tend to be less visible and will require more consideration. Examples of indirect costs can be the following:

• Downtime and loss of production
• Management of the software code and documentation
• Replacement of obsolete components
• Inherent latency in the system

Maintenance costs are slightly more difficult to estimate; however, there is data available to base decisions on. Here the MTBF data, obtained from manufacturers, is commonly available. Interestingly, some people are more concerned about mean time to repair (MTTR) than MTBF. Without doubt MTTR is of interest; however, if it is possible to repair a fault quickly (low MTTR) but fails often (low MTBF) it will be the MTBF which is the dominant consideration. Always take MTBF data as a higher consideration than MTTR.

The warranty period is also of considerable interest at this stage of direct comparison. While failures range from rare to regular, it is important to assess this as it will impact the business greater as a loss of production than will direct hardware replacement cost. Failures outside of warranty have a three-fold cost: hardware cost, installation costs and production loss. It is often accounted for only as a hardware cost, especially when performed under a maintenance budget, but the major impact will be the loss of sales revenue, unutilised production floor labour and waste product.

Of course when a failure occurs, it is important to factor the actual losses a business will incur into the cost of this failure. Production rates ranging from $1K to $50K per hour (based on value of goods sold) are typical, with even higher rates possible when considering automotive or aerospace industries. Often, it is a matter of only one hour of lost production which can produce losses greater than the difference in cost of proven quality control system components and lower cost, low quality components. Even warranty periods for equipment vary greatly, with some suppliers offering twice the warranty period of others; a manufacturer’s confidence in their product is demonstrated by the warranty period offered.

Control system latency

Latency is a measure of the time delay in a system. All digital control devices inherently have latency due to the cyclic evaluation of a program and responses to changes of inputs. The size of the program, the use of tasks and interrupts and the processing power of the CPU all affect the rate at which the system will respond to a change in machine state.

This latency is a source of a hidden cost in manufacturing. The slower a particular control system is to process the user program, the slower it reacts to changes in input states. Although this is often measured in milliseconds, when accumulated over the many cycles per annum, this time can add up to be equivalent to weeks of production time.

There is a large discrepancy between vendors when it comes to cycle times, with some manufacturers achieving cycle rates in the order of 10 times faster than others. The fastest controllers are generally those used in controlling machinery, where quick update rates are paramount. This acts to slow down a production line, and the faster the cycle rate of the machine, the more this latency affects efficiency. However, since this is measured in tens to hundreds of milliseconds per cycle, it must be considered over the period of a production year to evaluate quantifiable differences.

To illustrate this effect, consider a generic machine control algorithm; sequential control by state transition. This sequence consists of five pneumatic cylinders controlled by a PLC, each cylinder having an extended and retracted proximity switch. Each cylinder will be extended, and a proximity sensor is used to trigger a change in PLC state, which commands the next cylinder. Once all cylinders are extended, they begin retracting, one after the other in a similar manner. Each proximity input causes a transition in the state of an output.

The timing of the reaction to the input and change of state of the output is related to the cycle time of the PLC. It can be assumed that the change of state of the input will occur on average halfway through a PLC cycle; hence the time taken to update the output will be 1.5 PLC scans (half a scan to read the input as on and another scan to update the output). If each cylinder extends or retracts in 0.5 s, then we can evaluate the PLC latency and effect over the course of one manufacturing year. The results are shown in Table 2.

	PLC scan time (ms)
	1	2	4	6	8	10
Cycle time (s)	5.015	5.030	5.060	5.090	5.120	5.150
Cycles per annum	6,288,335	6,269,583	6,232,411	6,195,678	6,159,375	6,123,495
Extra cycles per annum	0	18,752	55,924	92,657	128,960	164,840

Table 2: Comparison of number of machine cycles based on a 24 hours/day, 365 days/year production schedule.

As can be seen in the Table 2, the difference between a fast 1 ms PLC scan and a slower 10 ms scan can be almost 165,000 machine cycles per annum. As each machine cycle is 135 ms faster, it will not require additional labour to operate; this is effectively increasing the throughput of the machine.

The above example shows the hidden cost of latency which exists in almost all production lines. The benefits are easy to realise over the course of one year. The cost of a faster CPU will have a very short return on investment (ROI) for fast cycling, transition-dependent machines. PLC manuals should contain processing times for specific instructions, as well as overhead time for peripheral servicing and I/O updates, and be wary of any system which does not make these times clear.

Summary

When comparing industrial automation systems, most consideration tends to be incorrectly given to the initial cost. But it can be readily seen that this initial cost is the least consideration when it comes to analysing the true cost of an automation control system. Ongoing costs, such as product support, software licensing agreements and control system latency represent a much larger proportion of the overall cost than the initial purchase price. Not all PLCs are the same, and a very large discrepancy between cycle times can be found. When comparing data between different vendors, make sure to consider the mean time between failures and give appropriately more importance to this than the mean time to repair. Be wary of ongoing costs and consider their impact to the overall cost of the system.