PREDICTIVE MAINTENANCE OF INDUSTRIAL PRINTERS
Develop an AI-enabled reliability index for HP industrial printers, based on manufacturing parameters, field data and printer usage data, structured and unstructured.
Compared to consumer printing, commercial and industrial printing market has very different dynamics which result in different support expectations and requirements. The majority of commercial printing customers are PSPs (Print Service Providers) who sell printing and other value-adding services in the market. For them printers and related finishing equipment become key organizational assets (money-makers) and equipment uptime is a key performance indicator directly impacting the financial results.
Printer up-time is impacted by multiple factors including usage patterns, printer maintenance, support response time and product complexity/reliability. Maintenance profiles and routines vary a lot across printer installed base and are also impacted by operator training and customer internal processes. HP tries to educate customers on rigorous maintenance regimes but it is very hard to enforce it. One of the differenciator is the lifetime and the reliability of the machines within industrial environments.
HP Printing has started moving from a reactive support model to a proactive one, where printers “tell” service and support what is going on through telemetry data systems. In our journey we want to move ahead and start guessing and predicting issues and problems in the field, or even improve the effectiveness of the proactive support by giving the support teams key index that help them to focus on the printers that require more urgent attention.
Industrial pritners are complex systems with 1000+ parts and many moving & automatic precise systems that require a lot of engineering effort to make it reliable along time. It is almost impossible to have an index saying this prior to know the usage model and the workflows in which the customer fits the printer in.
Such a reliability index would have a major impact on the service associated to the printers. It would help prioritizing maintenance planning and the solution associated to the most probable ongoing failure. The objective of HP is to have such a predictive reliability index provided on a cloud platform to improve the service delivery, reduce downtime and improve customer satisfaction.
The desired outcome is an index characterizing printer reliability and where the most probable next failure can happen. This would enable each printer to be monitored and ranked across fleets depending on maintenance and support needs.
The main outputs for the challenge are two algorithms developed on Predix whose output is:
- Printer reliability index calculation (probablility of printer down in next 2-4 weeks).
- Printer next probable failure (what is the next more probable failure).
These outputs should be re-calculated regularly (daily/weekly) updating with the last days usage information and/or known failure modes.
The algorithms must be developed and ported on the Predix cloud. It will help securely connect the machines, data, and analytics to develop this predictive AI-enabled reliability index for HP industrial printers
The main use case will be embedding those algorithms in the Service Center solution that HP Latex industrial printers already have, and adding that prediction capabilities to the system so that partners and HP service can prioritize and maximize efficiency on how we maintain and support our customers’ printers. Initially this can be done manually in our proactive support teams, and later automate it in more sophisticated planning and prioritization tools that may evolve from the proof of concept.
The main users of the AI system are our proactive and predictive Service Center solutions team, which helps Support teams in prioritizing their work depending on this index.
Service Center solutions already helps regional Supportteams, and in planning their most relevant pending work in improving or maintaining the health of the HP installed base. For instance, when a printer fails, the system warns the engineer responsible of the printer. In some cases the system already anticipates actions to prevent failure. This index should help them prioritize their activity to ensure customers improve their uptime ratios.
Data sets that we will have:
- Top line tests results (pass/fail tests run at factory)
- Critical Printer Performance Parameter results (100+ parameters measured per printer in factory)
- Printer status timeline
- Printer usage (monthly ink use)
- Printer system errors (internal events warning about issues ongoing on the printer. Some of them related to real failures other just informative).
- Service & support events manual written feedback
We will help understanding the data sets and other questions with factory related experts, product experts and data mining experts. We are also working in additional data science challenges in our team, evolving our current solutions to even better customer focused ones that enable our customers succeed in their business by using our printing solutions.