Machine learning and AI to predict the failure of medium voltage (mv) pump sets
- Customer
- Rand Water
- Project manager on the customer side
-
- IT Provider
- Samotics B.V
- Year of project completion
- 2025
- Project timeline
- January, 2024 - February, 2025
- Project scope
- 5 automated workstations
- Goals
-
The aim of the project was determine condition monitoring, based on Electrical Signature Analysis (ESA), will be an operationally and financially sound investment to improve Rand Water’s approach to Electro-Mechanical Asset Reliability, Maintenance and Performance?
The objectives were to:
1. Obtain reliable, real-time condition monitoring and performance data of electro-mechanical assets (pumps);2. Thoroughly understand the chosen solution and all its functionalities;3. Prevent unplanned electrical and mechanical failures and avoid water service disruption –essentially an early warning system; and4. Obtain a financial analysis or quantify the benefits of such a solution.
The overall objective of the project was to assess whether the Generative AI solution (SAM4) linked to data from sensors can support Rand Water in changing their maintenance approach from a reactive to a proactive one, through early detection of upcoming asset failures and operational inefficiencies, thereby reducing costs of lost production, temporary mitigation infrastructure and intervention, additional personnel costs for call outs as a result of asset failure, etc.
- Project Results
-
Figure 1: Recorded runtime for all 4 pumps throughout pilot duration
Electrical health indicators - The following indicators for electrical asset health have been tracked over time throughout the pilot:
● Current unbalance may cause excessive heating, reduced efficiency, torque pulsations, increased mechanical stress, and potential overload trips, leading to higher energy consumption and premature equipment failure; Current unbalance for all 4 assets has been found relatively stable and well within range of what is generally considered healthy (<10%)
● Voltage unbalance may cause excessive current unbalance, overheating, reduced efficiency, torque pulsations, mechanical stress, and potential system failures, leading to increased energy losses and premature motor and equipment damage; Voltage unbalance for all 4 assets has been found stable and well within range of what is generally considered healthy (<1%)
● Total Harmonic Distortion (THD), caused by nonlinear loads, leads to excessive heating, reduced efficiency, torque fluctuations, nuisance tripping, resonance effects, and bearing damage in electric motors and pumps, increasing energy losses and the risk of premature failure.
○ Current THD for all 4 assets has been found stable and well within range of what is generally considered healthy (<10%); Pump Set 25 shows both a slightly higher level of THD as well as a higher variability of THD, but still well within healthy levels.○ Voltage THD for all 4 assets has been found stable and well within range of what is generally considered healthy (<3%).
The uniqueness of the project
Electrical Signature Analysis (ESA) is a non-intrusive condition monitoring technology that analyses the electrical signals of motors and driven equipment to detect faults. It works by capturing voltage and current waveforms and using advanced algorithms to identify anomalies related to mechanical and electrical issues. ESA is valuable for industries relying on motors, improving reliability, efficiency, and maintenance planning. Uses for Condition Monitoring by ESA:- Detection of Electrical Faults – Identifies issues like insulation degradation, rotor bar defects, and power supply imbalances.
- Mechanical Fault Identification – Detects bearing failures, misalignment, and unbalanced loads.
- Early Fault Diagnosis – Provides predictive maintenance insights, reducing unplanned downtime.
- Efficiency Monitoring – Helps optimize energy consumption and system performance.
Samotics applies Electrical Signature Analysis in a unique and unrivalled manner, by offering an end-to-end service through SAM4:
SAM4 has an excellent track record for detection of both electrical and mechanical failures across the full asset drivetrain. SAM4 works particularly well on pumps, given the relatively short and simple drivetrain.
Samotics manages the end-to-end data infrastructure and data flows, according to the following high-level infrastructure:
- Used software
-
SAMOTICS PLATFORM
All SAM4 insights and operational metrics are presented in Samotics’ platform. Any detection of anomalous asset behaviour or inefficient operation is communicated both by email and in the SAM4 platform. All historic and open incidents can be viewed at any time through the ‘Incidents’ section.
Figure 1: Example of SAM4 Detection Notification of mechanical unbalance.
Figure 2: Example of tabular overview of outstanding and close SAM4 detections
Each individual detection has a separate ‘Incident page’ in the platform, including an interactive timeline of the incident, overview of the drivetrain health indicating which drivetrain components are impacted and an overview of likely root-causes and failure modes. Customers can reply to notifications either by email or in the incident timeline itself. In addition, they can report maintenance activities, which are subsequently added to that timeline. The data underlying the alert is presented as ‘supporting evidence’ and is automatically updated.Figure 3: Example of an Incident Timeline for a detected current phase loss
Figure 4: Impacted drivetrain components and potential root causes and failures modes
Figure 5: 3-phase Current RMS as underlying data for the alert of a phase loss
- Difficulty of implementation
-
Challenges:
- Availability of asset information: Relevant information should be available for all assets monitored by SAM4, which includes motor nameplate information, pump curve documentation, relay readings of current / power to validate SAM4 calculations
- Connectivity: Currently SAM4 is a cloud-based solution only, which requires a relatively stable 4G/5G connection; In remote locations with limited signal strength, the use of high-quality antennas or a Starlink connection has proven successful
- Site access and ability to shut down asset for installation: In order to efficiently install SAM4 hardware, site access needs to be organized to the local contractor; In addition, a Rand Water operator needs to be onsite and available during installation in order to shut down and restart the equipment
- Availability of resources to act upon SAM4 notifications: To enable successful implementation of SAM4 technology, Rand Water, together with the OEM and local service provider, will need to ensure clear processes are designed and roles and responsibilities are assigned for the adequate follow-up on SAM4 notifications. This may require some management of change in existing processes and responsibilities among Reliability & Maintenance and Operational Teams
Limitations & exclusions:
- SAM4 cannot monitor: DC or Servo motors, 2-pole synchronous motors or generators
- SAM4’s detection capabilities can be limited by: High noise levels introduced by the grid or by certain types and settings of Variable Speed Drives may mask the motor’s rotational frequency component in sampled data; As a consequence, trends at the rotational frequency may only become visible at a very late stage or not at all
- SAM4 cannot detect partial discharges: SAM4 samples data in the kHz range and partial discharges only present themselves in the MHz range
- Project Description
-
Rand Water is a water board that was established and is governed in terms of Chapter 6 of the Water Services Act, No. 108 of 1997. Rand Water is listed as a National Government Business Enterprise in Schedule 3b of the Public Finance Management Act No.1 of 1999. The Government of the Republic of South Africa, through the Department of Water and Sanitation (“the Executive Authority”), duly represented by the Minister of Water and Sanitation, is the sole Shareholder of Rand Water. Since 1903, Rand Water has been playing an integral role in the development and growth of Gauteng, its cities, and its surrounding regions by providing water to people and industries.
Our mandate, as the largest bulk water utility in Africa and one of the largest in the world, is to provide bulk potable water to mainly municipalities who then supply water to more than 18 million people in Gauteng, parts of Mpumalanga, Free State and North West provinces - an area that stretches over 18 000km2. For 121 years, Rand Water has supplied bulk potable water to Gauteng and its surrounding areas. Rand Water’s success is based on sharing in the pioneering spirit that led to the growth of the City of Johannesburg. It is the same spirit that has driven Rand Water’s growth in terms of infrastructure and water quality, creating a reputation for supplying world-class quality water that is of international standards. During the year under review, Rand Water continued with its mandate of water and sanitation
services provision in its area of operations. Most of Rand Water’s core value chain involves purchasing and distributing potable water from the Integrated Vaal River System (IVRS). Rand Water obtains raw water from the Department of Water and Sanitation (DWS). For decades,
DWS has been developing and implementing various water transfer schemes, sourcing raw water from other countries, provinces and catchments with more abundant resources and delivering it into the IVRS. The latest DWS development project is the Lesotho Highlands Phase 2, which should be commissioned by 2028. The customer base served by Rand Water includes metropolitan municipalities, local municipalities, mines, and industries. We have a global reputation for providing water of high quality that ranks among the best in the world. We have consistently met and exceeded national and international standards on water quality.
Rand Water has piloted a SaaS solution for condition monitoring of electric motors and rotating equipment (e.g., pumps) called SAM4 on 2 sets of 2 medium voltage electric motor + pump combinations in the Rand Water Zuikerbosch Pump Station. The technology is provided by Samotics B.V., based in Leiden, The Netherlands and is based on Electrical Signature Analysis. SAM4 acquires current and voltage data through sensors installed in the motor control cabinets, applies various computations to the raw data and analyses outcomes in a cloud environment. AI models that run in that cloud, trigger alerts when anomalous behaviour or operational inefficiencies are detected. Samotics’ Asset Reliability Specialists diagnose those alerts and translate them into actionable maintenance advice for Rand Water’s technical teams (both electrical and mechanical) and operators. The hardware required for acquiring the current and voltage data is provided by Samotics.
The objective of the pilot was to assess whether SAM4 can support Rand Water in changing their maintenance approach from a reactive to a proactive one, through early detection of upcoming asset failures and operational inefficiencies, thereby reducing costs of lost production, temporary mitigation infrastructure and intervention, additional personnel costs for call outs as a result of asset failure, etc. Since the monitored pump sets are well maintained and operated near its designed best efficiency point, no asset failures were expected during the pilot. Therefore, a test was executed in which cavitation was intentionally introduced in the Stage 1 pump of one of the pumpsets. SAM4 successfully picked up on this cavitation. Additionally, SAM4 has proven to be able to monitor key indicators for electrical and mechanical faults in real-time. Their online platform offers insights into the real-time operation of the monitored assets, including operational metrics such as current, power, head and flow. In addition, the functioning of the platform in case of an incident was tested and found clear and actionable for technical & operational teams. The platform allows for easy communication between stakeholders at a time of an incident to keep all involved up to date.
Concrete results include: Clear identification of cavitation during the test performed; Identification of more motor starts than expected: Motor starts can be particularly harmful to Medium Voltage motors’ insulation, due to the high inrush currents during start up; Increasing noise in current & voltage signal coming from the grid on all assets throughout the pilot duration and Identification of operation with voltage levels below 11kV and the duration of those inefficient operations: Any operation below 11kV introduces energy losses.
The main benefits of using the SAM4 technology for Rand will include:
● Reduction of ‘lost water production’ as a result of asset failure
● Reduction of costs for mitigation infrastructure during downtime / asset failure
● Reduction of personnel costs for call-outs during asset failure● Reduction of reactive maintenance costsExpected benefits could add up to as estimated ZAR 65M per year and a full scale roll-out (354 pumps) has a potentially expected ROI of 200%. Although the full scale roll-out will require diligent preparation, no major challenges have been identified. The current Rand Water staff (Electrical Asset Management, Mechanical Asset Management and Plant Operators) have the knowledge, skills and capacity to process the output of the SAM4 system at scale.
Data flows and enrichment
● The raw current and voltage signal produced by the sensors is extremely bulky and useless by itself without any further processing and selection
● Only after sampling and calculations, data is fed into the SAM4 detection models in the cloud for failure detection and diagnosis
● All notifications can be shared both in the SAM4 platform and by email
● In parallel, operational metrics data (e.g.; current, active power, flow & head) is processed and provided in real-time in the SAM4 platform
● Various APIs are available to integrate SAM4 insights and notifications in existing systems
● All (raw) data can be shared with Rand Water. The project was successful in demonstrating the value of the SAM4 condition monitoring solution as an early warning system of electromechanical asset health: Installation of the SAM4 solution will enhance reliability of Rand Water’s water supply.•Given high costs of Rand Water’s current reactive maintenance approach to critical pumping infrastructure and the capabilities of the SAM4 ESA-based condition monitoring system, investment in a full-scale roll out SAM4 across all Rand Water’s pumps is a sound one from both a financial and operational perspective.•An estimated total all-in cost for deployment is approximately ZAR 62M (at current exchange rate), with an expected ROI full scale roll out: ~200% over a period of 3 years
Data flows and enrichment ● The raw current and voltage signal produced by the sensors is extremely bulky and useless by itself without any further processing and selection ● Only after sampling and calculations, data is fed into the SAM4 detection models in the cloud for failure detection and diagnosis ● All notifications can be shared both in the SAM4 platform and by email ● In parallel, operational metrics data (e.g.; current, active power, flow & head) is processed and provided in real-time in the SAM4 platform ● Various APIs are available to integrate SAM4 insights and notifications in existing systems ● All (raw) data can be shared with Rand Water. The project was successful in demonstrating the value of the SAM4 condition monitoring solution as an early warning system of electromechanical asset health: Installation of the SAM4 solution will enhance reliability of Rand Water’s water supply.•Given high costs of Rand Water’s current reactive maintenance approach to critical pumping infrastructure and the capabilities of the SAM4 ESA-based condition monitoring system, investment in a full-scale roll out SAM4 across all Rand Water’s pumps is a sound one from both a financial and operational perspective.•An estimated total all-in cost for deployment is approximately ZAR 62M (at current exchange rate), with an expected ROI full scale roll out: ~200% over a period of 3 years
- Project geography
-
South Africa