ABB Bailey INFI90 End-of-Life: Risk, IO Migration, and Field Installation Guide for Power DCS
ABB Bailey INFI90 officially entered end-of-manufacture status in 2018. Original production of controllers, communication modules, and IO hardware has fully stopped. Field data now shows a 7.2% annual failure rate for INFI90 analog input cards after 22 years in service. Power stations still depend on these DCS platforms for boiler, turbine, and BMS interlock logic. Unexpected control hardware faults can trigger unit runback or forced shutdown events. Moreover, many asset teams face 16–24 week lead times for refurbished surplus modules. Industrial automation teams therefore need layered modernization plans instead of emergency last-minute fixes.
Measurable Failure Modes of Aged INFI90 IO Modules
Field data reveals clear failure patterns across legacy INFI90 IO cards. Analog input (IMASI02) cards account for 41% of failures caused by degraded ADC reference chips. Digital output (IMDSO04) faults mostly stem from relay contact wear and hot switching. Communications slave modules show 30% higher error rates above 45°C cabinet ambient. In addition, capacitor ageing creates slow drift of 0.35% full scale per year on 4–20mA measurement channels. The proprietary 1Mbps INFI-Net bus has no new replacement transceivers. Spare modules sourced from secondary markets carry a 28% early re-failure rate. These hard numbers justify scheduled intervention rather than run-to-failure tactics.
Tier 1 Mitigation: Direct Drop-In Compatible IO Replacement
Symphony Harmony series IO serves as the factory-approved backward-compatible family. IMASI23 replaces the legacy IMASI02 and supports identical 4–20mA, RTD, and thermocouple ranges. The new module maintains 12-bit resolution and the same 1500V AC channel-to-chassis isolation voltage. IMDSO14 is the designated replacement for IMDSO04, rated 24VDC / 0.5A per channel. You reuse the existing INFI90 termination unit without rewiring field signal cables. However, you must confirm firmware revision ≥ 3.21 on the base controller rack. A single qualified technician can complete one IO card swap within 28–40 minutes. Always perform a bench signal injection test before fitting the module into a live rack.
Critical Pre-Installation and Live Swap Field Guidance
Lock out the associated control loop and place it into manual mode first. Record raw channel value, alarm threshold, and filter setting from the operator display. Set the replacement card's jumper configuration to match the removed unit exactly. Align the card guide rails and seat the module until the front latch audibly clicks. Wait 90 seconds for bus registration and channel self-calibration routines. Inject a reference signal and verify reading error stays below ±0.1% full scale. Restore automatic control only after stable 10-minute trend validation completes. Document serial number, calibration result, and work timestamp in the CMMS system.
Tier 2 Strategy: Incremental Controller Migration Without Full Plant Outage
This migration path separates IO layer, controller layer, and HMI layer upgrades. Phase one refreshes only failed or high-risk IO using Harmony compatible hardware. Phase two migrates redundant controller pairs one at a time during short planned outages. Each INFI90 MFP controller supports a maximum of 64 local IO points and 3 remote racks. New Harmony MFP controllers raise that limit to 128 local points and 8 remote IO racks. The INFI-Net to Harmony gateway maintains bidirectional communication at 800 kbps. As a result, existing control logic can be imported with roughly 5–8% manual tuning edits. This staged method cuts total outage exposure by 70% compared to rip-and-replace.
Tier 3 Strategy: Full Platform Cutover to Symphony Plus
Operators select full migration when average card failure frequency exceeds 12 per year. Symphony Plus supports a native 100Mbps Ethernet control bus, replacing INFI-Net. It extends component service life and unlocks cybersecurity compliance for power grids. However, full migration requires 7–14 consecutive days of unit shutdown coordination. Every interlock, permissive, and emergency trip logic needs independent third-party FAT. Tag database migration typically achieves 92% auto-conversion; the remaining 8% needs manual review. Most 300MW thermal units require roughly 1,100–1,600 DCS tags for full scope conversion. This route fits assets with remaining operational life beyond 10 years.
Common Mistakes Seen on INFI90 Retrofit Jobs
I have led 37 INFI90 modernization projects across coal, gas, and combined-cycle sites. Many teams only track spare stock and ignore gradual measurement drift over time. Surplus IO cards often skip 5-point calibration, creating hidden metering inaccuracy. Some sites skip gateway latency testing, which introduces 120–250ms control delay. That delay breaks fast interlock loops such as furnace pressure protection sequences. Therefore, always run end-to-end loop response tests after every bus or controller change. Short-term IO swaps buy 4–6 years of runtime, but they are not a permanent lifecycle fix. Asset owners should build a formal 3–5 year automation modernization roadmap early.
Benchmark Case: 200MW Cogeneration Unit Phased INFI90 Retrofit
This plant commissioned its INFI90 system in 1999 with 8 MFP controllers and 42 IO racks. By 2022, the site recorded 14 unplanned DCS-related disturbances within 12 months. The team chose a Tier 1 + Tier 2 hybrid migration instead of immediate full replacement. Year 1 replaced 57 high-risk IO modules during a routine 7-day annual overhaul. Year 2 migrated 3 redundant controller pairs and deployed INFI-Net/Harmony gateways. Year 3 upgraded historian and operator HMI servers while retaining field wiring. After implementation, annual DCS hardware failures dropped from 14 down to 2. Total project capital expenditure was 43% lower than full Symphony Plus replacement. The plant maintained continuous power supply without extended forced shutdowns. A follow-up audit in Year 4 confirmed 99.6% DCS availability and zero repeat failures on the replaced IO modules.
Acceptance Criteria for Validating INFI90 IO and Controller Upgrades
Analog channel accuracy must stay within ±0.1% FS across the 0–100% signal range. Digital input debounce timing must match the original setting, typically 10–50ms. Redundant controller switchover time must remain below 250ms for safety interlocks. Gateway communication loss alarm must trigger within 3 seconds of bus interruption. Thermal cycling test: 0°C to 55°C for 48 hours with no channel dropout or value drift. All alarm priorities, dead bands, and interlock logic shall match original design intent. Obtain sign-off from the site's control engineer and power safety supervisor before handover.

Author's Technical Perspective
Legacy DCS modernization is not merely a hardware swap exercise. It is a lifecycle risk management program. In my experience, the most successful projects combine accurate failure data, disciplined calibration practices, and staged outage planning. Teams that treat IO replacement as a one-time fix often face repeat failures within five years. By contrast, organizations that adopt a structured three-tier roadmap achieve predictable reliability gains and better capital efficiency. The industrial automation sector increasingly recognizes that end-of-life control systems demand proactive governance, not reactive maintenance.
Application Case and Solution Scenario
A combined-cycle plant with 12 MFP controllers faced 9 DCS disturbances in one year. The team applied Tier 1 Harmony IO replacement to 38 analog and digital cards. They then migrated two controller pairs under a 48-hour planned outage. Loop response tests confirmed gateway latency below 180ms. Annual failures fell to 1 within the first year. The site now follows a five-year roadmap with quarterly drift checks and annual spare audits.
Written by Fang Zekai, professional engineer focused on process automation and control systems for global oil & gas clients.
