failure ANALYSIS

A continuing discussion how data is being analyzed for predictive maintenance based on the existing dataset.

reliability

The concept of reliability, when examined through the perspective of machine operation, focuses on the ability of a machine to function properly over a designated period before experiencing any form of functional failure or degradation. This perspective emphasizes the importance of consistent performance and durability of machinery, ensuring that it can perform its required tasks without malfunctioning or losing efficiency over time. 

Reliability, in this context, is quantifiable and can be measured based on the duration a machine operates effectively before encountering issues that impact its functionality. This approach to reliability is critical in various industries where the dependability of machinery and equipment is crucial to operations, safety, and productivity. It involves not only assessing the performance of a machine under normal operating conditions but also considering factors that could potentially lead to its premature failure, thereby enabling preventative measures to be taken to extend the machine's operational lifespan.


failure type

This bar chart categorizes different types of failures, of a machine or system, and shows their frequency counts. The categories of failures are represented in different colors:

The x-axis is labeled "Target," which to indicate the specific area or part of the machine/system that was the subject of the analysis. 

In this chart, the Random Failure type is not considered as label and will not be included in the analysis.

distribution of failure

purpose of the pair plot is to identify potential relationships between various sensor readings and machine failures. The label "1" corresponds to a failure event, and the different plots in the pair plot show how the values of other variables like air temperature, tool wear, process temperature, and rotational speed relate to the occurrence of failures.


Air temperature: There to be a slight positive correlation between air temperature and failures. As air temperature increases, the number of failures also increases slightly.

Tool wear: There is a positive correlation between tool wear and failures. As tool wear increases, the number of failures also increases. This suggests that tool wear could be a useful predictor of failures.

Process temperature: There is a weak positive correlation between process temperature and failures. As process temperature increases, the number of failures also increases slightly.

Rotational speed: There appears to be no clear relationship between rotational speed and failures. The points are scattered evenly across the plot.

Overall, the pair plot suggests that tool wear is the most important factor related to failures, followed by air temperature and process temperature. Rotational speed does not seem to be a significant predictor of failures in this dataset.


distribution of failure per type

These series of bar graphs representing the average values of different features for each failure type within a certain category of machinery or processes, with the exclusion of the "No Failure" category.

Starting from the top left, the first graph depicts the average air temperature (in Kelvin) for each failure type. The second graph at the top right displays the average process temperature (in Kelvin) for each failure type. Both these graphs show a consistent pattern across different failure types, indicating that temperature readings are relatively stable across the failures.

The middle left graph shows the average rotational speed (in rpm) for each failure type. The middle right graph shows the average torque (in Newton-meters) for each failure type. These graphs illustrate some variation across different failure types, suggesting that rotational speed and torque may be more closely related to certain failures.

The bottom left graph shows the average tool wear (in minutes) for each failure type, and this graph has the most variation between the different failure types. This suggests that tool wear be a significant indicator of certain types of failure.

Lastly, the bottom right graph seems to summarize the average rate of failure per failure type, though the units are not specified. This graph provides a direct comparison of how each failure type ranks in terms of its average occurrence or severity.

LIKELIHOOD OF FAILURE DUE TO TOOLWEAR

To create a priority schedule for maintenance or inspection based on the table, we should consider both the likelihood of failure and the tool wear values for each type of failure across the different machine types. The goal would be to prioritize those failures that are most likely to occur and that have the highest tool wear values, as these indicate a greater severity or frequency of the issue.


Here is a suggested priority schedule:


Heat Dissipation Failure for Machine M: This has the highest likelihood of occurrence at 37.04% and significant tool wear values (2 - 213). Given the high probability and potential for a wide range of tool wear, this should be addressed first.

Power Failure for Machine M: Even though the tool wear values are both 0, which might suggest no immediate wear, the high likelihood of 38.27% indicates that this type of failure is common and should be prioritized.

Overstrain Failure for Machine L: It has a high likelihood of 30.04% with substantial tool wear values (177 - 251), indicating a need for attention.

Heat Dissipation Failure for Machine L: The likelihood of 30.45% is high, and the tool wear range (8 - 229) suggests that it could be a recurring issue.

Tool Wear Failure for Machine H: It has a likelihood of 25% with high tool wear values (206 - 246), indicating a frequent need for maintenance or replacement.

Power Failure for Machine H: With a 20.83% likelihood and a range of tool wear from 81 to 208, this indicates a moderately high priority.

Tool Wear Failure for Machine M: It has a slightly lower likelihood of 17.28% but high tool wear values (198 - 253), so it should be monitored closely.

Heat Dissipation Failure for Machine H: It has a substantial likelihood of 33.33% but a lower range of tool wear (20 - 147), so it is a mid-level priority.

Tool Wear Failure for Machine L: With a likelihood of 10.29% and moderate tool wear values (200 - 235), this can be scheduled after the higher priority items.

Random Failures for Machine H: While the likelihood is moderate at 16.67%, the tool wear range is quite wide (2 - 166), suggesting variability that should be investigated.

Overstrain Failure for Machine M: The likelihood is low at 4.94%, but since the tool wear values are somewhat high (187 - 207), it should not be ignored.

Random Failures for Machine L and M: These have lower likelihoods and should be lower priority.

REFINEMENT OF MAINTENANCE WINDOW

The table ABOVE, which lists counts and statistical measures (like mean, standard deviation, and quartiles) of tool wear for different types of failures across machine types H, L, and M, a priority schedule for maintenance can be suggested as follows:

The maintenance schedule should focus first on the most frequent and severe types of failures. It should also consider the machine type's operational criticality and the impact of failure on overall production or safety. Regular preventative maintenance could help mitigate these failures, and the schedule could be adjusted over time as more data on failures is collected.