The Industrial Internet of Things (IIoT) is a key application of 5th Generation Mobile Communication Technology (5G), with latency-sensitive data transmission forming the foundation of IIoT. In this paper, we study the latency-sensitive data transmission in IIoT with the assistance of relay nodes based on the Non-Orthogonal Multiple Access-Wireless Power Transfer (NOMA-WPT) technology. Specifically, we aim at minimizing the transmission delay by jointly optimizing the uplink transmit power of IIoT devices and relays, the downlink charge power and the uplink bandwidth allocated of relays, the charging time fraction, the IIoT device-to-relay transmission time fraction, and the IIoT device-relay grouping factor, while guaranteeing the amount of data transmitted. To solve the formulated non-convex problem, we equivalently transform it into the problem of maximizing the minimum data transmission across IIoT devices in the framework of bisection searching. That is, the minimum transmission delay can be obtained by alternatively maximizing the minimum data transmission across IIoT devices during the time durations derived by the bisection searching. For the maximization of minimum data transmission, the Karush-Kuhn-Tucker (KKT) condition and Lagrangian function are first used to obtain the optimal solutions to uplink transmit power of IIoT devices and relays, downlink charge power and uplink bandwidth allocated of relays, charging time fraction, and IIoT device-to-relay transmission time fraction. The cross-entropy (CE) algorithm is then used to get the optimal solution of the IIoT device-relay grouping. Simulation results show that the proposed solution reduces the transmission delay by 30.77% compared to using dual-layer Frequency Division Multiple Access (FDMA) in the system.