1. The Thinking behind Focused Improvement
The effects of much conventional improvement activity used to be transitory, and there was a strong tendency for improvements not to be capable of being sustained or properly established. In TPM, by contrast, activities are organised to ensure that the benefits of Focused Improvements are upheld, principally through Autonomous Maintenance.
1.2 Allocating Roles in Focused Improvement
Who should be involved with Focused Improvement? Some companies make it the exclusive province of Autonomous Maintenance teams, but this is a big mistake. If there is any activity that should be spearheaded by managers, supervisors and technical staff, it is Focused Improvement. It must also be done in such a way as to pursue maximum economic benefit and contribute to the company’s bottom line. However, managers, supervisors and technical staff each have their own responsibilities, as well as individual strengths and weaknesses, so the responsibility for Focused Improvement should be allocated along the lines indicated in Table 4.1 and Figure 4.3.
Table 4.1 Who Should Tackle What Type of Topic
Topics to be tackled
Department and area managers
Difficult individual topics: right-first-time rate, productivity, startup losses, speed losses
Moderately difficult topics: failures, quality defects, minor stops, changeover losses
Difficult topics requiring specialised knowledge and skills: achieving right-first-time startup, developing new manufacturing methods, prolonging the service life of cutting tools, eliminating startup losses
Special project teams
Difficult, broad-ranging topics encompassing an entire production line or work area: changing process sequences, layouts, or processing methods, or commissioning new equipment
Autonomous Maintenance teams
Easy topics: failures, quality defects, minor stops, changeovers, etc.
1.4 Rolling Out Focused Improvement Step by Step
To implement Focused Improvement efficiently following the kickoff, you will obviously need to select the topics that must be addressed in order to attain the goals, priorities and targets summarised on the activity boards. To eliminate the losses explained in Chapter 2, it is vital that each department, section or subsection decides which machine or line it is going to work on, and then implements Focused Improvements methodically, based on the site’s Focused Improvement Plan. Table 4.3 shows an example of a step-by-step procedure for this. The Focused Improvement Pillar Subcommittee and the TPM Office should carefully monitor and follow up the implementation of the Focused Improvements. Table 4.4 shows an example of a monthly progress chart for doing so, while Table 4.5 shows an example of a horizontal rollout chart. The step-by-step implementation procedure is repeated on a new topic after each horizontal rollout, to reduce the losses even further.
4. The QC Story
The basic procedure for solving any problem is to identify the problem, analyse its causes, put together a plan for solving it, and carry out that plan. The QC story (or Improvement Story) is a simple, practical way of doing this and thereby solving problems efficiently. It consists of the following steps:
Step 1: Choose a topic
Step 2: Assess the current situation and set targets
Step 3: Draw up a schedule of activities
Step 4: Identify the causes of the problem
Step 5: Formulate and carry out a plan for eliminating those causes Step 6: Check the results
Step 7: Ensure that the problem cannot recur (e.g. by standardising) Step 8: Decide on the next topic to tackle, and make a plan for doing so
Although the details of the steps may vary from one company to the next or from one textbook to the next, the basic QC Story remains the same, and should be studied until it is well understood. Table 4.10 outlines each of the steps of the QC Story and indicates which QC tool should be used at each step.
Table 4.10 The QC Story and the QC Tools
5. The 7 QC Tools and the 7 New QC Tools
Quality control needs to be based on facts rather than experience or intuition. The original purpose of quality control was to reduce the number of product defects in mass production, and their chief cause was thought to be variability, so statistical tools were used to address the problem.
Statistical tools used in quality control include control charts, histograms, Pareto diagrams, design of experiments, and sampling inspection. Non-statistical approaches are also used, in the form of cause-and-effect diagrams, QFD (quality function deployment), FMEA (failure mode and effects analysis) and FTA (fault-tree analysis), among others. These methods focus on graphic representation of data likely to prove useful in solving the problem, and their advantage is that they make the problem visible.
In contrast to the traditional 7 QC Tools explained in section 5.2, which are used to handle numerical data, the 7 New QC Tools are used mainly for analysing verbal data. They consist of relations diagrams, tree diagrams, matrix diagrams, process decision programme charts, arrow diagrams, affinity diagrams, and matrix data analysis.
Control makes it possible to achieve a planned value for a given task. The basic approach to control is to repeat the following sequence: Plan → Do → Check (inspect/diagnose) → Act (repair/improve). This is called the ‘PDCA cycle’ or the ‘control cycle’ (see Figure 4.6).
5.2 The 7 QC Tools
(1) Cause-and-effect diagrams
The cause-and-effect diagram (also known as the fishbone diagram or Ishikawa diagram) was developed by the Japanese quality guru Dr. Kaoru Ishikawa to express one of the basic principles of total quality management, that the desired product, target or other outcome is achieved not by directly controlling the results themselves but by controlling the processes that produce those results (since a process is a collection of causes leading to the results).
Note: Cause-and-effect diagrams are sometimes called as ‘4-M Diagram’ or ‘4-M Analysis’ (Man, Machine, Material, Method)
Checksheets are printed lists or diagrams of information required for control purposes. Once dealt with, each item is ticked, making data-collection more efficient and transparent.
Checksheets can be designed to investigate the distribution of process parameters, the occurrence or causes of product defects, or the location of equipment defects.
Although the overall distribution pattern can be deduced from the data points in a frequency distribution table, the distribution can also be represented as a column graph, known as a histogram (see Figure 4.8). Employed to reveal mean values and variation patterns, the histogram plays an important role as a process analysis technique for tasks such as checking for defectives by comparing product quality characteristics against standard values.
1. Frequency Distribution Tables
In a frequency distribution table, quality characteristic data are sorted into a number of regularly-spaced classes, arranged in size order, to find how many values there are in each class. The frequency of a class is the number of values appearing in that class (or the number of times the same value appears, if each data value is taken as a class), and a frequency distribution table is a tabulation of the number of data points in each class.
A good indication of the overall distribution pattern can usually be found by gathering a hundred or more data points, and drawing up a frequency distribution table by distributing them between 10 to 20 classes.
2. Normal Distribution
‘The normal distribution’ is the most common distribution pattern for a variable. It produces a bell-shaped distribution curve that is laterally symmetrical about a central line (see Figure 4.9).
3. Standard Deviation
Standard deviation is used to express the variability of data quantitatively. ‘Variability’ is the degree of dispersion of the data (how spread out they are), and it can be made visible in the form of the distribution curve created by plotting a histogram.
A ‘deviation’ is the gap between a particular data point and the mean of the sample from which the data point was taken. The sample’s ‘variance’, s2, is the sum of the squares of all the deviations, divided by the number of data points. The square root of a sample’s variance is called its ‘standard deviation’, and is represented by the letter ‘s’. It can be calculated using the following formula:
4. Pareto Diagrams
A Pareto diagram is a kind of frequency distribution graph. Imagine a set of data relating to equipment failures, rework, errors, complaints and other losses. The data specifies the financial cost of each loss, how often it occurs, what percentage of the total loss it accounts for, and so on. The losses may also be classified by their causes or by the situations in which they appear. If this data is represented in the form of a bar chart with the values in descending order, it will be obvious at a glance which category has the most failures, defects, etc. Plotting the cumulative total for each category on the resulting histogram will produce a Pareto diagram (see Figure 4.10). A Pareto diagram enables causes to be ranked, and thus prioritised and tackled more effectively.
5. Scatter Diagrams
With a set of data all of the same type, the overall distribution pattern can be found by a method such as plotting the frequency distribution. However, if there are two corresponding sets of data (such as height and weight), and the task is to find the relationship between the two, a scatter diagram, or scattergram, is used (see Figure 4.11). The relationship between two corresponding sets of data such as temperature and yield, or between workpiece dimensions before and after processing, is called a correlation. A correlation can be either positive or negative.
6. Control Charts
A control chart is a line graph on which control limit lines are plotted in order to find out if a process is in a stable condition, or in order to keep it so. Control charts can be used for attributes (discrete data that, by its nature, can only be expressed as positive integers, such as numbers of people and incidences of failure), or for variables (continuous data such as length, weight, time or temperature).
(1) X − R Control Chart
An X − R control chart is an X control chart combined with an R control chart.
An X control chart is mainly used to portray changes in the mean of a distribution, while an R control chart is used to portray changes in the distribution range or within- subgroup variation. An X − R control chart is used when the process characteristic is a variable such as length, weight, strength, purity, time or production volume.
(2) p Control Chart
A p control chart is also known as a ‘fraction defective’ control chart – the ‘fraction
defective’ p being the fraction of defective products within a sample – and is used on its own rather than in combination as in an X − R control chart.
A p control chart is classed as a control chart for attributes, and is used when the sample size n is not fixed (if, for example, 100 steel plates were received in one day, and 8 were defective, and 200 steel plates were received over 5 days, and 14 were defective).
(3) np Control Chart
An np control chart is also known as a ‘number defective’ control chart – the ‘number defective’ np being the number of defective products within a sample. It is classed as a control chart for attributes, and is used when the sample size n is fixed.
(4) Control Limit lines
Any graphical representation without control lines is simply a graph, rather than a control chart.
A control chart has a central line (CL), an upper control limit (UCL) and a lower control limit (LCL), collectively known as control lines (see Figure 4.12).
A 3σ control chart defines a range of 3σ above and below the central line, which represents the mean, and the upper and lower limits of this range become the control limit lines.
(7) Graphs and Charts
Graphs and charts are the basic tools of statistical control. The purpose of using statistical control techniques is to clarify the severity of problems, and identify their causes. Although problems cannot be solved by graphs and charts alone, they can certainly help in tracking down their causes. For this reason, graphs and charts are an essential part of the problem-solving toolkit.
1 Bar Charts
A bar chart is used to compare two or more quantities by representing them as bars with the same width, and lengths proportional to the size of the quantity. Bar charts can be either vertical or horizontal, although vertical bar charts are more commonly used. Bar charts rank with line graphs as the most useful form of graph or chart, and are in fact used even more often (see Figure 4.13).
2 Pie Charts
A pie chart is a graph in which a data set represented by a circle is broken down into categories, each represented by a segment of the circle, or a ‘slice of pie’, the size of each slice corresponding to the proportion of that category in relation to the whole. The principal purpose of a pie chart is to compare the sizes of the slices.
Pie charts are less commonly known as circle graphs. Sometimes, a concentric circle containing different data is drawn inside the first circle, and the resultant graph is known as a ‘doughnut chart’ (see Figure 4.14).
3 Line Graphs
As mentioned before, line graphs show how a quantity changes, especially over time. They are also known as function graphs, broken-line graphs, or time-series charts. In contrast to bar charts, which are characterised by the fact that they compare quantities at a fixed point in time, the chief feature of line graphs is that they provide an intuitive, overall insight into how a phenomenon varies over time. It is therefore important to emphasise the points rather than the lines when drawing them (see Figure 4.15).
4 Compound Bar Charts
Although we have now established that the three basic types of graph are bar charts, pie charts and line graphs, their applications are many and varied. Some of the more commonly used varieties include compound bar charts (also known as stacked bar charts or strip charts; see Figure 4.16), pictographs, area graphs, volume charts and combination graphs.
5 Radar Charts
A radar chart is a graph used to portray the balance between multiple indicators (in other words, the overall skew, or the relationship of the individual values to the mean), and the gaps between the actual and the ideal (see Figure 4.17). The chart has a central point, from which radiate a number of spokes, each corresponding to the axis for a certain indicator, and the value for each indicator is shown by plotting a point on its spoke at an appropriate distance from the centre. Radar charts are useful for representing quality awareness, 5S performance, training results, and other intangible benefits.
A matrix is a table in which the problem phenomena and their potential causes are arranged in opposing rows and columns. A symbol is inserted in a cell to indicate whether the potential cause associated with that cell is relevant to the corresponding problem, and if so, to what extent. This is a very effective way of triggering good ideas for solving problems. Figure 4.18 shows a QA matrix (called an L matrix because of its shape) in which two factors – quality characteristics and processes – are opposed. This kind of matrix is often used in Quality Maintenance. The figure also shows a T matrix contrasting three factors – processes, phenomena and causes.
Chapter 4. Focused Improvement. Part 2