Key Performance Indicators Improve Industrial Performance

(1)

Peer-review under responsibility of Applied Energy Innovation Institute doi: 10.1016/j.egypro.2015.07.474

Energy Procedia 75 ( 2015 ) 1785 – 1790

ScienceDirect

The 7

th

_{International Conference on Applied Energy – ICAE2015}

Key performance indicators improve industrial performance

Carl-Fredrik Lindberg

a,b,

_{*, SieTing Tan}

b,c

_{, JinYue Yan}

b,d

_{, Fredrik Starfelt}

e a_{ABB Corporate Research, Forskargränd 8, 721 78 Västerås, Sweden}

b_{Mälardalen University, Box 883, 721 23 Västerås, Sweden}

c_{Facultu of Chemical Engineering, Universiti Teknologi Malaysia, 81300 Skudai, Johor, Malaysia}

d_{School of Chemical Engineering and Technology, Royal Institute of Technology, Teknikringen 42, SE-100 44 Stockholm, Sweden} e_{MälarEnergi, Västerås, Sweden}

Abstract

Key Performance Indicators (KPIs) are important for monitoring the performance in the industry. They can be used to identify poor performance and the improvement potential. KPIs can be defined for individual equipment, sub-processes, and whole plants. Different types of performances can be measured by KPIs, for example energy, raw-material, control & operation, maintenance, etc.

Benchmarking KPIs with KPIs from similar equipment and plants is one method of identifying poor performing areas and estimating improvement potential. Actions for performance improvements can then be developed, prioritized and implemented based on the KPIs and the benchmarking results. An alternative to benchmarking, which is described in this paper, is to identify the process signals that are strongest correlated with the KPI and then change these process signals in the direction that improves the KPI. This method has been applied to data from a combined heat and power plant and a suggestion are given on how to improve boiler efficiency.

Selection and/or peer-review under responsibility of ICAE

Keywords: KPI; Benchmarking; Efficiency; Energy; Raw-material; Maintenance

1. Introduction

An industry contains numerous types of equipment and processes that are a challenge to control and maintain in order to achieve highest performance and profit for the plant. Key performance indicators (KPIs) are fundamental in measuring the performance and its progress. KPIs can provide information about the performance in different areas such as energy, raw-material, control & operation, maintenance, planning & scheduling, product quality, inventory, safety, etc.

* Corresponding author. Tel.: +46 21 32 32 23. E-mail address: [email protected].

(2)

In the literature different KPI related studies have been conducted, for instance [1] analysed the current state of the art on energy related production performance indicators to derive research gaps and industrial needs for equipment evaluation. Paper [2] identified indicators for reliability of a power plant, mostly focused on generation and transmission networks. Paper [3] established a list of performance indices based on benchmarking analysis of power plants in Australia. Paper[4] focused on economic performance indicators for power plants. Others have focused on technical performance analysis of energy efficiency in power plants in order to provide a guideline for different equipment [5]. Paper [6] identified factors for inefficiency in power plants.

There are several studies that has focused on industrial energy use and energy efficiency in various manufacturing sectors [7-10]. There has been KPI studies that focused on the control and operation performance [11], maintenance efficiency [12], as well as risk and safety [13]. There do also exists a standard, ISO 22400 Key performance indicators for manufacturing operations management [14], which is mainly intended for discrete production and hence only partly useful for continuous production as in process industry and utilities.

Many industries still lack appropriate guidelines on how to measure and improve their performance. Here different types of KPIs are proposed that are able to measure and track different types of performance in the plant. A method that is able to improve the KPIs, i.e. plant performance, is also presented and illustrated in a case study on a combined heat and power plant.

2. Key performance indicators (KPIs)

The reason for low performance is waste in different forms. By identifying the waste and implementing actions that reduces waste improves performance. Waste exists in different forms, for example energy, raw-materials, downtime, operation, maintenance, quality, etc. Below follows examples of KPIs from different areas that can be used to measure performance and identify waste.

The following KPIs are examples, many more are of course possible. This is a list to find inspiration from. The units of the KPIs are only important to keep track of when the KPIs are used for benchmarking in order to use the same units in the comparison. If a KPI only will be used to track trends and find correlated process signals from, then units are less or not important.

Most of the KPIs can be applied to individual equipment, sub-processes, and whole plants.

2.1 Energy KPIs

The energy could be in different forms for example electricity, gas, coal, oil, biomass, steam, etc. The produced output could e.g. be in units of tons/h, m3_{/h, units/h, etc.}

x Energy output / Energy input

x Energy input / Produced output

2.2 Raw-material KPIs

Raw-materials may not only be the main raw-material for a plant, it could also be water, chemicals, etc.

x Raw-material input / Produced output

x Emission / Produced output

(3)

2.3 Operation KPIs

The main Operation KPI is the Overall Equipment Effectiveness (OEE) [15] and its individual parts.

x OEE

x Percentage of scheduled operation time, over a time period, e.g. a day, week, month, etc.

x Percentage of actual uptime of the scheduled time, over a time period.

x Percentage of production rate of max rate for produced product type, over a time period.

x Percentage of full quality products of the production, over a time period.

2.4 Control performance KPIs

Control performance may influence product quality, production speed, equipment wear, etc.

x Number of control loops in manual mode / total number of control loops

x Variance of control error (set-point – measured value)

x Settling time after a set-point change

2.5 Maintenance KPIs

Too little maintenance causes an excessive number of unplanned stops resulting in lost production and emergency maintenance. Too much maintenance causes large maintenance costs and lost production during each planned maintenance.

x Maintenance costs / Produced output over a time period.

x Maintenance time / Produced output over a time period.

x Number of alarms over a time period.

x Same KPIs as for Operation KPIs and some of the Equipment KPIs presented below.

2.6 Planning KPI

Planning and scheduling impacts how well plant capacity is utilized. Since deriving the optimal production plan and comparing it with actual production, is not within the scope of the KPI calculations. Instead a KPI based on adherence to plan is suggested.

x Integrated sum of only positive values of (planned – actual production) over a time period. This KPI is not improved if production catches up later.

2.7 Inventory and buffer utilization KPIs

Large inventories are expensive, too small inventories may cause production disturbances. Buffer tanks should dampen disturbances, if they don’t they are either used in the wrong way or they could be replaced by a pipe.

x Throughput rate / Average Inventory

x Variance of buffer level

x Share of time buffer level is > 95% or <5% over a time period.

2.8 Equipment KPIs

These KPIs can be used to follow the condition of equipment and in some cases also predict when maintenance will be required.

x Different types of efficiencies e.g. heat transfer rate of heat exchangers, pump/fan efficiency, drying efficiency, etc.

x Equipment wear (based on e.g. operating hours, speed, load, startups).

o Number of valve openings for a valve or total valve opening travel distance.

x Vibration amplitude of an equipment.

(4)

3. Methodology

The method presented here is based on identifying process signals or combinations of process signals that are strongly correlated with the KPI of interest. The KPI is then improved by changing the correlated process signals in the direction that improves the KPI.

A simple example illustrates the method. Let a KPI be a flow rate, then we may find that a valve opening is strongly correlated with the flow rate. Hence this KPI can be improved by increasing the valve opening. Sometimes the method requires more work by the user, for example, if the flow rate controls a cooling process and we find that outdoor temperature is strongly correlated with the flow rate. Since we

can’t change the outdoor temperature we have to look for other correlated signals, and maybe also log

more signals in order to find correlated process signals or combinations of process signals that are able to improve the KPI.

The method works as follows:

x Select signals to log. It is better to select too many than too few signals. Select for example all signals related to a sub-process or sub-plant depending on KPI.

x Download historical data from e.g. the last six months with 1 hour sampling rate (1 hour average). The length of the data and sampling time may of course vary depending on type of KPI and plant dynamics.

x Remove signals with zero standard deviation. Remove data in all signals for time periods when the plant is shut down or are working under abnormal conditions.

x Calculate the KPI of interest from historical data.

x Search for signals or signal combinations in historical data that are strongly correlated with the KPI. Note these signals have to be possible to change.

x Change these process signals in the direction that the KPI improves, and verify the improvement. If no improvement then try the signal(s) that are next best correlated.

The signals are collected from historical data, logged at normal operation and/or during experiments that introduces additional excitation. Control loops with constant set-points, where the controlled signal is likely to influence the KPI, may have to be excited in experiments in order to find out if and how the signal impacts the KPI. However, if the controller tuning is poor and there are sufficiently large excitation in the signal, it may be possible to make a conclusion of the signal influence without exciting the controller set-point.

To find the signals or signal combination with strongest correlation to the KPI, all possible signals or signal combinations in the logged data have to be evaluated. This is done by using eq. (1).

(1) Where F(c) in (1) is calculated for each signal combination. The smallest F(c) identifies the signal combination with strongest correlation to the KPI. The process signals y(t) are applied in the function

f(y(t),c), which is given by the user where t is the discrete time with samples from 1 to N. The scaling parameters c in the function f are determined by minimizing the sum of square errors (=F(c)) between the KPI and the function f. In (2) an example is given where f is a linear function of two signals ya and yb. A

linear function is normally sufficient.

(5)

0 50 100 150 200 250 300 350 400 450 500 550 0 1 2 3 4

Sec. air upper front wall [Nm³/s]

0 50 100 150 200 250 300 350 400 450 500 550 420

440 460 480

Temperature main steam [°C]

time [h] 0 50 100 150 200 250 300 350 400 450 500 550 0.305 0.31 0.315 0.32 0.325 0.33 0.335 0.34 0.345 0.35 0.355 time [h]

KPI = Main steam flow [kg/s] / Power in fuel [MW] KPI

Prediction

More than two signals are of course possible to use, but it may generate very many combinations to evaluate. Non-linear functions f of y are also possible.

Note, the strongly correlated signals with the KPI have to be possible to change, i.e. the signals must be related to a control loop or they must be possible to change manually without causing negative impact on production or safety. Also note that the same set-point that improves one KPI may deteriorate another KPI. Some changes of set-points may also not be possible since they violate constraints in the plant.

4. Case study and result

The goal of this case study was to improve boiler efficiency in a combined heat and power plant that utilizes municipal solid waste as the fuel. Here we have defined the boiler efficiency KPI as flow rate of main steam [kg/s] / power in fuel [MW]. In Figure 1 (left), a plot of the KPI together with the best prediction is presented. Note that this boiler efficiency varies from 0.31 to 0.35, a large difference.

Figure 1 Left: KPI based on “Flow rate of main steam / Power in fuel” plotted over time, and the best prediction of the KPI. Right: Signals used in the prediction of the KPI (boiler efficiency).

Applying (2) to our logged data and searching for smallest F(c) gave the following signal combination f(y,c) = c1 *(Sec. air upper front wall) + c2*(Temperature main steam) + c3

where c1, c2 and c3 are identified constants. The signals with the best fit are shown in Figure 1 (right).

Hence a reducing secondary air flow at upper front wall is likely to improve our boiler efficiency KPI. Decreasing main steam temperature is not an option despite it is correlated with our KPI. This result needs however to be verified in further experiments, since the difference between measured and predicted KPI could have been smaller, and even if the difference was smaller, the causality between the KPI and the signals has to be verified.

Conclusions

Different types of KPIs that measures energy-, raw-material-, maintenance-, control-performance, etc. in the industry have been suggested. A method that improves the performance in the industry was also presented. The method is based on identifying the process signals or combinations of process signals that

(6)

have the strongest correlation with the KPI. These process signals are then changed in the direction that they improve the KPI. The process signals are normally changed by changing the set-points in corresponding controllers. The method uses logged data from normal operation to calculate the KPIs. If the excitation in the normal data is low, experiments that increases the excitation have to be done.

The method has been applied to data from a combined heat and power plant and a suggestion has been given on how to improve boiler efficiency by reducing the secondary air flow at upper front wall.

Acknowledgements

The project has been financially supported by the Swedish Governmental Agency for Innovation (VINNOVA).

References

[1] May, G., Taisch, M., Prabhu, V. V., Barletta, I. (2013). Energy Related Key Performance Indicators – State of the Art, Gaps and Industrial Needs. IFIP International Federation for Information Processing 2013. pp. 257–267, 2013.

[2] Čepin M. (2011). Assessment of Power System Reliability: Chapter 13 Reliability and Performance Indicators of Power

Plants. Springer-Verlag: London. DOI: 10.1007/978-0-85729-688-7_13.

[3] Pacific Power Association. (2011). Performance benchmarking for Pacific Power Utilities. Benchmarking report. December 2011. Pacific Power Association. Guan

[4] Curley, G. M. (2006, 0-0 0). Power plant performance indices in new market environment: IEEE standard 762 working group activities and GADS database. Paper presented at the Power Engineering Society General Meeting, 2006. IEEE.

[5] European Commission. (2009). Reference Document on Best Available Techniques for Energy Efficiency. Report.

[6] ABB Energy Efficiency Handbook. (2012). Energy Efficient Design of Auxiliary Systems in Fossil-Fuel Power Plants. ABB PSP Marketing. North America

[7] Tanaka, K.: Assessment of energy efficiency performance measures in industry and their application for policy. Energy Policy 36, 2887–2902 (2008)

[8]Gielen D, Taylor P. Indicators for industrial energy efficiency in India. Energy 2009;34(8):962e9.

[9] Palamutcu S. Electric energy consumption in the cotton textile processing stages. Energy 2010;35(7):2945e52.

[10] Giacone, E., & Mancò, S. (2012). Energy efficiency measurement in industrial processes. Energy, 38(1), 331-345. doi: http://dx.doi.org/10.1016/j.energy.2011.11.054

[11] Samuel, V. B., Agamuthu, P., &Hashim, M. A. (2013). Indicators for assessment of sustainable production: A case study of the petrochemical industry in Malaysia. Ecological Indicators, 24, 392-402.

[12] Kizim, A. V. (2013). Establishing the Maintenance and Repair Body of Knowledge: Comprehensive Approach to Ensuring Equipment Maintenance and Repair Organization Efficiency. Procedia Technology, 9(0), 812-818. doi: http://dx.doi.org/10.1016/j.protcy.2013.12.090

[13] Vestly Bergh, L. I., Hinna, S., Leka, S., & Jain, A. (2014).Developing a performance indicator for psychosocial risk in the oil and gas industry.Safety Science, 62, 98-106.

[14] ISO 22400, Automation systems and integration -- Key performance indicators (KPIs) for manufacturing operations management. https://www.iso.org/obp/ui/#iso:std:iso:22400:-1:ed-1:v1:en

[15] OEE Overall equipment effectiveness, http://en.wikipedia.org/wiki/Overall_equipment_effectiveness

Carl-Fredrik Lindberg

Received his M.Sc. in computer science and eng. (1989) at Luleå University and Ph.D. in automatic control (1997) at Uppsala University, Sweden. From 1997 he has been a member of the research and development staff at ABB Corporate Research and since 2013 adjunct professor in process automation with focus on energy efficiency at Mälardalen University.