Monetizing Data in Maintenance: Data-driven Spare Parts Management – Part 2
Today, digitization, Industry 4.0 and Maintenance 4.0 bring about vast volumes of data. Technologies such as IoT or IIoT allow large sets of devices to connect to data networks and send complex data continuously.
Organizations today maintain huge amounts of data, structured or unstructured. However, from research of renowned organizations like Gartner, we know that industrial firms today are not able to use 70—90 percent of data that are collected and stored. This paradox is described in the paper, and various generic models of big data monetization are proposed. Some of these models are shown in examples from spare parts management.
Spare parts inventory can lock in significant amounts of working capital. This article summarizes recommendations for effective spare parts inventory management and spare parts optimization using various sets of data and statistical analytical methods.
The management of spare parts and other materials needed for realization of maintenance processes is one of the key functions in physical asset management. Especially in power generation, oil and gas and heavy chemical industries, spare parts inventories can easily add up to tens of thousands of various items, at a value of hundreds of millions of euros.
It is obvious that efficient spare parts inventory management can have significant impact on the financial performance of the company. Better spare parts management can lead to improvement of financial performance of the company.
In previous research we discussed several recommendations for spare parts inventory management. Using these recommendations, companies can achieve better financial performance in different parts of the spare parts lifecycle. In some of these recommended practices, various data can be employed and analysed – especially in areas like portfolio segmentation, criticality assessment, forecasting, improving spare parts naming and identification, or cleaning and rectifying master data.
Eight Rules of Good Spare Parts Management
In our previous research, we refined the following eight rules – best practices – for good spare parts management:
- Focus on preventative maintenance – for preventative maintenance no inventories of spare parts need to be held.
- Solve problems in spare parts processes.
- Segment your spare parts portfolio.
- Analyse spare part’s criticality.
- Use suitable forecasting methods and verify their accuracy and reliability.
- Use special methods for intermittent demand items.
- Consider the whole lifecycle of your assets while making decisions related to spare parts.
- Implement a good information system for spare parts management so all above stated rules are supported and/or automated.
In this issue of Maintworld we will describe in more detail the importance of analysing spare parts critically and the need to use suitable forecasting methods and verify their accuracy and reliability.
Analyse Spare Part’s Criticality
In large organizations operating large production systems, the size of spares portfolio amounts to tens or hundreds of thousands of items. It is therefore essential to be able to distinguish the important ones from the others. Criticality of spare parts is after all the ultimate measure of spare parts’ importance.
The level of a spare part’s criticality is inevitably related to the criticality of the production equipment it is used for (so having an RCM analysis done will certainly help in assessing the criticality of spares). However, we need to keep in mind that criticality of spare parts is not equal to criticality of the device the spares are used for.
When analysing criticality, we need to collect and look at various areas of data linked with the item: cost of inventory holding, failure probability, impacts of spare part unavailability, lead-time and other parameters – as shown in Fig. 4. Based on the level of item’s criticality, appropriate service level targets should be set.
In practice, costs of inventory holding, and costs of spare parts unavailability should be carefully balanced. Costs can be compared using the following equation:
Where Cinv are costs of inventory holding per one year and Cun are costs of unavailability of spare part in case of need calculated per one day. LT is lead time calculated in days and f is frequency or probability of failure (need for spare part) as occurrences per year.
If all data for the equation is available, criticality can be calculated directly – and easily. But in practice typically some the variables in the equation are not known at all or are uncertain, blurred and inaccurate. This is where advanced analysis of available data comes to question. From our experience, in industrial organizations a number of interesting sets of data can be utilized to evaluate (or support evaluation) of spare parts’ importance (criticalness or criticality).
In the following diagram (Figure 6), a 2-level evaluation of criticality (or identification of critical items – materials or spare parts) is described. This 2-level approach allows for the “clever”, efficient process of evaluation of large numbers of items. Using various data sources like spare parts master data, history of spare parts transactions, RCM data, data from previous assessments of critical items, bills of materials etc., a preliminary separation of clearly non-critical items vs. suspicious (potentially critical) items can be done by means of data analysis without human interaction. After this preliminary evaluation, we can spot the relatively small group of potentially critical items and focus further evaluation on them. In this way the preliminary evaluation can save a lot of work and time otherwise required from maintenance technicians to assess each item individually.
In the second step, potentially critical items are scrutinized thoroughly to find out their level (score) of criticalness. This can be done either in a quantitative way (if required data is available) or qualitative way (data must be collected by means of questionnaires filled-in by maintenance technicians or engineers).
If data for quantitative calculation is not available, we need to rely on information from maintenance engineers or technicians. To objectivize their subjective view on spare parts (maintenance engineers are often strongly biased towards keeping excessive inventory, “just to be safe”), we have proposed a structured questionnaire.
Questions about spare parts realisability, probability of failure, and impact of unavailability, lead time, etc. should be designed to fit specific conditions of the organization (industry, technology/production equipment used etc.). The equation (1) should be taken into logarithm, so we can change variables for indices which can be added and subtracted instead of multiplicated. Answers should be given weights, and the questionnaire should be balanced so that we can have sums for each area (index) as shown in the following equation: IP-Iun-ILT-If=0 (2)
This allows for summing weights for answers in each area into a single index. If the left side of the equation is greater than 0, we should keep at least one item of spare part in stock. If it is less than 0 we should not – this spare part is not critical.
Weights should be tested on selected parts with known (or agreed) criticality. A pilot mix of spare parts should include some parts which are critical for sure, some which are not, and some which are in between.
It is essential to include maintenance engineers in both selecting questions for the questionnaire and in selecting weights for answers. This helps to create a better understanding of the questions and the whole purpose of the criticality assessment. A maintenance engineer should be able to fill the questionnaire in an average time of 2-10 minutes, so that the criticality assessment will not consume much of working time.
Although assessment can be done on paper or in Excel, today it makes more sense to use available services like Google Forms or SurveyMonkey or others, that can be used to collect needed data and minimize the work with collecting, processing and analysing the data.
Spare Parts Management Starts with Good Forecasting
The next step in the specification of optimum spare parts inventory management regime is the prediction of future demand (consumption) for the items in stock. The forecast is always based on transactional data from information systems – history of spare parts consumptions, which must be representative (meaning sufficiently long). In the case of spare parts, we usually work with a history of three to ten years (depending on industry). Three years of recorded history seems to be the minimum for intermittent items. A general rule here applies: the longer the history, the better and more reliable the forecast.
When analysing historical consumption, we need to carefully distinguish between material consumed for planned maintenance (planned shutdowns, turnarounds, preventive maintenance) and spare parts issued for unplanned (corrective) maintenance – repairs. In forecasting, we must adjust the history for planned maintenance.
In the forecasting process, items should be treated individually, according to the character of their consumption. Items with common demand patterns (high runners – fast moving items like fasteners, etc.) can be forecast using a number of standard statistical methods normally used in inventory management (moving average, exponential smoothing, Holt’s exponential smoothing, trends, seasonal indexes, Winter’s method, etc.). Items with intermittent demand require a special suitable method to be applied. The use of standard methods of prediction and inventory management in case of intermittent items results often in a substantial overestimate of future consumption and therefore excessive inventory level.
It is known—and obvious—that the people who spend the longest time around the equipment are the machines’ operators. With that in mind, it makes sense to train some of them to track equipment performance and behavior, instead of just reporting when the asset stops working altogether.
Is this lift safe to use until its next periodic inspection? Lift experts and maintenance personnel are expected to answer this question clearly and unambiguously – potentially challenging in practice, as safety functions in modern lifts are software-monitored and digitally controlled. The solution is based on effective and continuous verification of the firmware’s product safety and its safety in use.