Most senior executives recognize the economic potential of adopting Maintenance 4.0 practices.
The easy part is developing a high-level strategy and business case for bringing the Internet of Things, Machine Learning and Automation into the maintenance and reliability arena. At the same time, planners tend to overlook the plant floor reality: Existing systems and processes cannot be replaced in the short term.
Computerized Maintenance Management Systems (CMMS) contain detailed operational information including asset performance data and work order management.
Given the disruptive nature of Industry 4.0 (the fourth Industrial Revolution), does a CMMS system that dates to the mid-1990s belong in the smart factory of the future?
Before I address this issue, I would like to explain my role at Presenso. As a Professional Services Engineer, I work closely with industrial plants to deploy our AI-driven Industrial Analytics solution. We apply Artificial Intelligence and Machine Learning to sensor data generated by Industrial Machinery. Although access to historic Root Cause Analysis (RCA) data is not critical for our algorithms, Failed Component, Component Problem and Cause Code are helpful benchmarks for us to evaluate.
Here is the problem: In most cases, my client contacts are unable to extract consistent RCA data from the CMMS. These are challenges that are typically identified:
- CMMS often “compete” with other tools. Very often, RCA data is stored by technicians in an Excel spreadsheet but is not input into the CMMS.
- When RCA data is input, it is often mislabelled. Mandatory fields are used inconsistently, and business rules are not applied. I have seen anything ranging from a 2-word cryptic code to a mini-encyclopaedia. Neither of these can be used systemically.
- RCA data within the CMMS is not considered accurate and therefore not used operationally.
The result is that valuable information that could be used to predict the likelihood of machine failure is wasted. Not surprisingly, the culprits are obvious – either there are no consistent processes and rules for inputting RCA data or O&M employees are not trained in them.
Even without an Industrial Analytics solution, existing CMMS are simply not being used optimally. The remedy is not technical and does not require an investment in new software. First, the problem will not be addressed until plant management understands the importance of RCA and prioritizes it. Second, O&M employees must understand the benefit of inputting RCA information: how what is learned from one failure can be applied broadly within the plant (or plants) and used to improve overall OEE. Finally, change does not happen without formalizing the process, job performance metrics and ongoing training.
Ironically, the most forward-thinking industrial plants apply a completely non-technical approach to RCA by relying on human intelligence. This includes interviewing repair crews, photographing damage, reviewing previous incidence reports and making physical assessments (including a raw materials analysis). Although this level of inquiry is not always necessary or feasible, its application is indicative of a culture that prioritizes a systematic understanding of RCA.
In terms of whether existing CMMSs are compatible with advances in Machine Learning and Artificial Intelligence, I cannot provide a definitive answer. In the case of Presenso, we apply our algorithms to the sensor-generated data to detect patterns of anomalous behaviour. Based on this, we can trace the failure to the initial failure signals. With the right internal processes and employee training, the RCA data that our solution provides can be input into any CMMS.
Going back to the question of CMMS and Industrial IoT deployments, those industrial plants with legacy CMMSs that are decades old may wish to update them in parallel with other solution upgrades. Without advocating for or against upgrading a plant’s CMMS, there are best practices that can be implemented independent of Maintenance 4.0 adoption.