中子加速器高压电源在核材料长期辐照模拟实验中的安全运行机制
Accelerator-based irradiation of materials is used to reproduce the damage that neutrons produce in a reactor, and the reproduction depends on the ion energy, the flux and the total fluence. The high voltage supply sets the energy, and the stability of that energy over a campaign that lasts months is what makes the accumulated damage comparable between specimens.
Energy stability in this context means more than a short-term regulation figure. A drift of a tenth of a percent changes the penetration depth by a fraction of a percent, which changes the damage profile slightly, and over a long campaign the accumulated difference between two specimens can become significant. A supply with a stability of one part in ten thousand over a month, achieved by a temperature-controlled reference and a divider that is checked against a standard cell, is the practical basis for comparable specimens.
Beam current stability matters for the flux. The damage rate scales with the flux, and a flux that varies by ten percent produces a damage rate that varies by ten percent, which changes the required time to reach a given fluence. A current loop that holds the beam within a percent, with the measurement taken at a Faraday cup that is shielded from secondary electrons, is adequate when the accumulated fluence is computed from the integrated charge rather than from the elapsed time.
Safety in an accelerator facility is organised around the interlock chain rather than around the supply. The supply is one element in that chain: it must remove the high voltage when the interlock opens, and it must not restore it until the chain is reset deliberately. A design in which the supply's own logic can override the interlock is unacceptable, and a design in which the interlock only signals the control computer without a hardwired path to the supply is equally unacceptable. The hardwired path is what protects against a control fault.
Radiation-hardened design is another requirement. The components near the accelerator see neutron and gamma fields that degrade semiconductors, and the degradation appears as increased leakage and as logic errors. Locating the control electronics away from the beam line, using fibre for the control links and using radiation-tolerant devices where proximity is unavoidable, keeps the supply's behaviour predictable over a campaign.
Personnel protection relies on the dosimetry and on the access control, and the supply's role is to ensure that no high voltage exists when access is granted. A voltage-presence indication that is independent of the control system, wired to the door status panel, gives the access controller a trustworthy signal. Trusting a software flag alone has failed in facilities before, and the consequence is not a machine problem.
Records should include the energy, the integrated charge, the interlock events and the radiation field at the electronics. Those four quantities allow a specimen's history to be reconstructed, which is what makes the irradiation comparable and the experiment defensible in later analysis.
The accelerator front end deserves separate attention because the supply that feeds the ion source is the element most exposed to the machine's own faults. A source that arcs draws a current spike that can propagate into the supply if the cable is long and poorly terminated. Terminating the cable with a network matched to its characteristic impedance, and placing a spark gap at the source flange, limits the spike that reaches the supply. The gap also protects the source from a supply fault, which in a long campaign is as important as the reverse direction.
Redundancy is a practical measure in a facility that runs continuously for months. A supply that can be swapped in a shift without recalibration, by using a plug-in module that carries its own calibration data, keeps the campaign on schedule when a module fails. The calibration data carried by the module should be verified at installation against the facility reference, so the swap does not introduce a hidden energy error.
Monitoring of the high voltage divider is a maintenance item with a long lead time. A divider that has begun to drift gives an energy reading that is wrong in a way that no other instrument detects, because all the other measurements agree with the divider. Comparing the divider against an independent measurement at each scheduled shutdown, and recording the difference as a trend, catches the drift before it affects a specimen.
The control architecture should separate the protection functions from the regulation functions. Protection that shares a processor with regulation is protection that can be disabled by a regulation fault. Independent hardware comparators that trip a series contactor, with a separate test circuit that proves the comparator at each power-up, give a level of assurance that software cannot.
Documentation of the interlock chain, with the logic drawn as a single-line diagram and the test procedure for each element, is the item that makes the safety case auditable. In a facility where the machine is used for months with occasional changes of experiment, the documentation is what prevents a well-intentioned modification from defeating a protection that nobody remembers.
A final practical point is training. The behaviour of an accelerator supply under fault is not intuitive, and an operator who resets an interlock without understanding why it opened can defeat the protection. A short annual exercise in which the chain is tested with the machine at low energy teaches the behaviour better than any manual, and it is cheap compared with the consequences of a defeat.
Beyond the electrical detail, the safety case rests on the principle that the high voltage exists only when the machine is ready for it. Implementing that principle means the supply is the last element to be enabled and the first to be removed, and every interlock in the chain should be able to remove it without relying on any other element. That single rule, applied consistently over the life of the facility, is what keeps a long irradiation campaign both productive and uneventful.
The choice of energy for an irradiation campaign follows from the damage that the material is expected to accumulate. Too low an energy produces damage confined to a thin layer, which does not represent the bulk behaviour; too high an energy produces damage that is spread out and diluted. The range that gives a representative damage profile is narrow, and the supply must hold the energy within it for the whole campaign. That requirement is the reason the energy is specified as a window rather than as a nominal value, and why the calibration is repeated at each shutdown.
The dose rate must also be recorded, because the damage that develops depends on the rate as well as on the total fluence. At high rates the defects interact before they can migrate, and the resulting microstructure differs from that produced at low rates. Keeping the beam current within a band and recording the band with the specimen keeps the comparison honest, and it allows specimens from different campaigns to be compared only when the rates are comparable.
Cooling of the specimen is a thermal problem that intersects with the supply through the beam current. A high current heats the specimen, and the heating changes the defect mobility. A current that is regulated for the irradiation requirement rather than for the thermal requirement can overheat a thin specimen, and the resulting damage is not the damage that was intended. A specimen temperature measurement feeding a limit on the beam current resolves the conflict, and the limit should be part of the interlock chain rather than an operator instruction.
Vacuum quality influences the beam as much as the supply does. Residual gas produces charge exchange along the beam path, which creates a halo of lower energy particles that strike the specimen outside the intended area. Improving the vacuum reduces the halo, and monitoring the pressure with an interlock that removes the beam when the pressure rises above a threshold keeps the specimen from being irradiated by a beam whose composition has changed.
The campaign record should be written so that a specimen can be described years later without reference to the machine's memory. Energy, current, fluence, rate, temperature, vacuum and interlock events, in a file tied to the specimen identifier, is the minimum. Facilities that keep such records can answer questions about their own past work; facilities that do not repeat the work.
The engineering conclusion is that the supply in an irradiation facility is judged by the comparability of the specimens it produces, and comparability depends on stability, on calibration discipline and on the completeness of the records. A supply that is accurate but undocumented produces specimens that cannot be defended, and in long-duration work the defence is the product as much as the material.
Documentation that allows a specimen to be described without the machine is the final product of a long irradiation campaign.
A campaign that can be defended in this way is a campaign whose results can be used.

